CN109523479A - Visual detection method for gap on surface of pier - Google Patents

Abstract

The invention relates to a visual detection method for a gap on the surface of a pier, which obtains a color image of the surface of the pier through a linear array CCD camera; compressing and storing the color image in blocks, converting the color image into a gray image, and performing noise reduction treatment; performing image enhancement smoothing processing on the gray level image by using the linear characteristic of the bridge pier gap and adopting an image mask filtering algorithm; respectively calculating the interval and the intra-area distance of the area A and the area B by adopting an interval and intra-area distance method, enabling the ratio of the interval and the intra-area distance of the area A to the interval and the intra-area distance of the area B to be maximum by changing parameter values, taking a threshold result at the moment as a comprehensive segmentation threshold of a bridge pier gap image, and detecting the gap edge of the bridge pier image by adopting a wavelet transform gradient image segmentation algorithm; and finally, extracting the gap edge of the pier image by using a multi-scale morphological image edge detection method, and identifying the gap in the pier image by using a morphological watershed algorithm controlled by a marker, so that the efficient identification and detection of the pier gap can be realized.

Description

A visual detection method on the surface gap on the pier surface

technical field

The invention relates to image detection and recognition technology, in particular to a visual detection method for gaps on the surface of bridge piers.

Background technique

With the continuous development of society, the construction of bridge engineering in our country has also made great progress. However, due to the inadequacy of bridge design, construction period and future maintenance process, coupled with the influence of the natural environment and the service of bridge piers The length of time will cause damage to the pier structure gradually; the structural gaps on the surface of the pier will clearly reflect the damage of the pier. Hidden dangers.

Among many bridge pier diseases, bridge pier gap is a kind of damaged state that endangers the safety of bridge piers but is difficult to measure; at present, the detection of such diseases mostly stays in the manual operation stage, and the cracks are usually detected by close-range detection instruments. After zooming in, the width is detected, and the length of the crack is obtained by inaccurate measurement or estimation. This method requires inspection personnel to use a road inspection vehicle or a frame, so the work intensity is high, the inspection cost is high, and the safety requirements for personnel are high. high.

In recent years, with the advancement of science and technology, the improvement of computer data processing capabilities, speed, capacity and other performances, and the development of digital camera technology, digital image processing technology has been widely used in office automation, industrial robots, geographic data processing, and medicine. Related fields; in industrial engineering, it has been successfully applied to the detection of surface deformation and damage of components, the area of blades and the research of complex stress on components; in view of the shortcomings of current domestic bridge pier detection technology and the advantages of digital image technology, the development of image processing technology based The application practice of bridge pier gap automatic identification technology is of great significance to ensure the safety of bridge pier operation, reduce the cost of detection, and promote the rapid development of my country's transportation industry.

SUMMARY OF THE INVENTION

The purpose of the present invention is to propose a visual detection method for bridge pier surface gaps in order to solve the background technology, obtain the color image of the bridge pier surface through a linear array CCD camera; compress and store the color image in blocks, convert the color image into a grayscale image, and Carry out noise reduction processing; use the linear characteristics of the bridge pier gap, use the image mask filter algorithm to perform image enhancement and smoothing on the gray image; use the interval and intra-area distance method to calculate the interval and intra-area distance of area A and area B respectively, and pass Change the parameter value so that the distance ratio between the interval and the area of area A and area B reaches the maximum, and the threshold value at this time is used as the comprehensive segmentation threshold of the gap image of the pier, and the edge of the gap of the pier image is detected by using the wavelet transform gradient image segmentation algorithm; finally Using the multi-scale morphological image edge detection method to extract the gap edge of the bridge pier image, and using the marker-controlled morphological watershed algorithm to identify the gap in the bridge pier image can realize efficient recognition and detection of the bridge pier gap, and has a good application prospect.

Due to the influence of various factors, such as image acquisition, image transmission, and various image transformations, etc., the quality of the image of the bridge pier is often affected, and it is difficult to identify the gap of the bridge pier; It is more suitable for specific applications; for bridge pier gap images, it is to enhance the information of bridge pier gaps and reduce various noises such as textures, shadows, and uneven illumination.

In order to achieve the above object, the present invention adopts the following technical solutions:

A visual detection method for the surface gap of bridge pier, in which a linear array CCD camera and a displacement sensor are fixed on the same detection platform, during the movement of the detection platform, the displacement sensor provides a displacement signal, and the linear array CCD camera obtains a color image of the bridge pier surface according to the displacement signal;

Compress and store the color image in blocks, convert the color image into a grayscale image, and perform noise reduction processing;

Using the linear characteristics of the gap between the bridge pier, the image mask filter algorithm is used to enhance the image enhancement and smooth processing of the gray image;

The interval and intra-area distance methods are used to calculate the interval and intra-area distances of area A and area B respectively, and the ratio of interval and intra-area distance between area A and area B is maximized by changing the parameter value, and the threshold result at this time is used as the pier gap image The comprehensive division threshold;

Using the wavelet transform gradient image segmentation algorithm to detect the gap edge of the pier image;

The multi-scale morphological edge detection algorithm is used to segment the pier image according to the type of structural element, the size of the structural element, and the number of expansion operations, and then extract the gap edge of the pier image.

In a grayscale image, the grayscale values of adjacent pixels in a region change gently, the difference is small, and the statistical variance is small; on the contrary, at the edge of the region, the grayscale values between image pixels change greatly, so the statistical variance is large ; The purpose of mask processing is to filter the image while not destroying the details of the edge of the area as much as possible. When this method is applied to image enhancement of bridge pier, it can not only remove the noise in the image, but also better protect the edge of the gap details; the processing of the image neighborhood is to process the pixel values of the neighborhood image and the pixel values of the sub-images that have similar or the same pixel distribution as the neighborhood.

The mechanism of filtering processing is to move the pre-defined mask point by point in the image to be processed. The response of the filter at each point (x, y) is calculated according to the pre-defined formula. Generally, the response output is determined by the mask template The coefficient and filtering mask move the area of the relative pixel value of the pixel value.

A digital image can be divided into multiple areas or sub-images. When partitioning, try to make the changes in the adjacent areas large, while the changes in the same area are small, and in the same area, the change of the middle pixels is smaller than that of the edge pixels. Variety.

In order to improve the quality of the image, reduce and remove the noise in the image, only the linear features of the image can be extracted. When implementing specific calculations, the center point of the mask is coincident with the pixel points that need to be processed in the image, and each element in the mask is The value is multiplied by the pixel value corresponding to the image area covered by the mask, and the gray value of the pixel where the mask center point is located is replaced by the output response of the mask, which is the sum of the products of the elements of the template calculated in the previous step.

If a 3×3 mask template is used, according to the above method, the mask linear filter response value R at the point (x, y) in the image is:

R = w (-1, -1) f (x-1, y-1)+w (-1, 0) f (x-1, y)+...+w (0,0) f (x, y )+w (1, 0) f (x+1, y)+w (1, 1) f (x+1, y+1)

When implementing mask smoothing filtering, when the mask moves in the image, the center point of the filter is close to the image contour. If a square smooth mask of n×n size is used, the distance between the mask center and the edge of the image is (n-2)/2 pixels, the edge of the image coincides with at least one edge of the mask; when the mask continues to move to the edge of the image, the edge of the mask will be outside the edge of the image; therefore, when moving the mask, try not to let the center of the mask The point is more than (n-2)/2 pixels away from the edge of the image.

Since the pier image gap has linear characteristics and has directionality, according to this characteristic, the single template discussed in the previous section is extended to construct a single mask in 8 directions; generally speaking, a pixel in the image has 8 neighbors Domain, that is to say, a pixel has 8 directions, as shown in Figure 2.

Construct 8 templates with a size of 3×3 in 8 directions, process the 8 templates in turn in the same image window, and in the image window, compare the gray value of each pixel with the element value of the corresponding position of the template in turn. Ride, then accumulate the sum, the result is C _i = 0,1, ... 7.

The output value of the central pixel of the image window is represented by the following formula: g(m, n)=max(C _i ).

When the image window is 5×5, and the center position (hollow dot in the window) is (j, k), 9 different mask templates can be determined in this window, as shown in Fig. 3 .

Therefore, when performing mask filtering on an image, first calculate the mean and variance of each template:

In the formula: i represents the serial number of each mask template, i=1, 2, ..., 9, q is the number of pixels contained in the corresponding mask template, (m, n) is the pixel in the mask template relative to the central pixel (J, K) displacement.

Calculate the variance of the 9 templates and compare them, and use the gray mean value corresponding to the template with the smallest variance as the new gray value of the mask smooth output:

Therefore, as a further limitation of the present invention, the image enhancement smoothing process is specifically:

First, calculate the mean E _i and variance Ω _i of each template:

In the formula: i represents the number of each mask plate, i=1, 2,..., 9,

q is the number of pixels contained in the corresponding mask, and (m, n) is the displacement of the pixel in the mask relative to the central pixel (j, k);

Calculate the variance of 9 templates and compare it. The ash the average value of the ash with the smallest variance template is used as the new gray value of the smooth output of the mask.

Since the image consists of a target and a background, and the target is the gap object to be identified, it is necessary to divide the image into two parts according to the pixel value. The most common method to segment the gap from the background is to select a threshold, which is greater than the threshold. Other composition backgrounds.

When determining the threshold, if the threshold is set too high, occasional object points will be considered as the background; if the threshold is too low, the opposite will happen; therefore, when the image of the pier surface is complex, the histogram is difficult. There are obvious peaks, which are often connected together, and it becomes difficult to use the histogram to select the optimal threshold; at this time, the threshold can be determined according to the global information of the entire image, and the gaps in the pier image can be extracted by using the maximum area and interval distance threshold criteria Can achieve better results.

Therefore, as a further limitation of the present invention, the comprehensive segmentation threshold is obtained in the following manner:

Segment the entire road damage image into a series of sub-images; then calculate the segmentation threshold of each sub-image separately, and calculate the average gray level of the whole image, the average gray level of area A and area B; use interval and intra-area distance method Distribution calculation area A and region B range and district distance, and finally change The value ratio of the area A and the region B of the area A and the area B is maximum.

in, The distance between regional A and region B; The distance in the area A and region B.

In the pier image, the gap and background of a certain position are not exactly the same as those of another area, and there are some differences. This is due to the influence of the image acquisition equipment and the acquisition environment, that is, the non-gap area in some areas is gray If the intensity value is low, the gray value of the gap area in some areas will be high; at this time, using the global maximum interval and intra-area distance method to calculate the segmentation threshold of the entire image may easily cause some gap areas to be misjudged as non-gap areas , some non-seam regions are often misjudged as seam regions; therefore, using the global maximum inter-class and intra-class distance threshold segmentation algorithm to calculate the segmentation threshold cannot separate the damaged area from the non-damaged area.

However, the interval and intra-area distance method is used to distribute and calculate the interval and intra-area distance between area A and area B, and finally by changing Calculate the value of the value of the area A and area B, and the distance ratio between the area and the area reaches the maximum. As the comprehensive segmentation threshold of the damaged image, it can be achieved in the background area and weighted by the threshold of the global threshold to obtain a new threshold, that is, the comprehensive segmentation threshold, the comprehensive segmentation threshold is smaller than the original global threshold; using the new segmentation threshold to segment the road surface damage image, the small block can basically be determined as the target pixel.

In the automatic recognition of image gaps, it is difficult to select reconstructed sub-images because it is impossible to know the direction of noise in the image in advance. In fact, only reconstructing smooth images can basically achieve the elimination of random noise and enhancement of gaps. The purpose of the image; however, if the image has a large amount of texture in the horizontal and vertical directions, a reconstruction strategy of selecting horizontal sub-images and oblique sub-images can be adopted at this time.

For the reconstruction of the above-mentioned horizontal sub-image and oblique line sub-image, it is necessary to realize multi-resolution analysis of image reconstruction by wavelet transform; therefore, as a further limitation of the present invention, the wavelet transform gradient image segmentation algorithm is specifically: by wavelet transform And its inverse transformation for reconstruction, the gradient direction and the maximum modulus are stored in the wavelet coefficients, the maximum value of the wavelet coefficients is analyzed and transformed, and the wavelet coefficients corresponding to the noise and features are processed separately to realize the detection of the edge of the image gap and Noise separation can achieve effective smoothing and sharpening of the gaps in the image.

The purpose of image gap edge enhancement is to enhance the feature information of the target while suppressing the influence of noise; however, the image enhancement of bridge piers with noise is a difficulty faced by all enhancement algorithms, because the noise is the same as the edge of the real image with significant changes, in the frequency domain Both correspond to high-frequency sub-bands. When the enhancement algorithm is used to highlight the high-frequency part to improve the contrast of the edge and improve the image quality, the noise will inevitably be amplified.

Therefore, the wavelet coefficients corresponding to the noise and features can be treated differently, and the position and attribute of the edge can be determined according to the two components of the wavelet coefficient modulus maximum value and the gradient direction at each scale 2j; while reducing the noise , has an enhanced effect on the edge of the gap; the specific method is to transform the two components W ₁ and W ₂ of the wavelet transform on each scale 2j, and decompose the sub-image W _j (x,y) obtained by the jth layer Adapt to adaptive adjustments, make the following changes:

in, For the threshold, For gain, It is the edge on the scale J.

In order to make the algorithm have a good adaptive adjustment ability, parameters They are as follows:

in, W ₁ and W ₂ are the two components of the gradient image (x, y) wavelet transform.

Therefore, as a further limitation of the present invention, the wavelet coefficients corresponding to noise and features are respectively processed as follows:

Adaptively adjust the sub-image W _j (x, y) obtained by decomposing the jth layer, and make the following changes:

in, For the threshold, For gain, It is the edge on the size J;

In order to make the algorithm have a good adaptive adjustment ability, parameters They are as follows:

in, W ₁ and W ₂ are two components of the wavelet transform of the gradient image W _j (x,y).

A) and b) in Fig. 5 are the results of edge detection performed by the above-mentioned algorithm of the present invention, and the image after noise reduction, a) and b) in Fig. 5 have significantly reduced the noise of the image.

Due to the influence of various factors such as bridge pier texture, pier material, and photographing environment, the background noise of bridge pier images is relatively complex and contains a lot of noise. In order to avoid affecting the effect of gap edge detection, morphological filtering should be performed on bridge pier images first. The morphological opening operation process eliminates the small gray value details in the bridge pier image that are smaller in size compared with the structural elements, while maintaining the gray value and larger areas of the entire bridge pier image, so that sometimes the fine gaps in the image are taken as It is not easy to be recognized.

The morphological closing operation can eliminate the large gray value details of the small size compared with the structural elements in the pier image, and basically maintain the gray value and larger image area of the entire pier image; therefore, the morphological opening operation and closing operation The operation can be used for filtering of bridge pier images.

However, in actual application, the determination of structural elements should consider the shape characteristics and size of the gap components of the pier image to be processed, because it cannot handle objects with the same shape but different sizes; To consider its size, the present invention proposes to continuously increase the scale of structural elements for alternating sequential filtering, which is defined as follows according to the sequence of opening and closing operations:

Among them, B ₁ , B ₂ . . . B _i are structural elements with increasing size.

In Figure 6, a) is the original pier image, b) is the result of open-close filtering of square structural elements on the original image a), c) is the result of alternate filtering of image a) with structural elements of increasing scale, d ) is the median filtering result of the original image a). From the comparison of the processed images in Figure 6, it can be seen that the opening and closing filtering has a certain smoothing effect on the image. Compared with the median filtering alone, the filtering effect is better than that of the median filtering The value filtering is slightly worse, and the median filtering is better than the opening and closing filtering; Figure c) alternately filters the original image a) with structural elements of increasing scale, and its effect is better than the median filtering, and even better than Open and closure filtering; therefore, alternating opening and closing filtering algorithms can achieve better smooth effects.

In the pier image, the gaps generally have low gray value and show a long strip structure, so the gap object with the minimum gray value can be detected according to the gray value of the image, so the improved top-hat transformation (original image The image after subtracting the opening operation is called the top-hat transformation) as a gap detection to detect the image that has been morphologically filtered.

The process of the top hat change algorithm is to use the structural element to start the operation of the original bridge pier image Calculate the difference between the original image and the opening operation As a result, the gray bridge pier image top hat transformation.

Since the gaps in bridge pier images often have very small brightness, they look like dark areas in a bright background. Therefore, to extract gap features, the image must be negated first, and then use Perform a top hat transformation.

Correspondingly, in order to avoid negating the image, the image can be closed with the structure set nB first, and then the original image f(x) can be subtracted, which can also be called the improved bottom hat transformation, and the expression is: f ^nB (x )-f(x), where f ^nB (x) is the opening operation of the structural element nB on the grayscale image f(x), (n times); the output image after the above processing is a grayscale gradient image, and the threshold operation on the image can separate the gap from the pier image.

Therefore, as a further limitation of the present invention, the multi-scale morphology image edge detection method is specifically:

Firstly, the image is alternately opened and closed by using the ever-increasing structural elements to smooth the pier gap image and remove the noise, then negate the image, and then use the multi-scale top-hat transformation to perform top-hat transformation, using a multi-scale morphology edge detector The edge of the gap in the pier image is extracted, and finally the morphological watershed algorithm controlled by the marker is used to extract the gap object in the pier image for gap image recognition.

The present invention uses the morphological watershed algorithm controlled by markers to extract the gaps in the pier image. The watershed segmentation algorithm regards the image as a "topographic map". Small, segment the image by finding "catchment basins" and "watershed boundaries".

The effect of directly applying the watershed segmentation algorithm is often not good. If you mark the foreground object and the background object in the image, and then apply the watershed algorithm to achieve a better segmentation effect; the specific steps of the watershed segmentation method based on marker control are as follows:

Calculate the segmentation function, the darker area in the image is the object to be segmented, that is, the initial calculation of the position of the local minimum area in the image, which marks the position of the segmented object (referred to as the gap in this experiment); the above steps represent the same as our The irrelevant details of the segmentation problem, in order to eliminate these details, calculate the set of the minimum area of the image through a certain height threshold T, so as to better mark the position of the local minimum area; use the watershed distance transformation to calculate the external mark, generally use the watershed ridge As an external mark; after obtaining the internal and external markers, use the forced minimum technique to modify the grayscale image so that the local minimum area only appears at the marked position; perform watershed transformation on the image with the modified markers to segment out the gap objects.

A) in Figure 7 is the image of the bridge pier gap after morphological transformation, and b) in Figure 7 is the image of a) in Figure 7 after watershed transformation; it can be seen that the morphological watershed transformation has an important role in the gap segmentation of the pier image It has a certain role and has a better division effect.

The beneficial effects of the present invention are:

1. The present invention aims at the distribution characteristics of the pixel gray value of the pier gap image, and the needs of the pier gap image segmentation, adopts the image mask preprocessing algorithm, realizes the pier gap image filtering, and compares the pier gap with different low-pass filtering and high-pass filtering methods. Comparing the filtering effect of the seam image, on the basis of reducing image noise and improving image quality, the edge of the image seam is effectively protected, which provides a basis for subsequent seam identification and segmentation.

2. The present invention calculates the gradient and modulus maximum of wavelet transform, searches for the wavelet modulus maximum point on the image, obtains the crack information that needs to be identified, and then processes the wavelet coefficients corresponding to noise and features separately, while reducing noise It has an enhanced effect on the edge of the gap between the pier.

3. By improving the conventional morphological algorithm and using the multi-scale morphological edge detection algorithm to segment the pier image according to the type of structural element, the size of the structural element, the number of expansion operations, etc., better results can be achieved.

Description of drawings

Fig. 1 is the schematic diagram of bridge pier surface gap visual inspection method proposed by the present invention

Fig. 2 is a template direction definition diagram in the bridge pier surface gap visual inspection method proposed by the present invention

Figure 3 is a schematic diagram of the mask window and 9 different mask plates

Figure 4 is a comparison chart of a variety of image enhancement methods

Figure 5 is the result of using a wavelet change gradient image segmentation algorithm to deal with the gap edge of the pier image

Figure 6 is the result of different filtering methods for the gap image of the bridge pier

In Figure 7, a) is the image of the bridge pier gap after morphological transformation, and b) is the image of a) after watershed transformation

Detailed ways

The present invention will be further described below in conjunction with specific embodiments.

A visual detection method for the surface gap of bridge pier, in which a linear array CCD camera and a displacement sensor are fixed on the same detection platform, during the movement of the detection platform, the displacement sensor provides a displacement signal, and the linear array CCD camera obtains a color image of the bridge pier surface according to the displacement signal;

Compress and store the color image in blocks, convert the color image into a grayscale image, and perform noise reduction processing;

Using the linear characteristics of the gaps in the bridge piers, the image mask filter algorithm is used to perform image enhancement and smoothing on the grayscale image;

The interval and intra-area distance methods are used to calculate the interval and intra-area distances of area A and area B respectively, and the ratio of interval and intra-area distance between area A and area B is maximized by changing the parameter value, and the threshold result at this time is used as the pier gap image The comprehensive division threshold;

Use the wavelet change gradient image segmentation algorithm to deal with the gap edge of the pier image;

The multi-scale morphological image edge detection method is used to extract the gap edge of the bridge pier image, and finally the marker-controlled morphological watershed algorithm is used to identify the gap in the bridge pier image.

Among them, the image enhanced smooth processing is specifically:

First, calculate the average value of each template e _i and square difference ω _i :

In the formula: i represents the number of each mask plate, i=1, 2,..., 9,

q is the number of pixels contained in the corresponding mask, and (m, n) is the displacement of the pixel in the mask relative to the central pixel (j, k);

Calculate the variance of 9 templates and compare it. The ash the average value of the ash with the smallest variance template is used as the new gray value of the smooth output of the mask.

Among them, the comprehensive segmentation threshold is obtained by the following method:

Segment the entire road damage image into a series of sub-images; then calculate the segmentation threshold of each sub-image separately, and calculate the average gray level of the whole image, the average gray level of area A and area B; use interval and intra-area distance method Distribution calculation area A and region B range and district distance, and finally change The value ratio of the area A and the region B of the area A and the area B is maximum.

in, The distance between regional A and region B; The distance in the area A and region B.

The wavelet transform gradient image segmentation algorithm is specifically: through the wavelet transform and its inverse transform for reconstruction, the gradient direction and the maximum value of the modulus are stored in the wavelet coefficients, the maximum value of the wavelet coefficients is analyzed and transformed, and the wavelet corresponding to the noise and features The coefficients are processed separately to realize the detection of image gap edges and noise separation.

The small wave coefficients corresponding to noise and characteristics are processed separately:

Adaptively adjust the sub-image W _j (x, y) obtained by decomposing the jth layer, and make the following changes:

in, is the threshold, for gain, It is the edge on the size J;

In order to make the algorithm have a good adaptive adjustment ability, parameters They are as follows:

in, W ₁ and W ₂ are two components of the wavelet transform of the gradient image W _j (x,y).

Among them, the multi-scale morphological image edge detection method is as follows: first, use the ever-increasing structural elements to perform alternate opening and closing filtering on the image, smooth the pier gap image and remove the noise, then negate the image, and then use the multi-scale top-hat change Perform top-hat transformation, use multi-scale morphological edge detection algorithm to segment the pier image according to the type of structural element, the size of the structural element, and the number of expansion operations, and then extract the gap edge of the pier image, and finally use the morphological watershed controlled by the marker The algorithm extracts the gap object in the bridge pier image, and the gap image recognition is performed.

The invention obtains the color image of the bridge pier surface through a linear array CCD camera; compresses and stores the color image in blocks, converts the color image into a grayscale image, and performs noise reduction processing; utilizes the linear characteristics of the bridge pier gap, and adopts an image mask filter algorithm to The grayscale image is processed for image enhancement and smoothing; the interval and intra-area distances of area A and area B are calculated using the interval and intra-area distance methods, and the interval and intra-area distance ratios of area A and area B are maximized by changing parameter values. The threshold value at this time is used as the comprehensive segmentation threshold of the bridge pier gap image, and the wavelet transform gradient image segmentation algorithm is used to detect the gap edge of the bridge pier image; finally, the multi-scale morphology image edge detection method is used to extract the gap edge of the bridge pier image, and the marker The controlled morphological watershed algorithm identifies the gaps in the bridge pier image, which can realize the efficient identification and detection of bridge pier gaps, and has a good application prospect.

The above description is only a preferred embodiment of the present invention, but the protection scope of the present invention is not limited to this. The equivalent replacement or change of the inventive concept thereof shall be included within the protection scope of the present invention.

Claims (6)

1. A visual detection method for a gap on the surface of a pier is characterized by comprising the following steps: fixing the linear array CCD camera and the displacement sensor on the same detection platform, wherein the displacement sensor provides displacement signals in the moving process of the detection platform, and the linear array CCD camera acquires a color image of the surface of the pier according to the displacement signals;

compressing and storing the color image in blocks, converting the color image into a gray image, and performing noise reduction treatment;

performing image enhancement smoothing processing on the gray level image by using the linear characteristic of the bridge pier gap and adopting an image mask filtering algorithm;

respectively calculating the interval and the intra-area distance of the area A and the area B by adopting an interval and intra-area distance method, enabling the ratio of the interval and the intra-area distance of the area A and the area B to be maximum by changing parameter values, and taking the threshold result at the moment as a comprehensive segmentation threshold of the bridge pier gap image;

processing the gap edge of the pier image by adopting a wavelet transform gradient image segmentation algorithm;

and extracting the gap edge of the pier image by using a multi-scale morphological image edge detection method, and finally identifying the gap in the pier image by using a morphological watershed algorithm controlled by a marker.

2. The visual detection method for the gap between the surfaces of the piers according to claim 1, characterized in that: the image enhancement smoothing processing specifically comprises:

first, the mean value E of each template is calculated_iAnd variance Ω_i：

In the formula: i denotes the number of each mask, i is 1, 2, …, 9,

q is the number of pixels contained in the corresponding mask plate, and (m, n) is the displacement of the pixels in the mask plate relative to the central pixels (j, k);

then calculating the variances of the 9 templates and comparing, taking the gray average value corresponding to the template with the minimum variance as a new gray value of the mask smooth output,

3. the visual detection method for the gap between the surfaces of the piers according to claim 2, characterized in that: the comprehensive segmentation threshold is obtained by the following method:

dividing the whole pavement damage image into a series of sub-images; then respectively calculating the segmentation threshold of each sub-image, and solving the average gray scale of the whole image and the average gray scale of the area A and the area B; calculating the interval and intra-area distance of the area A and the area B by adopting the interval and intra-area distance method distribution, and finally changingThe maximum distance ratio between the section of the area a and the area B is obtained as a total segmentation threshold of the damaged image:

wherein,the interval distance between the area A and the area B;the intra-zone distance is the zone a and the zone B.

4. The visual detection method for the gap between the surfaces of the piers according to claim 3, characterized in that: the wavelet transformation gradient image segmentation algorithm specifically comprises the following steps: storing the gradient direction and the maximum modulus value in the wavelet coefficient through wavelet transformation and inverse transformation for reconstruction, analyzing and transforming the maximum wavelet coefficient, and respectively processing the wavelet coefficient corresponding to noise and characteristics to realize the detection of the slit edge of the image and the noise separation.

5. The visual detection method for the gap between the surfaces of the piers according to claim 4, characterized in that: the wavelet coefficients corresponding to the noise and the features are respectively processed specifically as follows:

sub-image W obtained by decomposing the j-th layer_j(x, y) is adaptively adjusted, as follows:

wherein,is a threshold value, and is,in order to achieve the gain,is the edge on the dimension j;

in order to make the algorithm have good adaptive adjustment capability, parametersRespectively as follows:

wherein,W₁and W₂Is a gradient image W_jTwo components of the (x, y) wavelet transform.

6. The visual detection method for the gap between the surfaces of the piers according to claim 5, characterized in that: the multi-scale morphological image edge detection method specifically comprises the following steps:

the method comprises the steps of firstly, conducting alternate opening and closing filtering on an image by using continuously-increased structural elements, smoothing a bridge pier gap image and removing noise, then performing image negation, then conducting top cap transformation by using multi-scale top cap change, segmenting the bridge pier image by using a multi-scale morphological edge detection algorithm according to the types of the structural elements, the sizes of the structural elements and the times of expansion operation, further extracting the gap edge of the bridge pier image, and finally extracting a gap object in the bridge pier image by using a morphological watershed algorithm controlled by a marker, and conducting gap image identification.