CN105139388A - Method and apparatus for building facade damage detection in oblique aerial image - Google Patents
Method and apparatus for building facade damage detection in oblique aerial image Download PDFInfo
- Publication number
- CN105139388A CN105139388A CN201510494876.0A CN201510494876A CN105139388A CN 105139388 A CN105139388 A CN 105139388A CN 201510494876 A CN201510494876 A CN 201510494876A CN 105139388 A CN105139388 A CN 105139388A
- Authority
- CN
- China
- Prior art keywords
- buildings facade
- facade
- buildings
- door
- window
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20021—Dividing image into blocks, subimages or windows
Abstract
The present invention discloses a method for building facade damage detection in an oblique aerial image. The method comprises the steps of: 1, using a k-means clustering algorithm based on a rough set theory to divide a building facade to obtain doors and windows of the building facade; 2, using a canny algorithm to perform edge detection on the doors and windows of the building facade to obtain edge features of the doors and windows; 3, using a Gini coefficient in economics to perform statistics on the edge features to obtain a Gini coefficient of the building facade; and 4, determining whether the building facade is broken according to the Gini coefficient. The method for building facade damage detection in the oblique aerial image is capable of easily and efficiently performing damage detection on the building facade without the need for prior information and pre-disaster data, thereby lowering the complexity of the method and reducing the production cost. The Gini coefficient in economics is introduced as an index for building facade damage detection, and structural features of the building facade can be made full use of to determine damage, thereby providing a solution for improving the precision and automation of building damage evaluation.
Description
Technical field
The present invention relates to remote sensing image applied technical field, especially relate to the method and apparatus that in a kind of oblique aerial image, the damage of buildings facade detects.
Background technology
Disaster makes the life of the mankind and property take a bath for a long time, is the huge obstacle of human survival and development.Remote sensing technology has that revisiting period is short, investigative range is large, aggregation of data high, for disaster monitoring and assessment provide a kind of favourable means.Along with the development of various monitoring means and new and high technology, traditional disaster test and assessment is progressively by the future development of qualitative statistical estimation to quantitative meticulous assessment.Buildings, as the key element of people's productive life, has great importance to the detection of its damage information and extraction after disaster occurs, and after it can be calamity emergency response and calamity, restoration and reconstruction provide important decision foundation.In view of the complicacy that buildings damage detects, not only to judge the change of the information such as buildings elevation and area, also to judge buildings end face and facade damage information, therefore how the focus that omnibearing high-precision quantitative Damage assessment is research at present be carried out to buildings.
The traditional aviation image of the oblique aerial photography technological breakthrough of fast development in recent years can only from the limitation of vertical angle shooting, by carrying multiple stage sensor simultaneously from vertical, a different angle acquisition image such as three or more tilts on same flying platform, the multi-faceted information of buildings can be gathered very well.Many scholars have done a large amount of deep research work utilizing oblique photograph measuring technique in the complete information extracting buildings, obviously utilize this technology to detect the deficiency that can make up traditional detection method to buildings damage.Buildings Top-print information due to texture information fairly simple, utilize it carry out damage detect research more, also achieve good achievement; But buildings facade information texture information is very abundant, what building window and damage occurred collapses, have certain interference between crack and breakage mutually, and this detects the damage of buildings facade and causes very large difficulty.
Therefore, how to carry out damaging to the buildings facade information in oblique aerial image the accurate quantification assessment detected for improving buildings damage to have great importance.
A large amount of abundant information is contained in high resolving power oblique aerial remote sensing image, particularly the texture information of buildings facade is very abundant, and current Chinese scholars utilizes oblique aerial image to carry out damaging the typical method detected to buildings facade and is included in following two aspects: the buildings facade damage 1) based on Mono temporal detects.Because calamity rear-inclined aviation image is easy to obtain, the therefore more realistic Production requirement of these class methods.These class methods are divided into again the damage of structure based information and textural characteristics to detect, mainly utilize the feature such as gray level co-occurrence matrixes or Tamura to carry out damage based on textural characteristics to detect, but because buildings facade textures roughness is larger, and textural characteristics detects the detection being mainly intended for close grain feature, be therefore not suitable for the damage distinguishing Mono temporal; Structure based information approach mainly extracts the information such as buildings facade crack to detect damage, but can extract and cause very big interference by fracture due to the abundant structural information of buildings facade, therefore these class methods also comparatively difficulty.2) the buildings facade damage based on multidate detects.The method that these class methods mainly detect based on change is carried out the damage of buildings facade and is detected, but because the oblique aerial image before calamity is generally difficult to obtain, the inclination image particularly closing on the disaster-stricken front time period is difficult to obtain more, and before calamity, after calamity, how the obliquity effects of multidate carries out high registration accuracy is also current difficult point.Therefore, need a kind of data of urgent searching easily to obtain, judge that automaticity is high, extraction result is relatively accurately high and the damage detection method of realistic need of production.
Summary of the invention
The object of the invention is to propose the method and apparatus that in a kind of oblique aerial image, the damage of buildings facade detects, the present invention takes full advantage of the structural information of buildings facade in inclination image, Gini coefficient simultaneously in integrated economics is as damage index, significantly improve the precision that the damage of buildings facade detects, feature is:
(1) the method has when not needing data before prior imformation and calamity, can carry out the damage of buildings facade simply efficiently and detect, reduce the complexity of method, also save production cost.
(2) index that the Gini coefficient in economics detects as the damage of buildings facade is introduced, the architectural feature that can make full use of buildings facade judges damage, not only improve automaticity and the precision of judgement, and the needs of realistic production.
For reaching this object, the present invention by the following technical solutions:
The method that in oblique aerial image, the damage of buildings facade detects, comprising:
Step one, utilizes the k-means clustering algorithm based on rough set theory to split buildings facade, obtains the door and window of buildings facade;
Step 2, utilizes canny algorithm to carry out rim detection to the door and window of buildings facade, obtains the edge feature of door and window;
Step 3, utilizes the Gini coefficient in economics to add up described edge feature, obtains the Gini coefficient of buildings facade;
According to described Gini coefficient, step 4, judges whether buildings facade is damaged.
Wherein, described step one, utilizes the k-means clustering algorithm based on rough set theory to split buildings facade, obtains the door and window of buildings facade, be specially:
In S110, image, the gray-scale value of pixel is f, wherein f=0,1,2 ... 255, k the central point utilizing rough set theory to obtain is as preliminary classification average μ
1, μ
2, μ
3..., μ
k;
Distance D between S120, the gray-scale value f calculating each pixel in image and previous step preliminary classification average μ, composes the initial classes average class nearest apart from it, namely by each pixel
D|f
p-μ
k|=min{D|f
p-μ
i|,(i=1,2,…k)}(1)
Carry out iteration to (1) formula, wherein p is the central point in iterative process;
S130, for i=1,2 ..., k calculates new cluster centre, upgrades class average:
in formula, N
ibe
in number of pixels, m is iterations;
S140, all pixels to be investigated one by one, if i=1,2 ... k, has μ
i (m+1)=μ
i (m), then algorithm convergence, terminates, otherwise returns S120 continuation next iteration.
Wherein, described step 2, utilizes canny algorithm to carry out rim detection to the door and window of buildings facade, obtains the edge feature of door and window, is specially:
S210, utilize canny algorithm to carry out rim detection to the door and window of buildings facade, obtain the door and window edge of buildings facade;
S220, be all perpendicular to ground due to most of buildings facade, first statistics is parallel to the range distribution between the parallel lines on ground, then calculates the histogram of distance, finally obtains the edge feature of door and window; Flow process is as follows:
A) because buildings facade may be damaged, therefore the outline line that door and window rim detection obtains may not be parallel to each other, the method of therefore adding up the range distribution employing between the parallel lines being parallel to ground is: add up to vertical direction buildings facade along horizontal direction every a fixed step size, calculate vertical direction and close on distance between two pixels, be designated as d
i, whole facade image can obtain distance vector d=[d
1, d
2, d
3.., d
k];
B) formula statistics with histogram function D (d is utilized
i)=n
icount distance vector histogram, then to histogram variable n
icarry out ascending sort, obtain vector n=[n
1, n
2, n
3.., n
k], wherein n
1≤ n
2≤ n
2≤ ...≤n
k; Vector n is the edge feature of buildings facade door and window.
Wherein, described step 3, utilizes the Gini coefficient in economics to add up described edge feature, obtains the Gini coefficient of buildings facade, is specially:
S310, supposing that the edge feature of the door and window extracted in image is f, is histogram f=[f by the distribution statistics of f
1, f
2, f
3..., f
k], the element in histogram is sorted from small to large, obtain new histogram set be f '=[f '
1, f '
2, f '
3..., f '
k], the Gini coefficient formula so measuring image rule degree is:
Wherein, || f||
1for first normal form, K is the classification sum of statistics with histogram, and the scope of G is from 0 to 1, and G is larger, and buildings facade is more complete, and G is less, and the damage of buildings facade is serious; Find through statistical experiment, buildings damage threshold value is 0.45, and statistical law in economics is basically identical;
S320, the statistics edge feature vector n in step 2 to be brought in (2) formula as f, obtain the Gini coefficient of elevation of building.
Wherein, according to described Gini coefficient, described step 4, judges whether buildings facade is damaged, and is specially:
When Gini coefficient G is greater than 0.45, represent that buildings facade is intact; Otherwise, when Gini coefficient G is less than 0.45, represent that elevation of building there occurs damage.
The device that in oblique aerial image, the damage of buildings facade detects, comprising:
Buildings facade cutting unit, for utilizing the k-means clustering algorithm based on rough set theory to split buildings facade, obtains the door and window of buildings facade;
Door and window edge feature calculation unit, for utilizing canny algorithm to carry out rim detection to the door and window of buildings facade, obtains the edge feature of door and window;
Gini coefficient computing unit, for utilizing the Gini coefficient in economics to add up described edge feature, obtains the Gini coefficient of buildings facade;
According to described Gini coefficient, damage judging unit, for judging whether buildings facade is damaged.
Wherein, described buildings facade cutting unit is f specifically for the gray-scale value of pixel in a, image, wherein f=0,1,2 ... 255, k the central point utilizing rough set theory to obtain is as preliminary classification average μ
1, μ
2, μ
3..., μ
k; Distance D between b, the gray-scale value f calculating each pixel in image and described preliminary classification average μ, composes the initial classes average class nearest apart from it, namely by each pixel
D|f
p-μ
k|=min{D|f
p-μ
i|,(i=1,2,…k)}(1)
Carry out iteration to (1) formula, wherein p is the central point in iterative process; C, for i=1,2 ..., k calculates new cluster centre, upgrades class average:
in formula, N
ibe
in number of pixels, m is iterations; D, all pixels to be investigated one by one, if i=1,2 ... k, has μ
i (m+1)=μ
i (m), then algorithm convergence, terminates, otherwise returns b continuation next iteration.
Wherein, described door and window edge feature calculation unit, specifically for a, utilizes canny algorithm to carry out rim detection to the door and window of buildings facade, obtains the door and window edge of buildings facade; B, be all perpendicular to ground due to most of buildings facade, first statistics is parallel to the range distribution between the parallel lines on ground, then calculates the histogram of distance, finally obtains the edge feature of door and window; Flow process is as follows: b10, may damage due to buildings facade, therefore the outline line that door and window rim detection obtains may not be parallel to each other, the method of therefore adding up the range distribution employing between the parallel lines being parallel to ground is: add up to vertical direction buildings facade along horizontal direction every a fixed step size, calculate vertical direction and close on distance between two pixels, be designated as d
i, whole facade image can obtain distance vector d=[d
1, d
2, d
3.., d
k]; B20, utilize formula statistics with histogram function D (d
i)=n
icount distance vector histogram, then to histogram variable n
icarry out ascending sort, obtain vector n=[n
1, n
2, n
3.., n
k], wherein n
1≤ n
2≤ n
2≤ ...≤n
k; Vector n is the edge feature of buildings facade door and window.
Wherein, described Gini coefficient computing unit, specifically for a, supposing that the edge feature of the door and window extracted in image is f, is histogram f=[f by the distribution statistics of f
1, f
2, f
3..., f
k], the element in histogram is sorted from small to large, obtain new histogram set be f '=[f '
1, f '
2, f '
3..., f '
k], the Gini coefficient formula so measuring image rule degree is:
Wherein, || f||
1for first normal form, K is the classification sum of statistics with histogram, and the scope of G is from 0 to 1, and G is larger, and buildings facade is more complete, and G is less, and the damage of buildings facade is serious; Find through statistical experiment, buildings damage threshold value is 0.45, and statistical law in economics is basically identical; B, the statistics edge feature vector n in step 2 to be brought in (2) formula as f, obtain the Gini coefficient of elevation of building.
Wherein, described damage judging unit, specifically for when Gini coefficient G is greater than 0.45, represents that buildings facade is intact; Otherwise, when Gini coefficient G is less than 0.45, represent that elevation of building there occurs damage.
Beneficial effect:
The method that in a kind of oblique aerial image of the present invention, the damage of buildings facade detects, comprising: step one, utilizes the k-means clustering algorithm based on rough set theory to split buildings facade, obtains the door and window of buildings facade; Step 2, utilizes canny algorithm to carry out rim detection to the door and window of buildings facade, obtains the edge feature of door and window; Step 3, utilizes the Gini coefficient in economics to add up described edge feature, obtains the Gini coefficient of buildings facade; According to described Gini coefficient, step 4, judges whether buildings facade is damaged.The present invention proposes in a kind of aviation inclination image and utilize Gini coefficient to detect the method for buildings facade damage, the method first utilizes the k-means algorithm of improvement namely the door and window of buildings facade and metope to be split based on the k-means clustering algorithm of rough set theory, obtains the door and window of buildings; Then use the facade door and window of canny algorithm to buildings to carry out rim detection, obtain the edge feature of building window; Finally make use of Gini coefficient in economics as damage index overall tolerance buildings facade damage information, thus reach and judge whether buildings facade is damaged.Visible, the present invention takes full advantage of the structural information of buildings facade in inclination image, Gini coefficient simultaneously in integrated economics is as damage index, significantly improve the precision that the damage of buildings facade detects, feature is: 1) the method has when not needing data before prior imformation and calamity, the damage of buildings facade can be carried out simply efficiently detect, reduce the complexity of method, also save production cost.2) introduce the index that the Gini coefficient in economics detects as the damage of buildings facade, the architectural feature that can make full use of buildings facade judges damage, not only improves automaticity and the precision of judgement, and the needs of realistic production.
Accompanying drawing explanation
The process flow diagram of the method that the damage of buildings facade detects in a kind of oblique aerial image that Fig. 1 provides for the specific embodiment of the invention.
The structural representation of the device that the damage of buildings facade detects in a kind of oblique aerial image that Fig. 2 provides for the specific embodiment of the invention.
Fig. 3 is the range distribution statistics schematic diagram between buildings facade parallel lines.
Embodiment
Should be appreciated that specific embodiment described herein only in order to explain the present invention, be not intended to limit the present invention.Technical solution of the present invention is described in detail below in conjunction with drawings and Examples.
Embodiment 1
The process flow diagram of the method that the damage of buildings facade detects in a kind of oblique aerial image that Fig. 1 provides for the specific embodiment of the invention.As shown in Figure 1, the method that in a kind of oblique aerial image of the present invention, the damage of buildings facade detects, comprising:
Step one, utilizes the k-means clustering algorithm based on rough set theory to split buildings facade, obtains the door and window of buildings facade;
Step 2, utilizes canny algorithm to carry out rim detection to the door and window of buildings facade, obtains the edge feature of door and window;
Step 3, utilizes the Gini coefficient in economics to add up described edge feature, obtains the Gini coefficient of buildings facade;
According to described Gini coefficient, step 4, judges whether buildings facade is damaged.
The present invention proposes in a kind of aviation inclination image and utilize Gini coefficient to detect the method for buildings facade damage, the method first utilizes the k-means algorithm of improvement namely the door and window of buildings facade and metope to be split based on the k-means clustering algorithm of rough set theory, obtains the door and window of buildings; Then use the facade door and window of canny algorithm to buildings to carry out rim detection, obtain the edge feature of building window; Finally make use of Gini coefficient in economics as damage index overall tolerance buildings facade damage information, thus reach and judge whether buildings facade is damaged.Visible, the present invention takes full advantage of the structural information of buildings facade in inclination image, Gini coefficient simultaneously in integrated economics is as damage index, significantly improve the precision that the damage of buildings facade detects, feature is: 1) the method has when not needing data before prior imformation and calamity, the damage of buildings facade can be carried out simply efficiently detect, reduce the complexity of method, also save production cost.2) introduce the index that the Gini coefficient in economics detects as the damage of buildings facade, the architectural feature that can make full use of buildings facade judges damage, not only improves automaticity and the precision of judgement, and the needs of realistic production.
In this programme, described step one, utilizes the k-means clustering algorithm based on rough set theory to split buildings facade, obtains the door and window of buildings facade, be specially:
In S110, image, the gray-scale value of pixel is f, wherein f=0,1,2 ... 255, k the central point utilizing rough set theory to obtain is as preliminary classification average μ
1, μ
2, μ
3..., μ
k;
Distance D between S120, the gray-scale value f calculating each pixel in image and previous step preliminary classification average μ, composes the initial classes average class nearest apart from it, namely by each pixel
D|f
p-μ
k|=min{D|f
p-μ
i|,(i=1,2,…k)}(1)
Carry out iteration to (1) formula, wherein p is the central point in iterative process;
S130, for i=1,2 ..., k calculates new cluster centre, upgrades class average:
in formula, N
ibe
in number of pixels, m is iterations;
S140, all pixels to be investigated one by one, if i=1,2 ... k, has μ
i (m+1)=μ
i (m), then algorithm convergence, terminates, otherwise returns S120 continuation next iteration.
Because the buildings facade notable feature in oblique aerial image is dispersed with neat door and window, therefore buildings vertical door window edge feature can be extracted as gini index distribution statistics feature, by differentiating whether buildings vertical door window edge neatly arranges, and judges whether buildings facade is damaged with this.In order to carry out the extraction of Gini coefficient statistical nature, first k-means clustering algorithm is utilized to be partitioned into the door and window of buildings facade herein, k-means clustering algorithm is quick to the one of feature space division, easy sorting technique, can dynamic clustering, there is adaptive advantage, belong to the category of not supervised classification, but the impact that k-means clustering algorithm is chosen by initial cluster center, if the unreasonable complexity that will increase computing chosen, mislead cluster process, therefore directly utilize k-means clustering algorithm to split buildings facade and will obtain irrational cluster result.Because rough set can when keeping classification capacity constant, by knowledge abbreviation, carry out good approximate classify, therefore the space-division method herein by rough set theory carries out preliminary classification to image, then on the basis of preliminary classification, utilize k-means clustering algorithm that buildings facade is divided into door and window and wall two parts, specific algorithm flow process is as follows:
Rough set is utilized to carry out preliminary classification.According to rough set theory, can using the information expressed by a width image as a knowledge system K=(I, R), I represents image, R is defined in the relation of equivalence in image I, utilize define initial center point and the number thereof that relation of equivalence R marks off cluster.If the gray-scale value of pixel is f, wherein f=0 in image I, 1,2 ..., 255, D (f)=n represents that statistics gray-scale value is the number of pixels of f.The image histogram be made up of D (f) is generally Gu Feng distribution, the pixel that gray-scale value is approximate can be classified as a class by histogram, so image just can be similar to and be divided into some classes, therefore the gray value differences of pixel is defined as conditional attribute, so rough set relation of equivalence R can be defined as: if two pixel grey scale value differences are less than spacing d, then two pixels are relevant, belong to equivalence class, that is:
R={f||f
i-f
j<d}(i,j=0,1,...,255)
First determine gray value differences d, obtain number of greyscale levels L by the grey level histogram scope of image.Gray-scale values maximum for respective pixel number in tonal range is defined as central point μ.Calculate L central point spacing between two, if minor increment is less than spacing d, then respective center point is merged, and using the value of the arithmetic mean of 2 as this central point.Repeat until the spacing between two of all central points is all greater than spacing d, so the number of central point and numerical value are exactly number and the average of initial classes required for k-means cluster.If the k utilizing rough set theory a to obtain central point is as preliminary classification average μ
1, μ
2, μ
3..., μ
k.
Therefore, this method adopts the k-means clustering algorithm after improving namely based on the k-means clustering algorithm of rough set theory, buildings facade to be divided into door and window and wall two parts.First, obtain K central mean of initially birdsing of the same feather flock together by rough set theory, then do cluster segmentation, mainly metope is divided into door and window and wall two class.After above cluster process terminates, in order to strengthen display effect, each pixel of segmentation result is using cluster centre gray-scale value as such final gray scale.
In this programme, described step 2, utilizes canny algorithm to carry out rim detection to the door and window of buildings facade, obtains the edge feature of door and window, is specially:
S210, utilize canny algorithm to carry out rim detection to the door and window of buildings facade, obtain the door and window edge of buildings facade;
S220, be all perpendicular to ground due to most of buildings facade, first statistics is parallel to the range distribution between the parallel lines on ground, then calculates the histogram of distance, finally obtains the edge feature of door and window; Flow process is as follows:
A) because buildings facade may be damaged, therefore the outline line that door and window rim detection obtains may not be parallel to each other, the method of therefore adding up the range distribution employing between the parallel lines being parallel to ground is: as shown in Figure 3, buildings facade is added up to vertical direction along horizontal direction every a fixed step size, calculate vertical direction and close on distance between two pixels, be designated as d
i, whole facade image can obtain distance vector d=[d
1, d
2, d
3.., d
k];
B) formula statistics with histogram function D (d is utilized
i)=n
icount distance vector histogram, then to histogram variable n
icarry out ascending sort, obtain vector n=[n
1, n
2, n
3.., n
k], wherein n
1≤ n
2≤ n
2≤ ...≤n
k; Vector n is the edge feature of buildings facade door and window.
In this programme, described step 3, utilizes the Gini coefficient in economics to add up described edge feature, obtains the Gini coefficient of buildings facade, is specially:
S310, supposing that the edge feature of the door and window extracted in image is f, is histogram f=[f by the distribution statistics of f
1, f
2, f
3..., f
k], the element in histogram is sorted from small to large, obtain new histogram set be f '=[f '
1, f '
2, f '
3..., f '
k], the Gini coefficient formula so measuring image rule degree is:
Wherein, || f||
1for first normal form, K is the classification sum of statistics with histogram, and the scope of G is from 0 to 1, and G is larger, and buildings facade is more complete, and G is less, and the damage of buildings facade is serious; Find through statistical experiment, buildings damage threshold value is 0.45, and statistical law in economics is basically identical;
S320, the statistics edge feature vector n in step 2 to be brought in (2) formula as f, obtain the Gini coefficient of elevation of building.
Utilize the Gini coefficient in economics to add up buildings facade edge feature, obtain buildings facade Gini coefficient.Gini coefficient (Ginicoefficient), 20 beginning of the century Italy economist Geordies, the index of the fair degree of the judgement distribution of earnings defined according to lorenz curve is used for an important analysis index of the inner income disparity situation of integrated survey resident in the world.The numerical value of Gini coefficient is usually in 0 to 1 scope, and numerical value larger expression income difference is large, and unfair distribution is put down, and numerical value less expression income difference is little, fairness in distribution.In the world usually using 0.4 as the warning line of gap between the rich and the poor, be greater than this numerical value and easily occur social unrest.Gini coefficient is a kind of important measure index of adding up uneven distribution, and have good Scale invariant and clone's invariant feature, these characteristics well meet six characteristics of sparse measurement.
Intact buildings facade normal conditions have good rule degree, the door and window on surface becomes straight uniform to distribute, there is certain sparse distribution characteristic, Gini coefficient is introduced as the metric index of buildings facade damage by this method, illustrate that when Gini coefficient is larger buildings facade structures is loose, there is good rule, damage; Otherwise, illustrate that buildings facade structures is mixed and disorderly, there occurs damage.
In this programme, according to described Gini coefficient, described step 4, judges whether buildings facade is damaged, and is specially:
When Gini coefficient G is greater than 0.45, represent that buildings facade is intact; Otherwise, when Gini coefficient G is less than 0.45, represent that elevation of building there occurs damage.
In sum, Gini coefficient is utilized to take full advantage of the structural information of buildings facade in inclination image to the method detecting the damage of buildings facade in a kind of aviation inclination image that the present invention proposes, Gini coefficient simultaneously in integrated economics is as damage index, significantly improve the precision that the damage of buildings facade detects, feature is: 1) the method has when not needing data before prior imformation and calamity, the damage of buildings facade can be carried out simply efficiently detect, reduce the complexity of method, also save production cost.2) introduce the index that the Gini coefficient in economics detects as the damage of buildings facade, the architectural feature that can make full use of buildings facade judges damage, not only improves automaticity and the precision of judgement, and the needs of realistic production.
Embodiment 2
The present embodiment 2 is device embodiment, and embodiment 1 is embodiment of the method, and device embodiment and embodiment of the method belong to same technical conceive, and the content of not detailed description in device embodiment, refers to embodiment of the method.
The structural representation of the device that the damage of buildings facade detects in a kind of oblique aerial image that Fig. 2 provides for the specific embodiment of the invention.As shown in Figure 2, the device that in a kind of oblique aerial image of the present invention, the damage of buildings facade detects, comprising:
Buildings facade cutting unit, for utilizing the k-means clustering algorithm based on rough set theory to split buildings facade, obtains the door and window of buildings facade;
Door and window edge feature calculation unit, for utilizing canny algorithm to carry out rim detection to the door and window of buildings facade, obtains the edge feature of door and window;
Gini coefficient computing unit, for utilizing the Gini coefficient in economics to add up described edge feature, obtains the Gini coefficient of buildings facade;
According to described Gini coefficient, damage judging unit, for judging whether buildings facade is damaged.
The present invention proposes in a kind of aviation inclination image and utilize Gini coefficient to detect the device of buildings facade damage, this device first utilizes the k-means algorithm improved namely the door and window of buildings facade and metope to be split based on the k-means clustering algorithm of rough set theory by buildings facade cutting unit, obtains the door and window of buildings; Then use the facade door and window of canny algorithm to buildings to carry out rim detection by door and window edge feature calculation unit, obtain the edge feature of building window; Make use of Gini coefficient in economics as damage index overall tolerance buildings facade damage information finally by Gini coefficient computing unit, thus reach and judge whether buildings facade is damaged.Visible, the present invention takes full advantage of the structural information of buildings facade in inclination image, Gini coefficient simultaneously in integrated economics is as damage index, significantly improve the precision that the damage of buildings facade detects, feature is: 1) this device has when not needing data before prior imformation and calamity, the damage of buildings facade can be carried out simply efficiently detect, reduce the complexity of method, also save production cost.2) introduce the index that the Gini coefficient in economics detects as the damage of buildings facade, the architectural feature that can make full use of buildings facade judges damage, not only improves automaticity and the precision of judgement, and the needs of realistic production.
Described buildings facade cutting unit is f specifically for the gray-scale value of pixel in a, image, wherein f=0,1,2 ... 255, k the central point utilizing rough set theory to obtain is as preliminary classification average μ
1, μ
2, μ
3..., μ
k; Distance D between b, the gray-scale value f calculating each pixel in image and described preliminary classification average μ, composes the initial classes average class nearest apart from it, namely by each pixel
D|f
p-μ
k|=min{D|f
p-μ
i|,(i=1,2,…k)}(1)
Carry out iteration to (1) formula, wherein p is the central point in iterative process; C, for i=1,2 ..., k calculates new cluster centre, upgrades class average:
in formula, N
ibe
in number of pixels, m is iterations; D, all pixels to be investigated one by one, if i=1,2 ... k, has μ
i (m+1)=μ
i (m), then algorithm convergence, terminates, otherwise returns b continuation next iteration.
Described door and window edge feature calculation unit, specifically for a, utilizes canny algorithm to carry out rim detection to the door and window of buildings facade, obtains the door and window edge of buildings facade; B, be all perpendicular to ground due to most of buildings facade, first statistics is parallel to the range distribution between the parallel lines on ground, then calculates the histogram of distance, finally obtains the edge feature of door and window; Flow process is as follows: b10, may damage due to buildings facade, therefore the outline line that door and window rim detection obtains may not be parallel to each other, the method of therefore adding up the range distribution employing between the parallel lines being parallel to ground is: as shown in Figure 3, buildings facade is added up to vertical direction along horizontal direction every a fixed step size, calculate vertical direction and close on distance between two pixels, be designated as d
i, whole facade image can obtain distance vector d=[d
1, d
2, d
3.., d
k]; B20, utilize formula statistics with histogram function D (d
i)=n
icount distance vector histogram, then to histogram variable n
icarry out ascending sort, obtain vector n=[n
1, n
2, n
3.., n
k], wherein n
1≤ n
2≤ n
2≤ ...≤n
k; Vector n is the edge feature of buildings facade door and window.
Described Gini coefficient computing unit, specifically for a, supposing that the edge feature of the door and window extracted in image is f, is histogram f=[f by the distribution statistics of f
1, f
2, f
3..., f
k], the element in histogram is sorted from small to large, obtain new histogram set be f '=[f '
1, f '
2, f '
3..., f '
k], the Gini coefficient formula so measuring image rule degree is:
Wherein, || f||
1for first normal form, K is the classification sum of statistics with histogram, and the scope of G is from 0 to 1, and G is larger, and buildings facade is more complete, and G is less, and the damage of buildings facade is serious; Find through statistical experiment, buildings damage threshold value is 0.45, and statistical law in economics is basically identical; B, the statistics edge feature vector n in step 2 to be brought in (2) formula as f, obtain the Gini coefficient of elevation of building.
Described damage judging unit, specifically for when Gini coefficient G is greater than 0.45, represents that buildings facade is intact; Otherwise, when Gini coefficient G is less than 0.45, represent that elevation of building there occurs damage.
These are only the preferred embodiments of the present invention; not thereby the scope of the claims of the present invention is limited; every utilize instructions of the present invention and accompanying drawing content to do equivalent structure or equivalent flow process conversion; or be directly or indirectly used in other relevant technical fields, be all in like manner included in scope of patent protection of the present invention.
Claims (10)
1. the method that in oblique aerial image, the damage of buildings facade detects, is characterized in that, comprising:
Step one, utilizes the k-means clustering algorithm based on rough set theory to split buildings facade, obtains the door and window of buildings facade;
Step 2, utilizes canny algorithm to carry out rim detection to the door and window of buildings facade, obtains the edge feature of door and window;
Step 3, utilizes the Gini coefficient in economics to add up described edge feature, obtains the Gini coefficient of buildings facade;
According to described Gini coefficient, step 4, judges whether buildings facade is damaged.
2. the method that in a kind of oblique aerial image according to claim 1, the damage of buildings facade detects, it is characterized in that, described step one, utilize the k-means clustering algorithm based on rough set theory to split buildings facade, obtain the door and window of buildings facade, be specially:
In S110, image, the gray-scale value of pixel is f, wherein f=0,1,2 ... 255, k the central point utilizing rough set theory to obtain is as preliminary classification average μ
1, μ
2, μ
3..., μ
k;
Distance D between S120, the gray-scale value f calculating each pixel in image and previous step preliminary classification average μ, composes the initial classes average class nearest apart from it, namely by each pixel
D|f
p-μ
k|=min{D|f
p-μ
i|,(i=1,2,…k)}(1)
Carry out iteration to (1) formula, wherein p is the central point in iterative process;
S130, for i=1,2 ..., k calculates new cluster centre, upgrades class average:
In formula, N
ibe
in number of pixels, m is iterations;
S140, all pixels to be investigated one by one, if i=1,2 ... k, has μ
i (m+1)=μ
i (m), then algorithm convergence,
Terminate, otherwise return S120 continuation next iteration.
3. the method that in a kind of oblique aerial image according to claim 2, the damage of buildings facade detects, is characterized in that, described step 2, utilizes canny algorithm to carry out rim detection to the door and window of buildings facade, obtains the edge feature of door and window, is specially:
S210, utilize canny algorithm to carry out rim detection to the door and window of buildings facade, obtain the door and window edge of buildings facade;
S220, be all perpendicular to ground due to most of buildings facade, first statistics is parallel to the range distribution between the parallel lines on ground, then calculates the histogram of distance, finally obtains the edge feature of door and window; Flow process is as follows:
A) because buildings facade may be damaged, therefore the outline line that door and window rim detection obtains may not be parallel to each other, the method of therefore adding up the range distribution employing between the parallel lines being parallel to ground is: add up to vertical direction buildings facade along horizontal direction every a fixed step size, calculate vertical direction and close on distance between two pixels, be designated as d
i, whole facade image can obtain distance vector d=[d
1, d
2, d
3.., d
k];
B) formula statistics with histogram function D (d is utilized
i)=n
icount distance vector histogram, then to histogram variable n
icarry out ascending sort, obtain vector n=[n
1, n
2, n
3.., n
k], wherein n
1≤ n
2≤ n
2≤ ...≤n
k; Vector n is the edge feature of buildings facade door and window.
4. the method that in a kind of oblique aerial image according to claim 3, the damage of buildings facade detects, it is characterized in that, described step 3, utilize the Gini coefficient in economics to add up described edge feature, obtain the Gini coefficient of buildings facade, be specially:
S310, supposing that the edge feature of the door and window extracted in image is f, is histogram f=[f by the distribution statistics of f
1, f
2, f
3..., f
k], the element in histogram is sorted from small to large, obtain new histogram set be f '=[f '
1, f '
2, f '
3..., f '
k], the Gini coefficient formula so measuring image rule degree is:
Wherein, || f||
1for first normal form, K is the classification sum of statistics with histogram, and the scope of G is from 0 to 1, and G is larger, and buildings facade is more complete, and G is less, and the damage of buildings facade is serious; Find through statistical experiment, buildings damage threshold value is 0.45, and statistical law in economics is basically identical;
S320, the statistics edge feature vector n in step 2 to be brought in (2) formula as f, obtain the Gini coefficient of elevation of building.
5. the method that in a kind of oblique aerial image according to claim 4, the damage of buildings facade detects, is characterized in that, described step 4, judges whether buildings facade is damaged, and is specially according to described Gini coefficient:
When Gini coefficient G is greater than 0.45, represent that buildings facade is intact; Otherwise, when Gini coefficient G is less than 0.45, represent that elevation of building there occurs damage.
6. the device that in oblique aerial image, the damage of buildings facade detects, is characterized in that, comprising:
Buildings facade cutting unit, for utilizing the k-means clustering algorithm based on rough set theory to split buildings facade, obtains the door and window of buildings facade;
Door and window edge feature calculation unit, for utilizing canny algorithm to carry out rim detection to the door and window of buildings facade, obtains the edge feature of door and window;
Gini coefficient computing unit, for utilizing the Gini coefficient in economics to add up described edge feature, obtains the Gini coefficient of buildings facade;
According to described Gini coefficient, damage judging unit, for judging whether buildings facade is damaged.
7. the device that in a kind of oblique aerial image according to claim 6, the damage of buildings facade detects, it is characterized in that, described buildings facade cutting unit, be f specifically for the gray-scale value of pixel in a, image, wherein f=0,1,2 ... 255, k the central point utilizing rough set theory to obtain is as preliminary classification average μ
1, μ
2, μ
3..., μ
k; Distance D between b, the gray-scale value f calculating each pixel in image and described preliminary classification average μ, composes the initial classes average class nearest apart from it, namely by each pixel
D|f
p-μ
k|=min{D|f
p-μ
i|,(i=1,2,…k)}(1)
Carry out iteration to (1) formula, wherein p is the central point in iterative process; C, for i=1,2 ..., k calculates new cluster centre, upgrades class average:
in formula, N
ibe
in number of pixels, m is iterations; D, all pixels to be investigated one by one, if i=1,2 ... k, has μ
i (m+1)=μ
i (m), then algorithm convergence, terminates, otherwise returns b continuation next iteration.
8. the device that in a kind of oblique aerial image according to claim 7, the damage of buildings facade detects, it is characterized in that, described door and window edge feature calculation unit, specifically for a, utilize canny algorithm to carry out rim detection to the door and window of buildings facade, obtain the door and window edge of buildings facade; B, be all perpendicular to ground due to most of buildings facade, first statistics is parallel to the range distribution between the parallel lines on ground, then calculates the histogram of distance, finally obtains the edge feature of door and window; Flow process is as follows: b10, may damage due to buildings facade, therefore the outline line that door and window rim detection obtains may not be parallel to each other, the method of therefore adding up the range distribution employing between the parallel lines being parallel to ground is: add up to vertical direction buildings facade along horizontal direction every a fixed step size, calculate vertical direction and close on distance between two pixels, be designated as d
i, whole facade image can obtain distance vector d=[d
1, d
2, d
3.., d
k]; B20, utilize formula statistics with histogram function D (d
i)=n
icount distance vector histogram, then to histogram variable n
icarry out ascending sort, obtain vector n=[n
1, n
2, n
3.., n
k], wherein n
1≤ n
2≤ n
2≤ ...≤n
k; Vector n is the edge feature of buildings facade door and window.
9. the device that in a kind of oblique aerial image according to claim 8, the damage of buildings facade detects, it is characterized in that, described Gini coefficient computing unit, specifically for a, supposing that the edge feature of the door and window extracted in image is f, is histogram f=[f by the distribution statistics of f
1, f
2, f
3..., f
k], the element in histogram is sorted from small to large, obtain new histogram set be f '=[f '
1, f '
2, f '
3..., f '
k], the Gini coefficient formula so measuring image rule degree is:
Wherein, || f||
1for first normal form, K is the classification sum of statistics with histogram, and the scope of G is from 0 to 1, and G is larger, and buildings facade is more complete, and G is less, and the damage of buildings facade is serious; Find through statistical experiment, buildings damage threshold value is 0.45, and statistical law in economics is basically identical; B, the statistics edge feature vector n in step 2 to be brought in (2) formula as f, obtain the Gini coefficient of elevation of building.
10. the device that in a kind of oblique aerial image according to claim 9, the damage of buildings facade detects, is characterized in that, described damage judging unit, specifically for when Gini coefficient G is greater than 0.45, represents that buildings facade is intact; Otherwise, when Gini coefficient G is less than 0.45, represent that elevation of building there occurs damage.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510494876.0A CN105139388B (en) | 2015-08-12 | 2015-08-12 | The method and apparatus of building facade damage detection in a kind of oblique aerial image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510494876.0A CN105139388B (en) | 2015-08-12 | 2015-08-12 | The method and apparatus of building facade damage detection in a kind of oblique aerial image |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105139388A true CN105139388A (en) | 2015-12-09 |
CN105139388B CN105139388B (en) | 2017-12-15 |
Family
ID=54724721
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510494876.0A Active CN105139388B (en) | 2015-08-12 | 2015-08-12 | The method and apparatus of building facade damage detection in a kind of oblique aerial image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105139388B (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105931191A (en) * | 2015-12-21 | 2016-09-07 | 南京邮电大学 | Blind image deconvolution method based on concave-convex mixed regular prior |
CN109035206A (en) * | 2018-06-28 | 2018-12-18 | 中国地震局地震预测研究所 | A kind of building collapse state detection method |
CN109703465A (en) * | 2018-12-28 | 2019-05-03 | 百度在线网络技术(北京)有限公司 | The control method and device of vehicle-mounted imaging sensor |
CN109886585A (en) * | 2019-02-27 | 2019-06-14 | 广东三维家信息科技有限公司 | A kind of ventilation of room evaluation method and device |
CN111126183A (en) * | 2019-12-09 | 2020-05-08 | 武汉大学 | Method for detecting damage of building after earthquake based on near-ground image data |
CN111126184A (en) * | 2019-12-09 | 2020-05-08 | 武汉大学 | Post-earthquake building damage detection method based on unmanned aerial vehicle video |
CN111881915A (en) * | 2020-07-15 | 2020-11-03 | 武汉大学 | Satellite video target intelligent detection method based on multiple prior information constraints |
CN114782846A (en) * | 2022-06-20 | 2022-07-22 | 绵阳天仪空间科技有限公司 | Post-disaster rescue guiding system and method |
CN114782826A (en) * | 2022-06-20 | 2022-07-22 | 绵阳天仪空间科技有限公司 | Safety monitoring system and method for post-disaster building |
CN115424178A (en) * | 2022-09-05 | 2022-12-02 | 兰州大学 | Enhancement method for improving pavement crack data identification |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112200808B (en) * | 2020-12-07 | 2021-06-15 | 领伟创新智能系统(浙江)有限公司 | Strip steel surface defect detection method based on local Gini coefficient |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20030061228A1 (en) * | 2001-06-08 | 2003-03-27 | The Regents Of The University Of California | Parallel object-oriented decision tree system |
CN103729853A (en) * | 2014-01-15 | 2014-04-16 | 武汉大学 | Three-dimensional GIS assisted high-resolution remote sensing image building collapse-damage detecting method |
US20140278587A1 (en) * | 2013-03-15 | 2014-09-18 | State Farm Mutual Automobile Insurance Company | Systems and methods for assessing property damage |
CN104268879A (en) * | 2014-09-28 | 2015-01-07 | 民政部国家减灾中心 | Physical building quantity damage evaluation method based on remote sensing multi-spectral images |
-
2015
- 2015-08-12 CN CN201510494876.0A patent/CN105139388B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20030061228A1 (en) * | 2001-06-08 | 2003-03-27 | The Regents Of The University Of California | Parallel object-oriented decision tree system |
US20140278587A1 (en) * | 2013-03-15 | 2014-09-18 | State Farm Mutual Automobile Insurance Company | Systems and methods for assessing property damage |
CN103729853A (en) * | 2014-01-15 | 2014-04-16 | 武汉大学 | Three-dimensional GIS assisted high-resolution remote sensing image building collapse-damage detecting method |
CN104268879A (en) * | 2014-09-28 | 2015-01-07 | 民政部国家减灾中心 | Physical building quantity damage evaluation method based on remote sensing multi-spectral images |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105931191B (en) * | 2015-12-21 | 2019-05-03 | 南京邮电大学 | Image Blind deconvolution method based on concave-convex mixing canonical priori |
CN105931191A (en) * | 2015-12-21 | 2016-09-07 | 南京邮电大学 | Blind image deconvolution method based on concave-convex mixed regular prior |
CN109035206B (en) * | 2018-06-28 | 2020-10-09 | 中国地震局地震预测研究所 | Building damage state detection method |
CN109035206A (en) * | 2018-06-28 | 2018-12-18 | 中国地震局地震预测研究所 | A kind of building collapse state detection method |
CN109703465A (en) * | 2018-12-28 | 2019-05-03 | 百度在线网络技术(北京)有限公司 | The control method and device of vehicle-mounted imaging sensor |
CN109886585B (en) * | 2019-02-27 | 2021-01-01 | 广东三维家信息科技有限公司 | Room ventilation evaluation method and device |
CN109886585A (en) * | 2019-02-27 | 2019-06-14 | 广东三维家信息科技有限公司 | A kind of ventilation of room evaluation method and device |
CN111126184A (en) * | 2019-12-09 | 2020-05-08 | 武汉大学 | Post-earthquake building damage detection method based on unmanned aerial vehicle video |
CN111126183A (en) * | 2019-12-09 | 2020-05-08 | 武汉大学 | Method for detecting damage of building after earthquake based on near-ground image data |
CN111126184B (en) * | 2019-12-09 | 2022-04-12 | 武汉大学 | Post-earthquake building damage detection method based on unmanned aerial vehicle video |
CN111881915A (en) * | 2020-07-15 | 2020-11-03 | 武汉大学 | Satellite video target intelligent detection method based on multiple prior information constraints |
CN111881915B (en) * | 2020-07-15 | 2022-07-15 | 武汉大学 | Satellite video target intelligent detection method based on multiple prior information constraints |
CN114782846A (en) * | 2022-06-20 | 2022-07-22 | 绵阳天仪空间科技有限公司 | Post-disaster rescue guiding system and method |
CN114782826A (en) * | 2022-06-20 | 2022-07-22 | 绵阳天仪空间科技有限公司 | Safety monitoring system and method for post-disaster building |
CN115424178A (en) * | 2022-09-05 | 2022-12-02 | 兰州大学 | Enhancement method for improving pavement crack data identification |
Also Published As
Publication number | Publication date |
---|---|
CN105139388B (en) | 2017-12-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105139388A (en) | Method and apparatus for building facade damage detection in oblique aerial image | |
Kang et al. | The comparative research on image segmentation algorithms | |
CN109389180A (en) | A power equipment image-recognizing method and inspection robot based on deep learning | |
CN108038846A (en) | Transmission line equipment image defect detection method and system based on multilayer convolutional neural networks | |
CN103208123B (en) | Image partition method and system | |
CN103839065A (en) | Extraction method for dynamic crowd gathering characteristics | |
CN110232379A (en) | A kind of vehicle attitude detection method and system | |
CN104182985A (en) | Remote sensing image change detection method | |
CN103559508B (en) | A kind of based on continuous Adaboost video vehicle detection method | |
CN110210452A (en) | It is a kind of based on improve tiny-yolov3 mine truck environment under object detection method | |
CN109935080B (en) | Monitoring system and method for real-time calculation of traffic flow on traffic line | |
CN102867183B (en) | Method and device for detecting littered objects of vehicle and intelligent traffic monitoring system | |
CN106408030A (en) | SAR image classification method based on middle lamella semantic attribute and convolution neural network | |
CN114998852A (en) | Intelligent detection method for road pavement diseases based on deep learning | |
CN104183142A (en) | Traffic flow statistics method based on image visual processing technology | |
CN108960142B (en) | Pedestrian re-identification method based on global feature loss function | |
CN103390046A (en) | Multi-scale dictionary natural scene image classification method based on latent Dirichlet model | |
CN110472652A (en) | A small amount of sample classification method based on semanteme guidance | |
CN104318241A (en) | Local density spectral clustering similarity measurement algorithm based on Self-tuning | |
CN104077609A (en) | Saliency detection method based on conditional random field | |
CN103679214A (en) | Vehicle detection method based on online area estimation and multi-feature decision fusion | |
CN105243387A (en) | Open-pit mine typical ground object classification method based on UAV image | |
CN115993096A (en) | High-rise building deformation measuring method | |
CN103679218B (en) | A kind of handwritten form keyword detection method | |
CN108491719A (en) | A kind of Android malware detection methods improving NB Algorithm |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |