Construction progress monitoring method for highway construction project monitoring platform

Abstract

The present invention relates to the field of image communication technology, and in particular to a construction progress monitoring method for a highway construction project monitoring platform, comprising: using a segmentation threshold to segment a differential image between a target image and a reference image to obtain a target pixel on the target image and a reference pixel on the reference image; grayscale values ​​of the target pixel are grayscale transformed, the sparse fusion degree of the transformation curve used and the historical curve is the largest, the sparse fusion degree is obtained by the sparsity of the difference between the grayscale value after the grayscale transformation and the grayscale value of the reference pixel, the similarity between the transformation curve and the historical curve, and the similarity of the grayscale distribution of the target pixel at the first two positions, grayscale transformation is performed on the target image using the transformation curve, and the target image is compressed and transmitted using the difference between the target image after the grayscale transformation and the reference image; the segmentation threshold is corrected according to the error of the transformation curve. The present invention increases the compression transmission efficiency of the target image and saves network resources.

Description

Technical Field

[0001] The present invention relates to the technical field of image communication, and in particular to a construction progress monitoring method for a highway construction project monitoring platform. Background Art

[0002] During the highway construction process, it is necessary to monitor the construction progress in a timely manner. The existing monitoring method is to establish a monitoring platform based on the Internet of Things. The construction workers at the construction site transmit the data related to the construction progress to the monitoring platform to realize online monitoring of the construction progress. The data related to the construction progress includes the data of construction materials such as asphalt sand and gravel, measurement data (such as roadbed height, pavement compaction, etc.) and monitoring data. Among them, the monitoring data is the data that can directly and intuitively reflect the construction progress. However, when transmitting the monitoring data to the monitoring platform, the monitoring image needs to be compressed and transmitted. Considering that the network resources in the monitoring platform based on the Internet of Things are limited, the existing image compression algorithm does not have enough compression strength or compression efficiency when compressing the monitoring images in the monitoring construction, resulting in more network resources occupied and slow transmission efficiency. Summary of the invention

[0003] In order to solve the above problems, the present invention provides a construction progress monitoring method for a highway construction project monitoring platform.

[0004] The construction progress monitoring method for a highway construction project monitoring platform of the present invention adopts the following technical scheme:

[0005] An embodiment of the present invention provides a construction progress monitoring method for a highway construction project monitoring platform, the method comprising the following steps:

[0006] Obtain a target image and a reference image at each position, where the reference image is acquired before the target image;

[0007] H1: compress and transmit the target image according to the target image and reference image at each position and in combination with the segmentation threshold, including:

[0008] The difference image between the target image and the reference image is segmented by using the segmentation threshold to obtain the target pixel on the target image and the reference pixel on the reference image; the gray value of the target pixel is transformed into gray, and the transformation curve used in the gray transformation makes the sparsity of the difference between the gray value after the gray transformation and the gray value of the reference pixel maximized; the target image is transformed into gray by using the transformation curve, and the target image is compressed and transmitted by using the difference between the target image after the gray transformation and the reference image; the segmentation threshold is corrected according to the error of the transformation curve;

[0009] H2: Use the transformation curve obtained in H1 as the history curve; obtain the target image and reference image at the next position; according to the method in H1, compress and transmit the target image at the next position based on the target image and reference image at the next position and the segmentation threshold corrected in H1;

[0010] The transformation curve used in grayscale transformation in H2 has the greatest similarity to the historical curve;

[0011] H3: Obtain the similarity of the grayscale distribution of the target pixel in H2 and the target pixel in H1; use the transformation curve obtained in H2 as the history curve; continue to obtain the target image and reference image at the next position; according to the method in H1, compress and transmit the target image at the next position in H3 based on the target image and reference image at the next position and combined with the corrected segmentation threshold in H2;

[0012] The degree of sparse fusion between the transformation curve used in the grayscale transformation in H3 and the historical curve is the greatest, and the degree of sparse fusion is obtained by the sparsity of the difference between the grayscale value after the grayscale transformation in H3 and the grayscale value of the reference pixel, the similarity between the transformation curve used in the grayscale transformation in H3 and the historical curve, and the similarity of the grayscale distribution.

[0013] Preferably, the transformation curve used in the grayscale transformation maximizes the sparsity of the difference between the grayscale value after the grayscale transformation and the grayscale value of the reference pixel, and includes the following specific steps:

[0014] For the nonlinear function f(x), x represents the grayscale value of any target pixel, and the result output by f(x) represents the grayscale value after grayscale transformation; the one-dimensional vector composed of all the undetermined coefficients of the nonlinear function f(x) is recorded as the transformation feature, and the transformation feature is randomly initialized; the absolute value of the difference between the grayscale value of each target pixel after grayscale transformation and the grayscale value of the reference pixel is obtained, and recorded as the transformation difference of each target pixel;

[0015] The sparsity of the transformation difference is obtained according to the transformation difference of all target pixel points; the transformation feature when the sparsity is maximum is obtained; and the nonlinear function corresponding to the transformation feature when the sparsity is maximum is used as the transformation curve used in the grayscale transformation.

[0016] Preferably, the step of obtaining the sparsity of the transformation difference according to the transformation differences of all target pixels comprises the following specific steps:

[0017] If the transformation difference of the target pixel point is less than or equal to the first preset threshold, then the transformation difference of the target pixel point is set to 0; the number of target pixels is recorded as N1, the number of target pixels with transformation difference equal to zero is obtained N2, and the ratio of N2 to N1 is recorded as the sparsity of the transformation difference.

[0018] Preferably, the transformation curve used in the grayscale transformation in H2 has the greatest similarity to the historical curve, and the specific steps included are as follows:

[0019] When H2 is executed according to the method in H1, the sparsity of the difference between the grayscale value after grayscale transformation and the grayscale value of the reference pixel is obtained in H2, which is recorded as M1, and the cosine similarity between the transformation feature of the history curve and the transformation feature of the nonlinear function in H2 is recorded as M2; let F=M1+M2; obtain the transformation feature when F is maximum;

[0020] The nonlinear function corresponding to the transformation feature when F is maximum is used as the transformation curve used for the grayscale transformation obtained in H2.

[0021] Preferably, the sparse fusion degree of the transformation curve used in the grayscale transformation in H3 and the historical curve is the largest, and the sparse fusion degree is obtained by the sparsity of the difference between the grayscale value after the grayscale transformation in H3 and the grayscale value of the reference pixel, the similarity between the transformation curve used in the grayscale transformation in H3 and the historical curve, and the similarity of the grayscale distribution, and the specific steps included are as follows:

[0022] When H3 is executed according to the method in H1, the sparsity of the difference between the grayscale value after grayscale transformation and the grayscale value of the reference pixel point is obtained in H3, which is recorded as M3, and the cosine similarity between the transformation feature of the history curve and the transformation feature of the nonlinear function in H3 is recorded as M4; the grayscale distribution similarity is recorded as Q;

[0023] Let the sparse fusion degree M = Q×M4+(1-Q)×M3; the nonlinear function corresponding to the transformation feature when the sparse fusion degree M is maximum is used as the transformation curve used for the grayscale transformation obtained in H3.

[0024] Preferably, the step of performing background segmentation on the differential image between the target image and the reference image by using the segmentation threshold to obtain target pixels on the target image and reference pixels on the reference image includes the following specific steps:

[0025] The target image is subtracted from the reference image to obtain a differential image; the area in the differential image where the absolute value of the grayscale value is less than or equal to the segmentation threshold is recorded as the background area; the pixel points in the background area on the target image are recorded as target pixels, and the pixel points in the background area on the reference image are recorded as reference pixels.

[0026] Preferably, the method of compressing and transmitting the target image by using the difference between the target image after grayscale transformation and the reference image comprises the following specific steps:

[0027] The difference image between the transformed target image and the reference image is recorded as the image to be transmitted; the image to be transmitted is compressed using the LZW algorithm and then transmitted to the highway construction project monitoring platform.

[0028] Preferably, the segmentation threshold value is corrected according to the error of the transformation curve, and the specific steps include the following:

[0029] The grayscale value of each target pixel is transformed using the transformation curve. The transformed grayscale value is recorded as the sparse grayscale value of each target pixel. The standard deviation of the sparse grayscale values ​​of all target pixels and the grayscale value of the reference pixel is obtained and recorded as the error h of the transformation curve.

[0030] Obtain several target pixel points with the largest sparse difference value, which are recorded as sample pixel points; record the area outside the target pixel points on the target image as the foreground area; obtain the connected domain formed by the sample pixel points and having a coincident boundary with the foreground area, which is recorded as the target connected domain; on the coincident boundary between the target connected domain and the foreground area, record the neighboring pixel points of all the pixel points on the boundary as local target pixel points; record the pixel points in the reference image at the same position as the local target pixel points as local reference pixel points;

[0031] The grayscale value of each local target pixel is transformed by using the transformation curve, and the absolute value of the difference between the grayscale value of each local target pixel after the grayscale transformation and the grayscale value of the local reference pixel at the same position is obtained, which is recorded as the transformation difference of each local target pixel; the sparsity w of the transformation difference is obtained according to the transformation differences of all local target pixels;

[0032] The segmentation threshold is modified according to the error h and sparsity w of the transformation curve.

[0033] Preferably, the step of correcting the segmentation threshold according to the error h and the sparsity w of the transformation curve includes the following specific steps:

[0034] The segmentation threshold before correction is recorded as T, and (1+p×h)×T is taken as the segmentation threshold after correction;

[0035] Among them, when w is greater than th21, p=w-th21, when w is less than th22, p=w-th22, when w is greater than or equal to th22 and less than or equal to th21, p=0;

[0036] th21 and th22 are preset thresholds, and th22 is greater than th21.

[0037] Preferably, the step of obtaining the grayscale distribution similarity between the target pixel in H2 and the target pixel in H1 includes the following specific steps:

[0038] All target pixel points obtained in H1 are recorded as Q1, and all target pixel points obtained in H2 are recorded as Q2; the one-dimensional vector composed of the mean and variance of the grayscale values ​​of all target pixel points in Q1 is recorded as the first distribution feature, and the one-dimensional vector composed of the mean and variance of the grayscale values ​​of all target pixel points in Q2 is recorded as the second distribution feature, and the cosine similarity between the first distribution feature and the second distribution feature is recorded as the grayscale distribution similarity.

[0039] The beneficial effects of the technical solution of the present invention are:

[0040] The difference image between the target image and the reference image is segmented by using a segmentation threshold to obtain a target pixel on the target image and a reference pixel on the reference image; the grayscale value of the target pixel is transformed into grayscale, and the transformation curve used in the grayscale transformation maximizes the sparsity of the difference between the grayscale value after the grayscale transformation and the grayscale value of the reference pixel; the target image is transformed into grayscale using the transformation curve, and the target image is compressed and transmitted using the difference between the target image and the reference image after the grayscale transformation. In this process, the present invention performs grayscale transformation on the construction unchanged area in the images (i.e., the target image and the reference image) collected at different times, and at the same time maximizes the sparsity of the difference between the grayscale value after the grayscale transformation and the grayscale value of the reference pixel, thereby ensuring that the difference between the target image and the reference image after the grayscale transformation can be easily and efficiently compressed, reducing the amount of data during transmission and reducing the occupation of network resources to a certain extent.

[0041] Furthermore, the segmentation threshold is corrected according to the error of the transformation curve; the target image and reference image of the next position are obtained; the target image of the next position is compressed and transmitted according to the target image and reference image of the next position and in combination with the corrected segmentation threshold, and the similarity between the transformation curve used in the grayscale transformation and the historical curve in this process is the greatest. In this process of the present invention, the situation where the background area where the target pixel point is located cannot accurately include the construction unchanged area when the above-mentioned segmentation threshold is not selected properly is avoided; at the same time, the historical curve (that is, the transformation situation when the target pixel point at the previous position is transformed into grayscale) is integrated to avoid the situation where the transformation curve has a large error, so that the compression strength of the target image is improved, and the transmission efficiency is further improved.

[0042] Furthermore, the target image and reference image of the next position are continuously acquired; the target image of the next position is compressed and transmitted according to the target image and reference image of the next position and in combination with the segmentation threshold value that has been corrected again; wherein the sparse fusion degree of the transformation curve used in the grayscale transformation and the historical curve is the greatest. The purpose of this process of the present invention is: when the first two positions are segmented to acquire the target pixel points, if there is an over-segmentation or under-segmentation situation (that is, the segmentation threshold value is too large or too small), it is possible to avoid introducing too much or too little of the transformation curves used in the first two compression transmissions into the current compression transmission process (that is, the next position in H3), further ensuring the compression transmission efficiency and saving network transmission resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0044] Figure 1 A flowchart of the steps of a construction progress monitoring method for a highway construction project monitoring platform provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the construction progress monitoring method for a highway construction project monitoring platform proposed by the present invention, its specific implementation method, structure, features and effects, in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0046] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0047] The specific scheme of the construction progress monitoring method for a highway construction project monitoring platform provided by the present invention is described in detail below with reference to the accompanying drawings.

[0048] Embodiment 1:

[0049] See also Figure 1 , which shows a flowchart of the steps of a construction progress monitoring method for a highway construction project monitoring platform provided by Embodiment 1 of the present invention, the method comprising the following steps:

[0050] Step S001: Capture a target image at each position and store a reference image at each position, where the reference image is captured before the target image.

[0051] The drone inspects the highway under construction according to the set inspection track. In this embodiment, the inspection is performed once a day. The drone is equipped with a camera and a positioning device. The camera looks down and collects images of the highway under construction. In this embodiment, the collected images are grayscale images. In other embodiments, if the collected images are color images, they are grayscaled into grayscale images. Several sampling points are marked on the inspection track (for example, a sampling point is set every 3 meters), and the drone collects an image when it flies to the location of each sampling point.

[0052] At the same time, the drone is also equipped with a communication module (such as a 5G communication module), which is used to transmit the images collected at each location (that is, each sampling point) to the highway construction project monitoring platform. The images collected during multiple inspections at the same location are displayed on the monitoring platform, indicating the construction progress of each location at different times.

[0053] During each inspection process, the image of each location collected in real time is recorded as the target image, which is the image that needs to be transmitted to the highway construction project monitoring platform. The image collected at each location during the last inspection process is recorded as the reference image, and the reference image is stored on the drone.

[0054] Step S002: Use the segmentation threshold to perform background segmentation on the difference image between the target image and the reference image to obtain target pixel points on the target image and reference pixel points on the reference image.

[0055] In order to speed up the transmission speed of the drone, reduce the amount of data transmitted, save network resources and power consumption when the drone transmits data, it is necessary to compress the target image. Although the existing compression technology, such as LZW compression technology, can achieve efficient compression of image and other data, its compression efficiency is not high enough for the image of the highway under construction (that is, the target image). In this embodiment, considering that during the construction of the highway, the highway and some areas around it may not change significantly, this embodiment compresses the image by the difference of images collected at different times in this part of the area (that is, the difference between the target image and the reference image), thereby speeding up the transmission speed of the drone, reducing the amount of data transmitted, saving network resources and power consumption when the drone transmits data.

[0056] First, the target image is subtracted from the reference image to obtain a differential image (that is, the grayscale value of each pixel in the target image is subtracted from the grayscale value of the pixel at the same position in the reference image to obtain the grayscale value of each pixel in the differential image). The smaller the absolute value of the grayscale value of the pixel in certain areas of the differential image, the less obvious changes there are in the construction scene in that area.

[0057] A segmentation threshold is preset, and the difference image is segmented using the threshold segmentation method. The specific method is: in the difference image, all pixels whose absolute grayscale values ​​are less than or equal to the segmentation threshold are obtained, and the area formed by these pixels is recorded as the background area. The pixels in the background area of ​​the target image are recorded as target pixels, and the pixels in the background area of ​​the reference image are recorded as reference pixels.

[0058] In this embodiment, the initial value of the segmentation threshold is set to 50. When the area of ​​the background region is less than 25, the initial value of the segmentation threshold is enlarged by 10%, and then the background region, target pixel points and reference pixel points are re-obtained.

[0059] The difference between the grayscale value of the target pixel and the grayscale value of the reference pixel is relatively small, which may indicate a location where the construction scene has not changed significantly.

[0060] Step S003: grayscale transform the grayscale value of the target pixel point, and the transformation curve used in the grayscale transformation makes the sparsity of the difference between the grayscale value after the grayscale transformation and the grayscale value of the reference pixel point maximum.

[0061] This embodiment takes into account that the grayscale value of the target pixel is not completely consistent with the grayscale value of the reference pixel. For example, when collecting the target image and the reference image, the weather environment, lighting conditions, and even dust interference conditions of the construction site are different, which causes the grayscale value of the target pixel to be not completely consistent with the grayscale value of the reference pixel.

[0062] This embodiment performs grayscale transformation on the grayscale value of the target pixel point. The grayscale transformation refers to mapping the grayscale value of the target pixel point to another grayscale value using a transformation function. The commonly used grayscale transformation method is linear grayscale transformation, and linear grayscale transformation uses a linear function. This embodiment uses a nonlinear function, such as a 5th-order polynomial function. The nonlinear function is recorded as f(x), x represents the grayscale value of any target pixel point, and the result output by f(x) represents the grayscale value after grayscale transformation (that is, the grayscale value after the transformation curve is mapped). The one-dimensional vector composed of all the undetermined coefficients of the nonlinear function f(x) is recorded as the transformation feature, and the transformation feature is randomly initialized. The absolute value of the difference between the grayscale value of each target pixel point after grayscale transformation and the grayscale value of the reference pixel point at the same position is obtained, which is simply recorded as the transformation difference of each target pixel point.

[0063] The sparsity of the transformation difference is obtained according to the transformation differences of all target pixels.

[0064] As an example, obtaining the sparsity of the transformation difference includes the following steps:

[0065] If the transformation difference of the target pixel is less than or equal to the first preset threshold th1, then the transformation difference of the target pixel is set to 0. This embodiment is described by taking th1=2 as an example, and th1 can be set to other values ​​in other embodiments. For example, in some embodiments, if more monitoring details are required for the area where the construction scene remains unchanged when monitoring the construction site, then th1 can be set to a smaller value, such as th1=1; for another example, in other embodiments, if the details of the area where the construction scene remains unchanged are not concerned when monitoring the construction site, but the data transmission speed needs to be accelerated to a greater extent to save network resources, then the value of th1 can be set to a larger value, such as th1=5.

[0066] The number of target pixels is recorded as N1, the number N2 of target pixels whose transformation difference is zero is obtained, and the ratio of N2 to N1 is recorded as the sparsity of the transformation difference.

[0067] In the above process, each given transformation feature corresponds to the sparsity of a transformation difference (that is, the sparsity of the transformation difference is the objective function of the transformation feature). This embodiment uses a simulated annealing algorithm to obtain the transformation feature when the sparsity of the transformation difference is the maximum. The maximum sparsity means that the transformation differences of all target pixels are the most sparsely distributed (that is, the number of data values ​​0 appears is large), which is conducive to increasing the compression transmission efficiency.

[0068] The nonlinear function corresponding to the transformation feature when the sparsity is the largest is the transformation curve used in the grayscale transformation. That is, after the grayscale value of the target pixel is transformed using the transformation curve, the sparsity of the difference between the grayscale value after the grayscale transformation and the grayscale value of the reference pixel is the largest.

[0069] Step S004: Use the transformation curve to perform grayscale transformation on the target image, and use the difference between the target image after grayscale transformation and the reference image to compress and transmit the target image.

[0070] The grayscale transformation of the entire target image is performed using the transformation curve used in the grayscale transformation in step S003 to obtain the transformed target image. In the transformed target image and the reference image, the differences between the areas where the construction scene has no obvious changes are sparsely distributed, that is, the differential image between the transformed target image and the reference image (referred to as the image to be transmitted) has high sparsity and can be efficiently compressed. When the absolute value of the grayscale value of a pixel in the image to be transmitted is less than or equal to th1, the grayscale value of the pixel is set to 0. In this embodiment, the image to be transmitted is compressed using the LZW algorithm to obtain a compressed image, and the compressed image is transmitted to the highway construction project monitoring platform.

[0071] After receiving the compressed image, the highway construction project monitoring platform uses the LZW algorithm to decompress the compressed image to obtain a differential image. Since the reference image of each position has been transmitted to the highway construction project monitoring platform, the reference image of each position is summed with the differential image (that is, the grayscale values ​​of the same pixel points are summed separately) to obtain the target image of each position received by the highway construction project monitoring platform.

[0072] At this point, this step uses the above method to transmit each target image collected by the drone to the highway construction project monitoring platform for monitoring the construction progress.

[0073] It should be noted that after the compression and transmission of the target image collected by the drone at each location is completed, the reference image of the location is deleted from the drone, and the target image collected at each location is stored in the drone. This target image will be used as a reference image during the next inspection.

[0074] Step S005: Correct the segmentation threshold according to the error of the transformation curve.

[0075] It should be noted that the above transformation curve is calculated based on the target pixel and reference pixel in the background area. However, when the above segmentation threshold is not selected properly, the background area cannot accurately include the scene unchanged area (that is, the construction scene area without obvious changes), which leads to a large error in the transformation curve obtained by the target pixel and the reference pixel, and finally the compression strength of the above compressed image is not high enough, affecting the transmission efficiency. Therefore, it is necessary to correct the segmentation threshold.

[0076] As an example, the segmentation threshold is modified by:

[0077] The grayscale value of each target pixel is transformed by using the transformation curve. The transformed grayscale value is recorded as the sparse grayscale value of each target pixel. The difference between the sparse grayscale value of each target pixel and the grayscale value of each reference pixel is obtained and recorded as the sparse difference of each target pixel. The sparse differences of all target pixels are linearly normalized, and the standard deviation of the linearly normalized sparse differences is recorded as the error h of the transformation curve.

[0078] Get the K target pixels with the largest sparse difference values, record them as sample pixels, and record the area outside the background area (i.e., target pixel points) as the foreground area on the target image. Get the connected domain formed by the sample pixels and having overlapping boundaries with the foreground area, record them as the target connected domain. Get the overlapping boundaries of all target connected domains with the foreground area, record each pixel on the boundary as a boundary point, record all pixels within a D×D square window centered on the boundary point as the neighboring pixels of the boundary point, and record all the neighboring pixels of the boundary point as local target pixels. The pixels in the reference image that are at the same position as the local target pixels are recorded as local reference pixels.

[0079] In this embodiment, K is half of the number of target pixels (rounded up), and D=9; in other embodiments, K and D may be set to other values, which are not specifically limited in this embodiment.

[0080] The grayscale value of each local target pixel is transformed using the transformation curve to obtain the grayscale value of each local target pixel after the grayscale transformation. The absolute value of the difference between the grayscale value after the grayscale transformation and the grayscale value of the local reference pixel at the same position is simply recorded as the transformation difference of each local target pixel. The sparsity w of the transformation difference is obtained according to the transformation differences of all local target pixels (the specific process is the same as step S003).

[0081] The larger the sparsity w, the greater the grayscale difference after local grayscale transformation at the junction of the foreground and background areas is, which means that the background area obtained in the above steps is incomplete and the segmentation threshold needs to be adjusted larger. The smaller the sparsity w, the greater the grayscale difference after local grayscale transformation at the junction of the foreground and background areas is, which means that the background area obtained in the above steps is over-segmented and the segmentation threshold needs to be adjusted smaller.

[0082] The segmentation threshold is recorded as T, and (1+p×h)×T is used as the corrected segmentation threshold.

[0083] The larger h is, the larger the error of the transformation curve is, and the larger the correction amplitude of the segmentation threshold T is.

[0084] Among them, when w is greater than th21, p=w-th21, at this time p is greater than 0, and the larger w is, the more T needs to be adjusted; when w is less than th22, p=w-th22, at this time p is less than 0, and the smaller w is, the smaller T needs to be adjusted. When w is greater than or equal to th22 and less than or equal to th21, p=0, at this time T is not adjusted.

[0085] This embodiment is described by taking th21=0.6 and th22=0.3 as an example. In other embodiments, th21 and th22 may be set to other values, which are not specifically limited in this embodiment.

[0086] As another example, the segmentation threshold is modified, including the following method:

[0087] When h is less than or equal to th3, the segmentation threshold T is not corrected. When h is greater than th3, if the segmentation threshold T is less than or equal to 50, the segmentation threshold T is increased by 10%. When the segmentation threshold T is greater than 50, the segmentation threshold T is reduced by 10%. Wherein, th3 represents an error threshold. This embodiment is described by taking th3=0.3 as an example. In other embodiments, it can be set to other values, which are not specifically limited in this embodiment.

[0088] The accuracy of the corrected segmentation threshold in this example is relatively low, but the amount of calculation is small and the calculation speed is fast.

[0089] So far, in steps S002 to S005, the target image at each position is compressed and transmitted according to the target image and the reference image at each position and in combination with the segmentation threshold, and the segmentation threshold is corrected; the compression transmission efficiency is improved to a certain extent, saving network resources.

[0090] Step S006: Acquire the target image and the reference image at the next position, and compress and transmit the target image at the next position according to the target image and the reference image at the next position and in combination with the corrected segmentation threshold.

[0091] The transformation curve used in the grayscale transformation in step S003 is recorded as the history curve; the UAV flies to the next position (ie, the next sampling point), and collects the target image of the next position and reads out the reference image of the next position.

[0092] The target image at the next position is compressed and transmitted according to the target image and the reference image at the next position and in combination with the segmentation threshold (the segmentation threshold corrected in step S004). The specific process is the same as the method in steps S002 to S005, and the general process is:

[0093] The difference image between the target image and the reference image at the next position is obtained, and the difference image is segmented using the segmentation threshold corrected in step S004 to obtain the background area again, and the target pixel point and the reference pixel point at the next position are obtained on the target image and the reference image at the next position. The specific process is the same as step S002.

[0094] Re-transform the grayscale value of the target pixel at the next position. The transformation function used for the grayscale transformation is still recorded as f(x), and all its undetermined coefficients are recorded as transformation features, which are randomly initialized. Obtain the absolute value of the difference between the grayscale value of each target pixel after grayscale transformation and the grayscale value of the reference pixel at the same position, which is simply recorded as the transformation difference of each target pixel. The sparsity of the transformation difference is obtained based on the transformation difference of all target pixels. The specific process is the same as step S003.

[0095] This step further takes into account that the target image at each position described in step S003 and the target image at the next position described in this step have the same or similar air environment or lighting environment. Therefore, in order to make the transformation curve more accurate and reliable, it is necessary not only to maximize the sparsity of the transformation difference, but also to incorporate some features of the historical curve. The specific method is to maximize the similarity between the transformation function f(x) in this step and the historical curve.

[0096] The specific method is:

[0097] The sparsity of the transformation difference in this step is denoted as M1, and the cosine similarity between the transformation feature of the history curve and the transformation feature of the transformation function f(x) used in this step is denoted as M2; let F=M1+M2.

[0098] In the above process, each given transformation feature corresponds to a value of F (that is, F is the objective function of the transformation feature of f(x)), and the simulated annealing algorithm is used to obtain the transformation feature when the sparse fusion degree M is the largest.

[0099] The transformation function f(x) corresponding to the transformation feature when the sparse fusion degree M is the maximum is the transformation curve used for grayscale transformation in this step. That is, after the grayscale value of the target pixel is transformed using the transformation curve obtained in this step, the sparseness of the difference between the grayscale value after grayscale transformation and the grayscale value of the reference pixel is the largest, and the similarity between the transformation curve and the historical curve is also the largest. This makes the transformation curve more accurate and reliable, which is conducive to increasing the subsequent compression transmission efficiency.

[0100] Furthermore, the grayscale transformation of the target image at the next position is performed using the transformation curve used in the grayscale transformation in this step, and the target image at the next position is compressed and transmitted using the difference between the target image after the grayscale transformation and the reference image. The specific process is the same as step S004.

[0101] Then, the segmentation threshold is corrected according to the error of the conversion curve mentioned above in this step (specifically, the segmentation threshold corrected in step S005) to obtain a segmentation threshold corrected again. The specific process is the same as step S005.

[0102] Step S007 , continue to obtain the target image and reference image at the next position, and compress and transmit the target image at the next position according to the target image and reference image at the next position and in combination with the corrected segmentation threshold.

[0103] The transformation curve used in the grayscale transformation in step S006 is recorded as the history curve, and the target image of the next position is continuously acquired and the reference image of the next position is read out. It can be understood that the next position in step S007 is described based on the next position in step S006, and for the convenience of description, this step is introduced based on the current next position.

[0104] The target image at the current next position is compressed and transmitted according to the target image and the reference image at the current next position and in combination with the segmentation threshold corrected in step S006. The specific process is the same as the method in steps S002 to S005. The general process is:

[0105] The difference image between the target image at the current next position and the reference image is obtained, and the difference image is segmented using the segmentation threshold corrected again in step S006 to obtain the background area again, and the target pixel point and the reference pixel point at the current next position are obtained on the target image at the current next position and the reference image. The specific process is the same as step S002.

[0106] Re-transform the grayscale value of the target pixel at the current next position. The transformation function used for the grayscale transformation is still recorded as f(x), and all its undetermined coefficients are recorded as transformation features, which are randomly initialized. Obtain the absolute value of the difference between the grayscale value of each target pixel after grayscale transformation and the grayscale value of the reference pixel at the same position, which is simply recorded as the transformation difference of each target pixel. The sparsity of the transformation difference is obtained based on the transformation differences of all target pixels, which is recorded as M3. The cosine similarity between the transformation features of the historical curve and the transformation features of the transformation function f(x) used in this step is recorded as M4.

[0107] The sparse fusion degree M is obtained again according to M3 and M4. The specific method is:

[0108] The similarity of the grayscale distribution of the target pixel in step S002 and the target pixel in step S006 is recorded as Q.

[0109] As an example, the method to obtain Q is:

[0110] All target pixels obtained in step S002 are recorded as Q1, and all target pixels obtained in step S006 are recorded as Q2. The one-dimensional vector formed by the mean and variance of the grayscale values ​​of all target pixels in Q1 is recorded as the first distribution feature, the one-dimensional vector formed by the mean and variance of the grayscale values ​​of all target pixels in Q2 is recorded as the second distribution feature, and the cosine similarity between the first distribution feature and the second distribution feature is recorded as Q.

[0111] Q describes the distribution difference of grayscale values. The smaller Q is, the greater the difference in grayscale distribution of the target pixel points obtained after segmentation using the segmentation threshold at each position and the next position of each position. Considering that the environmental interference of two adjacent positions is similar, it means that there is still an obvious error in the segmentation result of the segmentation threshold, and the historical curve has a large error at this time. At this time, this step pays more attention to the size of M3 when obtaining the sparse fusion degree M, avoiding introducing too much of the transformation curve used in the previous two compression transmissions into the current compression transmission process (that is, the next position at the current time), so as to ensure the compression transmission efficiency.

[0112] The larger Q is, the smaller the difference in grayscale distribution of the target pixel obtained by segmentation using the segmentation threshold at each position and the next position of each position is, indicating that the segmentation result error of the segmentation threshold is relatively small. At this time, the historical curve has a smaller error and is more reliable. At this time, this step pays more attention to the size of M4 when obtaining the sparse fusion degree M, and introduces more transformation curves used in the previous two compression transmissions into the current compression transmission process (that is, the next position at the current time) to ensure the compression transmission efficiency.

[0113] In this embodiment, the sparse fusion degree M=Q×M4+(1-Q)×M3.

[0114] In the above process of this step, each given transformation feature corresponds to a sparse fusion degree M (that is, the sparse fusion degree M is the target function of the transformation feature of f(x)). This embodiment uses a simulated annealing algorithm to obtain the transformation feature when the sparse fusion degree M is the maximum. The transformation function f(x) corresponding to the transformation feature when the sparse fusion degree M is the maximum is the transformation curve used in the grayscale transformation in this step.

[0115] Furthermore, the target image at the current next position is gray-scale transformed using the transformation curve mentioned above in this step, and the target image at the current next position is compressed and transmitted using the difference between the target image after gray-scale transformation and the reference image. The specific process is the same as step S004.

[0116] Then, the segmentation threshold is corrected according to the error of the above-mentioned transformation curve in this step (specifically, the segmentation threshold obtained in the correction step S006 after the second correction), and the specific process is the same as that of step S005.

[0117] In this step, the target image at the current position (i.e., the current lower position) is compressed and transmitted based on the similarity of the grayscale distribution of the target pixels under the segmentation thresholds of the first two positions and the historical curve (i.e., the transformation curve used for the grayscale transformation of the previous position).

[0118] For the target image at the subsequent position, continue to compress and transmit according to the method of this step, that is: based on the similarity of the grayscale distribution of the target pixel points under the segmentation threshold of the first two positions of each subsequent position, and the historical curve (that is, the transformation curve used when the grayscale of the previous position of the subsequent position is transformed), compress and transmit the target image at the subsequent position.

[0119] In other embodiments, at subsequent locations, steps S001 to S007 may be re-executed to compress and transmit the target image at each subsequent location.

[0120] It should be noted that if the drone does not store the reference image of each position, it directly compresses and transmits the target image of each position.

[0121] Embodiment 2:

[0122] This embodiment provides a highway construction project monitoring platform, which includes a server, a storage device and a display device, wherein the server is used to receive target images transmitted by drones, and the server also receives other data, such as measurement data of surveyors at highway construction sites, such as the weight of construction materials such as asphalt and gravel shipped out of the warehouse, road compaction, roadbed height, pipeline laying length and other data. The storage device is used to store target images transmitted by drones and measurement data of surveyors at highway construction sites. The display device is used to display target images transmitted by drones, so that supervisors can monitor construction progress.

[0123] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A construction progress monitoring method for a highway construction project monitoring platform, characterized in that: The method comprises the following steps: Obtaining a target image and a reference image at each location, wherein the reference image is acquired before the target image, including: during each inspection process, recording the image of each location acquired in real time as the target image, wherein the target image is the image transmitted to the highway construction project monitoring platform, and recording the image acquired at each location during the last inspection process as the reference image, wherein the reference image is stored on the drone; H1: compress and transmit the target image according to the target image and reference image at each position and in combination with the segmentation threshold, including: The difference image between the target image and the reference image is segmented by using the segmentation threshold to obtain the target pixel on the target image and the reference pixel on the reference image; the gray value of the target pixel is transformed into gray, and the transformation curve used in the gray transformation makes the sparsity of the difference between the gray value after the gray transformation and the gray value of the reference pixel maximized; the target image is transformed into gray by using the transformation curve, and the target image is compressed and transmitted by using the difference between the target image after the gray transformation and the reference image; the segmentation threshold is corrected according to the error of the transformation curve; H2: Use the transformation curve obtained in H1 as the history curve; obtain the target image and reference image at the next position; according to the method in H1, compress and transmit the target image at the next position based on the target image and reference image at the next position and the segmentation threshold corrected in H1; The transformation curve used in grayscale transformation in H2 has the greatest similarity to the historical curve; H3: Obtain the similarity of the grayscale distribution of the target pixel in H2 and the target pixel in H1; use the transformation curve obtained in H2 as the history curve; continue to obtain the target image and reference image at the next position; according to the method in H1, compress and transmit the target image at the next position in H3 based on the target image and reference image at the next position and combined with the corrected segmentation threshold in H2; The sparse fusion degree of the transformation curve used in the grayscale transformation in H3 and the historical curve is the largest, and the sparse fusion degree is obtained by the sparsity of the difference between the grayscale value after the grayscale transformation in H3 and the grayscale value of the reference pixel, the similarity between the transformation curve used in the grayscale transformation in H3 and the historical curve, and the grayscale distribution similarity; The transformation curve used in the grayscale transformation maximizes the sparsity of the difference between the grayscale value after the grayscale transformation and the grayscale value of the reference pixel, and includes the following specific steps: For the nonlinear function f(x), x represents the grayscale value of any target pixel, and the result output by f(x) represents the grayscale value after grayscale transformation; the one-dimensional vector composed of all the undetermined coefficients of the nonlinear function f(x) is recorded as the transformation feature, and the transformation feature is randomly initialized; the absolute value of the difference between the grayscale value of each target pixel after grayscale transformation and the grayscale value of the reference pixel is obtained, and recorded as the transformation difference of each target pixel; Obtaining the sparsity of the transformation difference according to the transformation difference of all target pixels; obtaining the transformation feature when the sparsity is maximum; and using the nonlinear function corresponding to the transformation feature when the sparsity is maximum as the transformation curve used in the grayscale transformation; The specific steps of obtaining the sparsity of the transformation difference according to the transformation difference of all target pixels are as follows: If the transformation difference of the target pixel point is less than or equal to the first preset threshold, then the transformation difference of the target pixel point is set to 0; the number of target pixels is recorded as N1, the number of target pixels with transformation difference equal to zero is obtained N2, and the ratio of N2 to N1 is recorded as the sparsity of the transformation difference.

2. The construction progress monitoring method for a highway construction project monitoring platform according to claim 1, characterized in that: The transformation curve used in the grayscale transformation in H2 has the greatest similarity to the historical curve, and the specific steps included are as follows: When H2 is executed according to the method in H1, the sparsity of the difference between the grayscale value after grayscale transformation and the grayscale value of the reference pixel is obtained in H2, which is recorded as M1, and the cosine similarity between the transformation feature of the history curve and the transformation feature of the nonlinear function in H2 is recorded as M2; let F=M1+M2; obtain the transformation feature when F is maximum; The nonlinear function corresponding to the transformation feature when F is maximum is used as the transformation curve used for the grayscale transformation obtained in H2.

3. The construction progress monitoring method for a highway construction project monitoring platform according to claim 1, characterized in that: The sparse fusion degree of the transformation curve used in the grayscale transformation in H3 and the historical curve is the largest, and the sparse fusion degree is obtained by the sparsity of the difference between the grayscale value after the grayscale transformation in H3 and the grayscale value of the reference pixel, the similarity between the transformation curve used in the grayscale transformation in H3 and the historical curve, and the similarity of the grayscale distribution, and the specific steps included are as follows: When H3 is executed according to the method in H1, the sparsity of the difference between the grayscale value after grayscale transformation and the grayscale value of the reference pixel point is obtained in H3, which is recorded as M3, and the cosine similarity between the transformation feature of the history curve and the transformation feature of the nonlinear function in H3 is recorded as M4; the grayscale distribution similarity is recorded as Q; Let the sparse fusion degree M = Q × M4 + (1-Q) × M3; The nonlinear function corresponding to the transformation feature when the sparse fusion degree M is maximum is used as the transformation curve used for the grayscale transformation obtained in H3.

4. The construction progress monitoring method for a highway construction project monitoring platform according to claim 1, characterized in that: The method of performing background segmentation on the difference image between the target image and the reference image by using the segmentation threshold to obtain target pixel points on the target image and reference pixel points on the reference image includes the following specific steps: The target image is subtracted from the reference image to obtain a differential image; the area in the differential image where the absolute value of the grayscale value is less than or equal to the segmentation threshold is recorded as the background area; the pixel points in the background area on the target image are recorded as target pixels, and the pixel points in the background area on the reference image are recorded as reference pixels.

5. The construction progress monitoring method for a highway construction project monitoring platform according to claim 1, characterized in that: The method of compressing and transmitting the target image by utilizing the difference between the target image after grayscale transformation and the reference image includes the following specific steps: The difference image between the transformed target image and the reference image is recorded as the image to be transmitted; the image to be transmitted is compressed using the LZW algorithm and then transmitted to the highway construction project monitoring platform.

6. The construction progress monitoring method for a highway construction project monitoring platform according to claim 1, characterized in that: The specific steps of correcting the segmentation threshold according to the error of the transformation curve are as follows: The grayscale value of each target pixel is transformed using the transformation curve. The transformed grayscale value is recorded as the sparse grayscale value of each target pixel. The standard deviation of the sparse grayscale values ​​of all target pixels and the grayscale value of the reference pixel is obtained and recorded as the error h of the transformation curve. Obtain several target pixel points with the largest sparse difference value, which are recorded as sample pixel points; record the area outside the target pixel points on the target image as the foreground area; obtain the connected domain formed by the sample pixel points and having a coincident boundary with the foreground area, which is recorded as the target connected domain; on the coincident boundary between the target connected domain and the foreground area, record the neighboring pixel points of all the pixel points on the boundary as local target pixel points; record the pixel points in the reference image at the same position as the local target pixel points as local reference pixel points; The grayscale value of each local target pixel is transformed by using the transformation curve, and the absolute value of the difference between the grayscale value of each local target pixel after the grayscale transformation and the grayscale value of the local reference pixel at the same position is obtained, which is recorded as the transformation difference of each local target pixel; the sparsity w of the transformation difference is obtained according to the transformation differences of all local target pixels; The segmentation threshold is modified according to the error h and sparsity w of the transformation curve.

7. The construction progress monitoring method for a highway construction project monitoring platform according to claim 6 is characterized in that: The specific steps of correcting the segmentation threshold according to the error h and sparsity w of the transformation curve are as follows: The segmentation threshold before correction is recorded as T, and (1+p×h)×T is taken as the segmentation threshold after correction; Among them, when w is greater than th21, p=w-th21, when w is less than th22, p=w-th22, when w is greater than or equal to th22 and less than or equal to th21, p=0; th21 and th22 are preset thresholds, and th22 is greater than th21.

8. The construction progress monitoring method for a highway construction project monitoring platform according to claim 7, characterized in that: The specific steps of obtaining the grayscale distribution similarity between the target pixel in H2 and the target pixel in H1 are as follows: All target pixel points obtained in H1 are recorded as Q1, and all target pixel points obtained in H2 are recorded as Q2; the one-dimensional vector composed of the mean and variance of the grayscale values ​​of all target pixel points in Q1 is recorded as the first distribution feature, and the one-dimensional vector composed of the mean and variance of the grayscale values ​​of all target pixel points in Q2 is recorded as the second distribution feature, and the cosine similarity between the first distribution feature and the second distribution feature is recorded as the grayscale distribution similarity.

Images