CN110517255A - Shallow crack detection method based on attractor model - Google Patents

Abstract

The title of this patent is "Shallow Crack Detection Method Based on Attractor Model", which belongs to the field of computer vision and image processing. Cracks are the most common defect or damage phenomenon in concrete structures. Severe cracks may endanger the integrity and stability of the structure, so the detection of cracks is very necessary. This patent proposes a method for detecting cracks using image processing, especially for the detection of shallow cracks. This patent proposes a shallow crack detection algorithm based on the attractor model, assuming that each pixel is a vector , and as an attractor, and then according to the degree of attraction between them, match them in pairs to obtain multiple groups of attractors; then determine the rectangular area between each group of attractors, and make an image in the corresponding rectangular area in the original image processing, and finally get the final shallow fracture skeleton map by splicing. The effects achieved by the present invention include: clearly and accurately finding and displaying the position of the shallow crack without manual intervention.

Description

Shallow crack detection method based on attractor model

technical field

The invention relates to a method for detecting shallow cracks on buildings, belonging to the fields of computer vision and image processing.

Background technique

Cracks are the most common defect or damage phenomenon in concrete structures. Severe cracks may endanger the integrity and stability of the structure, so the detection of cracks is very necessary. Manual detection and positioning of cracks can achieve high accuracy, but this is time-consuming and laborious; while the current image processing algorithms for crack detection face cracks with different depths on buildings, in order to ensure the detection effect, they often focus on the overall image However, this overall method usually makes shallow cracks undetectable or incompletely displayed after detection. For example, for the crack map shown in Figure 1, in order to ensure the detection effect, the crack detection map after overall image processing is shown in Figure 2. Figure 3 and Figure 4 are the extracted shallow crack map in the lower right corner of Figure 1 and the crack detection map after the overall processing of the extracted lower right corner in Figure 2, respectively. By comparing Figure 3 and Figure 4, it can be seen that this overall processing is not very effective for the detection of shallow cracks. Therefore, for the detection of shallow cracks, this patent proposes a detection method based on the attractor model, which can clearly and accurately detect shallow cracks.

For the detection of shallow cracks, this patent proposes a shallow crack detection algorithm based on the attractor model. After obtaining the overall processed crack skeleton diagram, each pixel on it is assumed to be a vector, and each pixel As an attractor; then according to their direction and degree of attraction, match each pair to determine each pair of attractors; then extract a rectangular area between each pair of attractors, and lower the threshold in the corresponding rectangular area in the original image , until the fracture skeleton appears; finally, the final shallow fracture skeleton map is obtained by splicing.

Contents of the invention

The purpose of the present invention is to solve the problem that shallow cracks on the image cannot be detected or displayed incompletely when processing the building image as a whole. In order to solve these problems, this patent uses image processing as the main technical means, and proposes a shallow crack detection method based on the attractor model. The process includes: first, do image processing on the overall fracture map, including grayscale, top-hat transformation, thresholding, and skeleton map extraction; then extract the incompletely displayed shallow fracture area, and use the attractor algorithm to obtain each fracture area in the shallow fracture area. pair of attractors; then determine a rectangular area between each pair of attractors, and lower the binarization threshold at the corresponding position in the original image until cracks appear in the rectangular area; finally splice the results after each processing, and get Skeleton diagram of shallow fractures.

The effects achieved by the present invention include: clearly and accurately finding and displaying the position of the shallow crack without manual intervention.

Description of drawings

Figure 1 Original image of cracks

Figure 2 Overall crack detection diagram

Fig.3 The shallow fracture map in the lower right corner of the original fracture map

Figure 4 The shallow crack detection map in the lower right corner of the original crack map

Figure 5 pixel point eight-neighborhood window diagram

Figure 6 Pixel direction area map

Figure 7 Determining the schematic diagram of the rectangular area

Figure 8. Shallow crack detection diagram during iterative process

Figure 9. Shallow crack detection diagram during the iterative process

Figure 10 The final shallow crack detection map

Detailed ways

1. Attractor Algorithm

Through the previous image processing, the overall detection map of the crack map has been obtained, and the detection map of shallow cracks with incomplete detection is obtained by extracting the area in the lower right corner, as shown in Figure 4. Aiming at Figure 4, an attractor algorithm is proposed, and its steps are as follows:

1. Determine the type of pixel

The main purpose of this step is to classify the pixels in Figure 4 in order to determine attractor pairs, and the classification is based on the number of pixels with a pixel value of 255 in the eight-neighborhood window of the pixel. Assuming that t(i,j) is the pixel value of the image at coordinates (i,j), record the eight-neighborhood window of t(i,j) as W, and the eight-neighborhood window diagram of the pixel point is shown in Figure 5. W contains (Z1, Z2,..., Z8) eight pixels. If t(i, j) is 255, then if the number of pixels with a pixel value of 255 in the eight-neighborhood window of t(i, j) is zero, then the pixel point (i, j) is defined as an "isolated point"; if the number of pixels with a pixel value of 255 in the eight-neighborhood window of t(i,j) is one, then define the pixel point (i,j) as an "endpoint"; if t(i,j ) in the eight-neighborhood window with a pixel value of 255 is two or more, then the pixel point (i, j) is defined as a "connection point". Classify the pixels in Figure 4 according to the above rules, and save these three types of pixels by using the stack method, and each pixel is defined as an attractor, so this step defines all types of attractors.

2. Determine the size and direction of the attractor

The main purpose of this step is to determine the size and orientation for each class of attractors. First determine the direction of the attractor. For "isolated point", this patent defines its direction as any direction; for "end point", this patent defines its direction as the reverse of the only pixel existing in its eight neighbors The direction of the extension line; for "connection points", this patent defines them as having no direction. Secondly, determine the size of the attractor. For "endpoints" and "connection points", this patent makes their sizes similar to the magnetic force of the magnetic ball. The magnetic force of the magnetic ball is related to the number of magnetic balls connected to it, and the magnetic force connected to it The more balls, the greater its magnetic force. Similarly, this patent determines the size of an attractor by determining how many attractors are connected to it in front of an attractor; and for the "connection point", this patent defines its size as zero.

3. Determine the attractor pair

The main purpose of this step is to identify attractor pairs. The previous steps have classified and saved all effective attractors, and then compare these attractors pairwise to determine each pair of attractors. For "end point" and "connection point", this patent first determines a sector for each attractor, where the direction of the sector is consistent with the direction of the attractor, and the extension length of the sector is determined by the size of the attractor. The larger the magnet, the longer the extension length of the sector, and after experimental demonstration, this patent adopts the sector angle of (-15°, 15°), as shown in Figure 5, if the sectors of the two attractors intersect pair the two attractors. As for the "joint points", they have no size and direction, so they do not participate in pairing, but only play an important role in determining the size of the attractor. This patent uses the method of stacking to compare these attractors. When two attractor points are determined to be a pair of attractors, delete the data of these two attractors in the stack, and then process them sequentially. Finally Identify attractor pairs.

2. Process the fracture map according to the attractor pair and obtain the shallow fracture skeleton map

The pairing of two attractors indicates that there is a potential shallow crack between the two attractors, so this step is mainly for these attractor pairs, determine the rectangular area between each pair of attractors, and in the original crack map The corresponding rectangular area changes the threshold until the fracture skeleton line appears in the rectangular area. The schematic diagram of the rectangular area is shown in Figure 7. The selection principle of the rectangular area is that the two attractor coordinates of a pair of attractors are the endpoints, and its length and width are the absolute values of the difference between the horizontal and vertical coordinates of the two attractors. Perform the above operations on all attractor pairs to obtain several rectangular areas; then lower the threshold in the corresponding rectangular areas in the original image until crack skeletons appear in the area; then stitch these rectangular areas to obtain a shallow crack skeleton map.

It is obviously impossible to achieve a good effect by only processing once, so this patent adopts an iterative idea to improve the experimental effect. The processed detection map is used as input to continue the above-mentioned attractor algorithm processing, and the final crack detection map is output when the set number of iterations is reached. Figure 8 and Figure 9 are the shallow crack detection diagrams in the iterative process after removing irrelevant interference points. Figure 10 is the final shallow crack detection map. Compared with the original crack map, it can be seen that the pixel attractor algorithm proposed in this patent clearly and accurately displays the skeleton map of shallow cracks, and solves the problem of undetectable or incomplete detection of shallow cracks in the overall processing.

The above is only an embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the description of the present invention and the contents of the accompanying drawings, or directly or indirectly used in other related technologies fields, all of which are equally included in the scope of patent protection of the present invention.

Claims (3)

1. based on the shallow fracture method for detecting for attracting submodel, it is characterised in that using the method for image procossing on building Crack is detected, and full automatic positioning and shows these cracks, is not necessarily to manual intervention, strong operability；It can be clear, quasi- True positioning crack.

2. according to claim 1 based on the shallow fracture method for detecting for attracting submodel, characterized in that examined for shallow fracture It surveys, this patent proposes a kind of based on the shallow fracture detection algorithm for attracting submodel, by artificially laying down a regulation, calculating pixel Between Attraction Degree, discovery and positioning and are shown shallow fracture.

3. according to claim 1 based on the shallow fracture method for detecting for attracting submodel, characterized in that include place below Reason process: fracture picture carries out global gray processing, does top cap transformation and the transformation of black cap, given threshold to the image after gray processing It extracts crack skeleton drawing, using attracting subalgorithm to handle shallow fracture, shows the skeleton drawing of shallow fracture striped.