CN110363706B - Large-area bridge deck image splicing method - Google Patents

Abstract

The invention discloses a large-area bridge deck image splicing method. With the development of computer vision inspection technology, image inspection is increasingly applied to bridge inspection engineering practice. However, because the bridge space is large, if the remote shooting and sampling are adopted, the resolution of the camera is limited, so that satisfactory detection accuracy cannot be obtained. The invention is as follows: 1. and acquiring images of the detected bridge deck one by one to obtain a bridge deck image set. After that, image preprocessing is performed. 2. And (5) image registration. 3. And (5) image fusion. The invention replaces human eyes to finish the automatic nondestructive detection of the bridge disease characteristic by the image acquisition and processing technology, and has very important practical significance for the research of the bridge deck damage detection technology in the complex terrain environment. On one hand, the construction safety is enhanced, and on the other hand, the operation maneuverability and flexibility are improved. The invention realizes the fidelity splicing of the large-area bridge deck image and improves the image splicing precision and efficiency.

Description

A large-area bridge deck image mosaic method

technical field

The invention belongs to the technical field of image detection, and in particular relates to a large-area bridge deck image splicing method.

Background technique

As an important transportation hub, the bridge, after long-term exposure to the sun and rain and load operation, the internal stress will be transmitted to some weak parts along the bridge structure, resulting in cracks and other disease characteristics on the surface of the structure at this position. Due to the disease characteristics of the bridge surface, the outside air and harmful media can easily penetrate into the concrete to produce carbonate through chemical reaction, resulting in a decrease in the alkalinity environment of the steel bars, and the surface purification film is more prone to rust after being damaged. In addition, concrete carbonation It will also aggravate shrinkage cracking and cause serious harm to the safe use of concrete bridges. Therefore, in order to ensure the service life and safety performance of the beam structure, it is necessary to detect and treat the surface disease characteristics of the bridge in time.

In order to detect bridge deck disease characteristics in time and take remedial measures to eliminate potential safety hazards, manual inspection and manual marking are usually used. However, the working environment of bridge bottom surface detection is often dangerous. This detection method has poor maneuverability, high risk and low efficiency. Moreover, bridge deck damage is manually measured by experienced inspectors and recorded with naked eyes, which has a certain degree of subjectivity. The detection accuracy is more dependent on the experience and knowledge of experts, and experience lacks objectivity in quantitative analysis.

With the development of computer vision detection technology, image detection is gradually applied to bridge detection engineering practice. However, due to the large space of the bridge body, if remote shooting and sampling are used, it will be limited by the resolution of the camera, resulting in the inability to obtain satisfactory detection accuracy. Therefore, it is necessary to carry out continuous multi-group sampling on the surface of the bridge, and to carry out large-scale image stitching on the collected sample images, and the stitching effect will directly affect the detection accuracy of bridge deck disease characteristics. In order to reduce the cumulative error as much as possible on the basis of wide viewing angle and high resolution, improve the stitching accuracy of bridge deck images, and realize the global display of real bridge detection surface information.

Contents of the invention

The purpose of the present invention is to provide a large-area bridge deck image mosaic method.

Concrete steps of the present invention are as follows:

Step 1. Collect images of the detected bridge deck one by one to obtain a bridge deck image set. Afterwards, the brightness component information of each bridge deck image in the bridge deck image set is extracted and counted, and the brightness component information of each bridge deck image is equalized respectively. Then transform each bridge deck image into the frequency domain by Fourier transform, and use the phase information of the normalized cross power spectrum in the phase correlation algorithm to obtain the translation parameters between the images, and complete the pre-estimation of the overlapping area between adjacent images .

Step 2, image registration

First, SIFT feature points are extracted in the overlapping regions between adjacent images. Then the SIFT feature points are screened by an adaptive contrast threshold method to obtain a feature descriptor composed of matching point pairs. And the RANSAC algorithm is used to calculate the projection transformation matrix between adjacent images.

The adaptive contrast threshold method is as follows:

(1) Set the lower limit of the number of feature points N _min =200, the upper limit of N _max =300, and the contrast threshold T _c =T ₀ . T ₀ is the initial threshold, which ranges from 0.02 to 0.04.

(2) Perform feature point detection, and count the number N of feature points whose contrast is higher than _Tc .

(3) If N _min ≤ N ≤ N _max , include feature points with a contrast higher than T _c into the initial matching point set, remove feature points with a contrast lower than the threshold T _c , and go directly to step (5). Otherwise, go to step (4).

并执行步骤(3)。若N＞N_max，则将对比度阈值增大为原数值的2倍，并执行步骤(3)。(4) If N<N _min , then reduce the contrast threshold T _c to the original value

And execute step (3). If N>N _max , increase the contrast threshold to twice the original value, and perform step (3).

(5) Eliminate false feature points in the initial matching point set by the method of nearest neighbor and second nearest neighbor, and generate a feature descriptor. The feature descriptor contains multiple matching point pairs consisting of pairs of feature points, and the distance and direction information between each matching point pair.

Step 3, image fusion

Firstly, according to the projection transformation matrix between adjacent images, the corresponding bridge deck images are projectively transformed. Then, the gradual-in and gradual-out fusion algorithm is used to carry out weighted and smooth transitions on the RGB three-color channels of each adjacent bridge deck image respectively, and the bridge deck mosaic image is obtained.

In the fade-in and fade-out fusion algorithm, the fade-in and fade-out weighting formula of each fusion point pixel value I(x, y) in the overlapping area of adjacent images is as follows:

和

x₁、x₂分别为重叠区域两侧边界的横坐标。x为对应融合点的横坐标。t为两相邻图像重叠区域在对应融合点上的灰度差阈值。Wherein, I ₁ (x, y) and I ₂ (x, y) are the pixel values of the corresponding fusion points in the overlapping area of two adjacent bridge deck images respectively. d ₁ and d ₂ are respectively the gradient weight factors of two adjacent bridge deck images at the corresponding fusion points.

and

x ₁ and x ₂ are the abscissas of the borders on both sides of the overlapping area, respectively. x is the abscissa of the corresponding fusion point. t is the gray level difference threshold of the overlapping area of two adjacent images at the corresponding fusion point.

As preferably, the method for collecting images in step 1 is specifically as follows:

(1) After calculating the internal parameter matrix of the CCD camera by Zhang Zhengyou's plane calibration method, the radial distortion coefficient is obtained by the least square method.

(2) Install the CCD camera calibrated in step 1-1 on the bridge inspection platform, and collect images of the complete bridge deck according to the preset shooting track. The preset shooting track is S-shaped.

(3) Perform image calibration on the bridge deck images collected in step 1-2 according to the CCD camera internal reference matrix and distortion coefficient obtained in step 1-1.

As a preference, in step 2, the process of solving the projection transformation matrix by the RANSAC algorithm is as follows:

(1) Construct the initial sample set S with each matching point pair in the feature descriptor. Calculate the Euclidean distance between each pair of matching points in the initial sample set S, and sort them from small to large.

(2) Take the first 85% matching point pairs of the sequence obtained in step (1) to construct a new sample set S'.

(3) Randomly select 4 groups of matching point pairs from the new sample set S′ to form an inlier set S _i , and calculate the H _i of the matrix model inlier set S _i , and proceed to step (4).

(4) The other matching point pairs in the new sample set S' are tested for the adaptability of the matrix model H _i . If there is a matching point whose verification error is smaller than the error threshold, add the pair of matching points whose verification error is smaller than the threshold to the inlier set S _i , and perform step (5). Otherwise, discard the matrix model H _i and execute (3) again.

(5) If the number of elements in the interior point set S _i is greater than the specified threshold, it is considered that a reasonable parameter model is obtained, and the matrix model H _i is recalculated for the updated interior point set S _i , and the cost function is minimized using the LM algorithm. Otherwise, discard the matrix model H _i and re-execute step (3).

(6) Repeat steps (3) to (5) l times, where l is the maximum number of iterations. Afterwards, comparing the interior point sets S _i obtained in the l iterations, the interior point set S _i with the largest number of elements is used as the final interior point set, and the calculated matrix model H _i is taken as the distance between adjacent bridge deck images. Projection transformation matrix.

As preferably, in step 3, the specific steps of projection transformation are as follows:

(1) According to the transitivity of the projection transformation matrix between adjacent images, the first bridge deck image of each row is used as the reference image of the corresponding row for splicing. The transfer transformation is performed on the transformation matrix H _ii-1 between each adjacent bridge deck image to obtain the transfer transformation matrix H _i1 between each bridge deck image and the reference image. Then, the corresponding bridge deck images are respectively mapped to the reference plane coordinate system through each transformation matrix H _i1 , so as to complete image splicing and fusion between adjacent images in the horizontal direction, and form multiple wide-angle horizontal panoramic images Image _i .

(2) The first horizontal panoramic image Image ₁ obtained in step (1) is used as the reference panoramic image for splicing. The transfer transformation matrix T _jj-1 between each horizontal panoramic image is performed to obtain the transfer transformation matrix T _j1 between each horizontal panoramic image Image _i and the reference panoramic image. Then the corresponding horizontal panoramic images are respectively mapped to the reference plane coordinate system through each transfer transformation matrix T _j1 to complete the image splicing and fusion between adjacent horizontal panoramic images in the vertical direction to form the final bridge deck panoramic image.

The beneficial effects that the present invention has are:

1. The present invention replaces human eyes to complete automatic non-destructive detection of bridge damage characteristics through image acquisition and processing technology, which has very important practical significance for the research of bridge deck damage detection technology in complex terrain environments. On the one hand, the construction safety is enhanced, and on the other hand, the operation mobility and flexibility are improved.

2. Aiming at the problems of large amount of calculation and insufficient precision of traditional image registration algorithms, the present invention proposes an improved image registration algorithm for multiple groups of adjacent bridge decks in order to extract bridge deck image disease feature data more completely and accurately , to achieve fidelity stitching of large-area bridge deck images, improve the accuracy and efficiency of image registration, lay a working foundation for subsequent bridge disease feature image detection, and provide a technical reference for image stitching detection in other fields.

3. For the special detection environment of the bridge surface, the present invention proposes an improved gradual-in and gradual-out image fusion algorithm, and introduces the gray level difference threshold of adjacent bridge deck images, which can effectively suppress the influence of irrelevant noise on the bridge deck image and preserve the The detailed feature information of the bridge deck disease can improve the signal-to-noise ratio of the stitched image on the basis of realizing the fidelity fusion of multiple groups of bridge deck images.

4. The present invention improves the anti-interference and stability of the bridge deck image mosaic algorithm under complex background, and has better robustness. Ensure the accuracy and precision of subsequent disease feature data extraction.

5. According to the working environment of bridge surface image collection, the present invention designs multiple groups of bridge deck image mosaic algorithms, and conducts reliability analysis and research on key calculations. It has high algorithm innovation and bridge inspection engineering reference value.

Description of drawings

Fig. 1 is the mosaic flowchart of large area bridge deck image in the present invention;

Fig. 2 is a schematic diagram of bridge deck image acquisition track in the present invention;

Fig. 3 is a flow chart of splicing processing of multiple bridge images in the present invention;

Fig. 4 is a flow chart of adaptive contrast threshold calculation in the present invention;

Fig. 5 is a schematic diagram of a progressive image mosaic strategy in the present invention;

Fig. 6 is a schematic diagram of the column-by-column image mosaic strategy in the present invention;

7a and 7b are schematic diagrams of weighted fusion between adjacent images in the present invention.

Detailed ways

The present invention will be further described below in conjunction with accompanying drawing.

As shown in Figure 1, a large-area bridge deck image mosaic method, the specific steps are as follows:

Step 1. Stitching preprocessing

1-1. Using Zhang Zhengyou’s plane calibration method, use the CCD camera to sample images of the calibration plate from different angles, calculate the internal parameter matrix of the CCD camera through the checkerboard corner coordinates of the detected calibration plate, and obtain the radial distortion coefficient by the least square method .

1-2. Place the CCD camera calibrated in step 1-1 on the bridge inspection platform, and collect multiple sets of images of the complete bridge deck according to the preset shooting trajectory. The preset shooting trajectory is shown in Figure 2, which is S-shaped.

1-3. According to the internal reference matrix and distortion coefficient of the CCD camera obtained in step 1-1, perform image calibration on each bridge deck image collected in step 1-2, so as to eliminate the distortion effect caused by lens distortion.

1-4. Image preprocessing

First pre-estimate the size of the panorama image. The size is determined according to the resolution and quantity of the images to be stitched, and the invalid area is removed after the stitching is completed; then, by extracting and counting the brightness component information of all the bridge deck images to be stitched, and respectively performing a calculation on the brightness component information of each bridge deck image Equalization to eliminate the influence of brightness difference caused by uneven illumination; finally, the bridge deck images are transformed into the frequency domain by Fourier transform, and the phase information of the normalized cross power spectrum in the phase correlation algorithm is used to obtain the difference between the images. The translation parameters of , so as to complete the pre-estimation of the overlapping area between adjacent images.

The phase correlation algorithm uses Fourier transform to transform the image to be stitched into the frequency domain, and then calculates the translation parameters of the two images through the normalized cross-power spectrum to obtain a two-dimensional impulse function: the peak value of the two-dimensional impulse function The size reflects the content correlation between adjacent bridge deck images, and its value of 1 means that the two images are completely the same, and 0 means that they are completely different. There are changes caused by perspective transformation and position movement between adjacent images collected by bridge image detection equipment. Although the energy of the impulse function will be dispersed from a single peak to many small peaks, the translation parameter corresponding to the maximum peak position will remain relatively stable. Therefore, the translation obtained by the phase correlation algorithm can roughly obtain the overlapping area between the images to be stitched, and the algorithm is not sensitive to changes in illumination brightness, and the maximum correlation peak detected has good robustness and stability.

Step 2, image registration

First, SIFT feature points are extracted in the overlapping area between adjacent images to reduce a large amount of unnecessary feature point detection calculations and improve the detection efficiency of SIFT feature points; then, through the adaptive contrast threshold method, the detected SIFT feature points The number of points is controlled within a reasonable range to screen out a stable set of feature points; and the improved RANSAC algorithm (random sampling consensus algorithm) is used to calculate the projection transformation matrix H between adjacent images.

The adaptive contrast threshold method is as follows:

After determining the overlapping area (Δx, Δy) between two adjacent bridge deck images, SIFT feature point detection is only performed on the overlapping area. Since the feature points with low contrast are more sensitive to the background noise of the bridge deck, the contrast threshold is set to filter out a stable set of feature points, which is denoted as C.

In the prior art, the contrast of each SIFT feature point is calculated through the Gaussian difference Taylor expansion, and a fixed contrast threshold is set to retain feature points higher than the contrast threshold as stable feature points. However, the above-mentioned contrast threshold T _c is a fixed value, and generally takes a value between 0.02 and 0.04. But in the crack image detection of different concrete bridges, the candidate feature point sets detected by SIFT are very different. If there are fewer feature points, it may not be able to meet the number of feature point matching requirements, which will affect the final splicing accuracy. (contrast thresholding step used in traditional stitching algorithms). The present invention sets a changing contrast threshold to ensure that the number of detected SIFT feature points is controlled within a reasonable range. After several sets of experiments, it is shown that the feature points of bridge crack image detection can be kept between 200 and 300 to meet better splicing accuracy.

As shown in Figure 4, the method for determining the contrast threshold _Tc among the present invention is specifically as follows:

(1) Set the lower limit of the number of feature points N _min =200, the upper limit of N _max =300, the contrast threshold T _c =T ₀ ; T ₀ is the initial threshold, and the value is 0.02-0.04.

(2) Perform feature point detection, and count the number N of feature points whose contrast is higher than _Tc .

(3) If N _min ≤ N ≤ N _max , then include the feature points with a contrast higher than T _c into the initial matching point set, remove the feature points with a contrast lower than the threshold T _c , and directly enter step (5); otherwise, execute Step (4).

并执行步骤(3)；若N＞N_max，则将对比度阈值增大为原数值的2倍，并执行步骤(3)。(4) If N<N _min , then reduce the contrast threshold T _c to the original value

And execute step (3); if N>N _max , increase the contrast threshold to twice the original value, and execute step (3).

(5) Eliminate false feature points in the initial matching point set by the method of nearest neighbor and second nearest neighbor, and generate a feature descriptor. The feature descriptor contains multiple matching point pairs consisting of paired feature points, as well as the distance and direction information between each matching point pair.

After the matching of feature points between the overlapping regions of adjacent images, enough matching point pairs are selected, and the transformation matrix between bridge crack sequence images is solved by matching point pairs, so as to complete the large-scale bridge deck image mosaic. In order to further improve the efficiency and accuracy of image registration, the RANSAC algorithm is improved.

The process of solving the projection transformation matrix H by the improved RANSAC algorithm is as follows:

(1) Construct the initial sample set S with each matching point pair in the feature descriptor. Count the Euclidean distance between each pair of matching points in the initial sample set S, and sort them from small to large;

(2) Get the first 85% matching points of the sequence obtained in step (1) to construct a new sample set S';

(3) Randomly select 4 groups of matching point pairs from the new sample set S′ to form an interior point set S _i , and calculate the H _i of the matrix model interior point set S _i , and enter step (4);

(4) The other matching point pairs in the new sample set S′ carry out adaptive testing for the matrix model H _i ; if there is a matching point whose test error is less than the error threshold, then add the matching point pair whose test error is less than the threshold to the inner point set S _i , and execute step (5); otherwise, discard the matrix model H _i and execute (3) again.

(5) If the number of elements in the interior point set S _i is greater than the specified threshold, it is considered that a reasonable parameter model is obtained, and the matrix model H _i is recalculated for the updated interior point set S _i , and the cost function is minimized using the LM algorithm; Otherwise, discard the matrix model H _i and re-execute step (3).

(6) Repeat steps (3) to (5) l times, where l is the maximum number of iterations. Afterwards, comparing the interior point sets S _i obtained in the l iterations, the interior point set S _i with the largest number of elements is used as the final interior point set, and the calculated matrix model H _i is taken as the distance between adjacent bridge deck images. Projective transformation matrix H.

The improved RANSAC algorithm calculates the Euclidean distance between all matching point pairs and performs sorting and screening, which not only reduces the sample set data of the point pairs to be matched, increases the proportion of internal points in the sample set, but also reduces the iteration of the projection transformation matrix The number of refinements is used to improve the matching accuracy of bridge deck images. According to the feature that the smaller the distance between the image feature point pairs, the higher the matching similarity, the Euclidean distance between all feature point pairs is calculated here and sorted in order from small to large for screening. According to the statistics of multiple groups of bridge deck image splicing test results, after the initial matching of feature point pairs in overlapping areas, the successful matching rate of the initial sample set S can reach more than 85%, and then the first 85% of the sequence of feature point pairs are used to construct a new model. Sample set S'. After screening the sample data, the sample set S' contains enough matching point pairs, which not only increases the proportion of inliers in the sample set, but also greatly reduces the number of iterations of the transformation matrix parameter model H.

Step 3, image fusion

First, according to the projection transformation matrix between adjacent images, the corresponding bridge deck image is projected and transformed; then the RGB three-color channels of each adjacent bridge deck image are weighted and smoothed by using the gradual in and gradual out fusion algorithm, and the bridge deck mosaic is obtained image.

The specific steps of projection transformation are as follows:

(1) As shown in Figure 5, according to the transitivity of the projection transformation matrix between adjacent images, the first bridge deck image of each row is used as the reference image of the corresponding row, and stitching is performed according to the image row stitching strategy. Perform transfer transformation on the transformation matrix H _ii-1 between adjacent bridge deck images to obtain the transfer transformation matrix H _i1 between each bridge deck image and the reference image; then transform the corresponding bridge deck images through each transformation matrix H _i1 Mapped to the reference plane coordinate system to complete image splicing and fusion between adjacent images in the horizontal direction to form multiple horizontal panoramic images Image _i with wide viewing angles.

The transformation formulas of each transfer matrix are as follows:

H ₂₁ =H ₂₁

H ₃₁ =H ₃₂ ×H ₂₁

H _n1 =H _nn-1 ×H _n-1n-2 ×...×H ₂₁

Among them, H _ii-1 is the transformation matrix between the i-1th bridge deck image and the i-th bridge deck image in the same row, and its value is calculated in step 2-2; H _i1 is the first image in the same row The transformation matrix between the bridge deck image and the i-th bridge deck image; n is the number of images on the same row.

(2) As shown in Fig. 6, the first horizontal panoramic image Image ₁ obtained in step (1) is used as a reference panoramic image, and stitching is performed according to the image sequence stitching strategy. Carry out transfer transformation to the transformation matrix Tjj _{- 1} between each horizontal panoramic image, obtain the transfer transformation matrix T _j1 between each horizontal panoramic image Image _i and the reference panoramic image; The horizontal panoramic images are respectively mapped to the reference plane coordinate system to complete the image splicing and fusion between adjacent horizontal panoramic images in the vertical direction. In the image splicing and fusion, the transfer matrix transformation formula refers to the description in step (1) to form the final Panoramic image of the bridge deck.

In this embodiment, the fade-in and fade-out fusion algorithm has been improved, see the following for details.

After image registration, in order to further eliminate the interference of bridge deck image splicing seams on crack detection processing, the pixel values of adjacent bridge deck images are usually weighted and averaged, as shown in Figure 7b, the pixels in the overlapping area are stitched to both sides The distance of the line is used as the basis for judging the fusion weight.

However, due to the change of the image acquisition position, the reflected light on the bridge surface may cause jumps in the gray value of individual pixels in the overlapping area. a threshold t. Calculate the gray value difference corresponding to the target pixel point in the overlapping part in the two original images. If the difference value is less than the threshold value, it means that the pixel point has no obvious difference in the original bridge deck image, and its weighted average value can be directly taken as the point On the contrary, it means that the image to be spliced has a sudden change of light and dark at the pixel position, and the pixel value with a larger weight before smoothing should be taken as the fusion pixel value of this point.

In the fade-in and fade-out fusion algorithm, the fade-in and fade-out weighting formula of each fusion pixel value I(x, y) in the overlapping area of adjacent images is as follows:

和

x₁、x₂分别为重叠区域两侧边界的横坐标；x为对应融合点的横坐标；t为两相邻图像重叠区域在对应融合点上的灰度差阈值。Wherein, I ₁ (x, y) and I ₂ (x, y) are respectively the pixel values of the corresponding fusion points in the overlapping area of two adjacent bridge deck images, as shown in FIG. 7 a . d ₁ and d ₂ are the gradient weight factors of two adjacent bridge deck images at the corresponding fusion points; as shown in Figure 7b,

and

x ₁ and x ₂ are the abscissas of the boundaries on both sides of the overlapping area; x is the abscissa of the corresponding fusion point; t is the gray level difference threshold of the overlapping area of two adjacent images at the corresponding fusion point.

Claims (4)

1. A large-area bridge deck image splicing method is characterized by comprising the following steps: step 1, acquiring images of a detected bridge deck one by one to obtain a bridge deck image set; then, extracting and counting the brightness component information of each bridge deck image in the bridge deck image set, and respectively equalizing the brightness component information of each bridge deck image; then transforming the bridge deck images into a frequency domain through Fourier transform, obtaining translation parameters among the images by adopting phase information of a normalized cross-power spectrum in a phase correlation algorithm, and completing pre-estimation of an overlapping area among adjacent images;

step 2, image registration

Firstly, extracting SIFT feature points in an overlapping area between every two adjacent images; then sifting SIFT feature points by a self-adaptive contrast threshold method to obtain a feature descriptor consisting of matching point pairs; calculating a projection transformation matrix between every two adjacent images by adopting a random sampling consistency algorithm;

the adaptive contrast threshold method is specifically as follows:

(1) Setting a lower limit N of the number of characteristic points _min Upper limit N =200 _max =300, contrast threshold T _c ＝T ₀ ；T ₀ The initial threshold value is 0.02-0.04;

(2) Detecting the characteristic points and counting that the contrast is higher than T _c The number of feature points N;

(3) If N is present _min ≤N≤N _max Then contrast will be higher than T _c The characteristic points are brought into the initial matching point set, and the rejection contrast is lower than a threshold value T _c And directly entering the step (5); otherwise, executing the step (4);

(4) If N is less than N _min Then the contrast threshold T is set _c Reduced to the original value

And executing the step (3); if N > N _max Increasing the contrast threshold to 2 times of the original value and executing the step (3);

(5) Rejecting error feature points in the initial matching point set by a nearest neighbor comparison neighbor method, and generating a feature descriptor; the feature descriptor contains a plurality of matching point pairs consisting of paired feature points and distance and direction information between the matching point pairs;

step 3, image fusion

Firstly, performing projection transformation on corresponding bridge deck images according to projection transformation matrixes between adjacent images; then, respectively carrying out weighted smooth transition on RGB three-color channels of each adjacent bridge deck image by adopting a gradual-in and gradual-out fusion algorithm to obtain a bridge deck splicing image;

in the fade-in and fade-out fusion algorithm, the fade-in and fade-out weighting formula of each fusion point pixel value I (x, y) in the overlapping region of the adjacent images is as follows:

wherein, I ₁ (x,y)、I ₂ (x, y) are pixel values of corresponding fusion points of two adjacent bridge deck images in the overlapping area respectively; d is a radical of ₁ 、d ₂ Respectively is a gradual change weight factor of two adjacent bridge deck images at the corresponding fusion point;

and

x ₁ 、x ₂ respectively are the horizontal coordinates of the boundaries at the two sides of the overlapping area; x is the abscissa of the corresponding fusion point; and t is the gray level difference threshold value of the overlapped area of the two adjacent images on the corresponding fusion point.

2. The large-area bridge deck image splicing method according to claim 1, wherein: the method for acquiring the image in the step 1 specifically comprises the following steps:

(1) Calculating an internal reference matrix of the CCD camera by adopting a Zhang Zhengyou plane calibration method, and then obtaining a radial distortion coefficient by adopting a least square method;

(2) Arranging the CCD camera calibrated in the step (1) on a bridge detection platform, and acquiring an image of the complete bridge floor according to a preset shooting track; the preset shooting track is S-shaped;

(3) And (3) respectively carrying out image calibration on each bridge deck image acquired in the step (2) according to the internal reference matrix and the distortion coefficient of the CCD camera obtained in the step (1).

3. The large-area bridge deck image splicing method according to claim 1, wherein the method comprises the following steps: in step 2, the process of solving the projective transformation matrix by the random sampling consistency algorithm is as follows:

(1) Constructing an initial sample set S by using each matching point pair in the feature descriptor; counting Euclidean distances between each matching point pair in the initial sample set S, and sequencing from small to large;

(2) Taking the first 85% of matching point pairs of the sequence obtained in the step (1) to construct a new sample set S';

(3) Randomly extracting 4 groups of matching point pairs from the new sample set S' to form an inner point set S _i And calculating an interior point set S _i Matrix model H of _i Entering the step (4);

(4) The rest matching point pairs in the new sample set S' are corresponding to the matrix model H _i Carrying out adaptability test; if the matched points with the detection errors smaller than the error threshold exist, adding the matched point pairs with the detection errors smaller than the threshold into the inner point set S _i And executing the step (5); otherwise, the matrix model H is discarded _i Re-executing (3);

(5) If inner point set S _i If the number of the middle elements is larger than the specified threshold value, a reasonable parameter model is considered to be obtained, and the updated interior point set S is subjected to _i Recalculating matrix model H _i Minimizing a cost function by using an optimization learning algorithm; otherwise, the matrix model H is discarded _i And re-executing the step (3);

(6) Repeating steps (3) to (5) for l times, wherein l is the maximum iteration number; then, comparing the inner point set S obtained in the iteration for one time _i Set S of interior points with the largest number of elements _i Taking the matrix model H as the final internal point set and calculating _i As a projective transformation matrix between adjacent bridge deck images.

4. The large-area bridge deck image splicing method according to claim 1, wherein the method comprises the following steps: in step 3, the projection transformation comprises the following specific steps:

(1) According to the transmissibility of the projective transformation matrix between the adjacent images, the first bridge deck image of each row is respectively used as a reference image of the corresponding row for splicing; for the transformation matrix H between the adjacent bridge floor images _ii-1 Carrying out transmission transformation to obtain a transmission transformation matrix H between each bridge deck image and the reference image _i1 (ii) a Then through each transformation matrix H _i1 Mapping the corresponding bridge deck images into a reference plane coordinate system respectively to finish Image splicing and fusion between every two adjacent images in the horizontal direction to form a plurality of transverse panoramic Image images with wide visual angles _i ；

(2) The first horizontal panoramic Image obtained in the step (1) is processed ₁ Splicing as a reference panoramic image; for transformation matrix T between transverse panoramic images _jj-1 Performing transfer transformation to obtain horizontal panoramic images _i Transfer transformation matrix T with reference panoramic image _j1 (ii) a Transforming the matrix T by each transfer _j1 And respectively mapping the corresponding transverse panoramic images into a reference plane coordinate system to complete image splicing and fusion between every two adjacent transverse panoramic images in the vertical direction to form a final bridge deck panoramic image.

Images