A method for detecting cracks in highways

Abstract

The present invention relates to the field of image analysis technology, and specifically to a method for detecting cracks on highways. The present invention utilizes the characteristics that road markings are periodic while crack defects are not periodic, analyzes the periodic characteristics of road markings, determines the historical periodic time point at the current moment, and analyzes the suspected defective area in the road image at the current moment in combination with the vehicle's driving direction and driving angle, and the distribution similarity characteristics of the suspected defective area in the road image at the historical periodic time point, determines the marking possibility of each suspected defective area in the road image at the current moment, and finally dynamically adjusts the acquisition frequency of the road image, reduces the waste of resources in highway crack detection, focuses on acquiring road images containing crack defects, and improves detection efficiency.

Description

Technical Field

[0001] The invention relates to the technical field of highway pavement image analysis, and in particular to a highway crack detection method. Background Art

[0002] A road inspection vehicle is a vehicle used to automatically inspect and assess road surface conditions. These vehicles are usually equipped with various sensors and equipment to collect various data on the road surface and use this data to assess road quality, detect road defects, and plan maintenance.

[0003] The camera above the road inspection vehicle can collect road images in real time. However, the probability of road defects is small. When a fixed acquisition frequency is used to collect normal road surfaces without crack defects, it will cause a waste of memory resources, increase the computing resources consumed by image analysis, and fail to collect defective road surfaces that need to be analyzed. Summary of the invention

[0004] In order to solve the technical problems of resource waste and low detection efficiency in highway crack detection, the purpose of the present invention is to provide a highway crack detection method, and the technical scheme adopted is as follows:

[0005] A method for detecting cracks in a highway, the method comprising:

[0006] Obtaining images of highway road surfaces through patrol vehicles;

[0007] According to the grayscale characteristics of the pixel points in each of the road surface images, the road surface images are threshold segmented to obtain suspected defect areas; according to the repetitive characteristics of the suspected defect areas in the continuous road surface images obtained by the patrol vehicle driving at a uniform speed in a straight line, combined with the uniform speed of the vehicle, the cycle length of the road surface marking is obtained; according to the cycle length and the vehicle speed at the current moment, the historical cycle time point at the current moment is obtained;

[0008] According to the suspected defective area in the road surface image at the current moment, the distribution similarity characteristics of the suspected defective area in the road surface image at the historical period time point, combined with the vehicle driving direction and driving angle, the possibility of the marking of each suspected defective area in the road surface image at the current moment is obtained;

[0009] The acquisition frequency of the road surface image is adjusted according to the possibility of the markings in all suspected defective areas in the road surface image at the current moment.

[0010] Furthermore, the method for obtaining the possibility of marking lines includes:

[0011] Mapping the road surface image at the current moment and the corresponding historical period time point to the same rectangular coordinate system; selecting any suspected defect area in the road surface image at the current moment as the area to be analyzed; calling the suspected defect area in the road surface image at the historical period time point corresponding to the current moment as the comparison area;

[0012] The center point of the area to be analyzed is taken as the center of the neighborhood window, and the neighborhood window of the area to be analyzed is constructed with a preset window size; according to the relative distribution characteristics of the area to be analyzed and the comparison area in the neighborhood window, combined with the driving speed and driving angle of the vehicle, the possibility of the marking line of the area to be analyzed is obtained.

[0013] Furthermore, the method for obtaining the possibility of the marking line of the area to be analyzed includes:

[0014] Select any of the comparison areas within the neighborhood window of the area to be analyzed as a target comparison area;

[0015] According to the horizontal offset characteristics of the center point of the area to be analyzed and the center point of the target comparison area, combined with the vehicle driving angle at the current moment, the horizontal possibility parameters corresponding to the area to be analyzed and the target comparison area are obtained;

[0016] According to the offset characteristics of the center point of the area to be analyzed and the center point of the target comparison area in the vertical direction, obtaining the vertical possibility parameters corresponding to the area to be analyzed and the target comparison area;

[0017] The weighted weights of the horizontal possibility parameter and the vertical possibility parameter are obtained according to the vehicle driving angle at the current moment; the vehicle driving angle at the current moment is positively correlated with the weighted weight of the horizontal possibility parameter; the vehicle driving angle at the current moment is negatively correlated with the weighted weight of the vertical possibility parameter; the sum of the weighted weight of the horizontal possibility parameter and the weighted weight of the vertical possibility parameter is 1; the weighted sum of the horizontal possibility parameter and the vertical possibility parameter corresponding to the area to be analyzed and the target comparison area is taken as the period rationality corresponding to the area to be analyzed and the target comparison area; the maximum period rationality corresponding to the area to be analyzed is taken as the line marking possibility of the area to be analyzed.

[0018] Furthermore, the method for obtaining the horizontal possibility parameter includes:

[0019] Obtaining an offset angle according to a horizontal offset distance between the center point of the area to be analyzed and the center point of the target comparison area in the horizontal direction and a distance between the center point of the area to be analyzed and the center point of the target comparison area;

[0020] According to the similarity characteristics between the vehicle driving angle at the current moment and the offset angle, the horizontal possibility parameter of the area to be analyzed is obtained; the similarity characteristics between the vehicle driving angle at the current moment and the offset angle are positively correlated with the horizontal possibility parameter.

[0021] Furthermore, the method for obtaining the vertical possibility parameter includes:

[0022] According to the offset angle and the vehicle speed of the vehicle at the current moment, the horizontal theoretical offset distance at the current moment is obtained by using the pre-trained neural network model; according to the distance between the center point of the area to be analyzed and the center point of the target comparison area and the horizontal theoretical offset distance, the vertical theoretical distance is obtained;

[0023] According to the vertical offset distance between the center point of the area to be analyzed and the center point of the target comparison area in the vertical direction, and the similarity characteristics with the vertical theoretical distance, the vertical possibility parameter of the area to be analyzed is obtained; the similarity characteristics between the vertical offset distance and the vertical theoretical distance are positively correlated with the vertical possibility parameter.

[0024] Furthermore, the method for adjusting the acquisition frequency of the road surface image includes:

[0025] Obtaining an adjustment coefficient according to the concentrated features of the marking possibilities of all suspected defective areas in the road surface image at the current moment; the concentrated features of the marking possibilities are negatively correlated with the adjustment coefficient;

[0026] The sampling frequency at the current moment is obtained according to the preset basic sampling frequency and the adjustment coefficient; the preset basic sampling frequency and the adjustment coefficient are both positively correlated with the sampling frequency at the current moment.

[0027] Furthermore, the method for obtaining the suspected defect area includes:

[0028] Pixels of the road surface image that do not belong to a preset grayscale threshold range are marked as suspected pixels, and a connected domain formed by the suspected pixels in the road surface image is used as a suspected defect area.

[0029] Furthermore, the method for obtaining the cycle length includes:

[0030] In the continuous road surface images obtained when the patrol vehicle travels in a straight line at a uniform speed, the minimum time interval between the appearance of the same suspected defect area in different road surface images is obtained; and the product of the minimum time interval and the uniform speed of the vehicle is taken as the cycle length.

[0031] Furthermore, the method for obtaining the cycle length includes:

[0032] The ratio of the cycle length to the vehicle speed at the current moment is taken as the historical cycle interval at the current moment; taking the current moment as the starting point and the historical cycle interval as the time domain length, the historical cycle time points are obtained in the negative direction of the time domain.

[0033] Furthermore, the preset grayscale threshold interval is [100, 200].

[0034] The present invention has the following beneficial effects:

[0035] The present invention first obtains the road surface image of the expressway through the patrol vehicle to provide an analysis basis; further threshold segmentation is performed on the road surface image to obtain the suspected defect area, narrow the analysis scope, facilitate the regular characteristics of the subsequent suspected defect area, and adjust the acquisition frequency of the road surface image; further, based on the repetitive characteristics of the suspected defect area in the continuous road surface image obtained by the patrol vehicle driving at a uniform speed and in a straight line, combined with the uniform speed of the vehicle, the period length of the road surface marking is obtained, and the periodic characteristic parameters of the road surface marking are determined, so as to facilitate the subsequent comparison of the different distribution patterns of the road surface marking and the cracks, and utilize the characteristics that the road surface marking has periodicity while the crack defect does not have periodicity; further, based on the period length and the vehicle speed at the current moment, the historical period time point at the current moment is obtained, so as to facilitate the subsequent comparison of the road surface image at the current moment with the historical period time point. The road surface images at periodic time points are used to analyze the periodic characteristics of the suspected defective areas. Further, considering that the vehicle's driving direction and driving angle will affect the distribution of vehicle markings, the interference of vehicle driving on the distribution of suspected defective areas is corrected based on the suspected defective areas in the road surface image at the current moment and the suspected defective areas in the road surface images at historical periodic time points. The possibility of markings in each suspected defective area in the road surface image at the current moment is obtained, which provides a basis for the final dynamic adjustment of the acquisition frequency of the road surface image. Finally, the acquisition frequency of the road surface image is adjusted according to the possibility of markings in all suspected defective areas in the road surface image at the current moment, so as to reduce the waste of resources in highway crack detection, focus on collecting road surface images containing crack defects, and improve detection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are 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.

[0037] Figure 1 A flow chart of a highway crack detection method provided by one embodiment of the present invention;

[0038] Figure 2 A schematic diagram of the operation of a road surface inspection vehicle provided by one embodiment of the present invention;

[0039] Figure 3 A schematic diagram of the distribution of marking deviations of a vehicle on a curve is provided in accordance with an embodiment of the present invention. DETAILED DESCRIPTION

[0040] 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 a highway crack detection method proposed by the present invention, its specific implementation, 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.

[0041] 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.

[0042] The specific scheme of a highway crack detection method provided by the present invention is described in detail below with reference to the accompanying drawings.

[0043] See also Figure 1 , which shows a flow chart of a highway crack detection method provided by an embodiment of the present invention, specifically comprising:

[0044] Step S1: Acquire a road surface image of a highway by using a patrol vehicle.

[0045] The embodiments of the present invention are mainly based on the fact that road markings on highways are periodic and appear regularly in continuous road surface images, while road gap defects are random. When road markings appear regularly, road marking areas in different road surface images have a certain consistency, while road gap defect areas have strong isolation and irregularity. By using this regularity, the possibility of containing defective areas in the image is analyzed, and then the acquisition frequency of the road surface image is dynamically adjusted, the memory resource and computing resource consumption of non-defective road surface images are reduced, and the acquisition is focused on defective roads, optimizing resource allocation, and improving the accuracy and efficiency of highway crack detection.

[0046] In order to dynamically adjust the acquisition frequency of road surface images, the road surface images of the highway are first obtained through patrol vehicles to provide an analysis basis.

[0047] See also Figure 2 , which shows a working schematic diagram of a road surface inspection vehicle provided by an embodiment of the present invention; Figure 2In the process, the road surface inspection vehicle is equipped with a vehicle-mounted camera to capture images of the road surface of the highway.

[0048] It should be noted that, in one embodiment of the present invention, the acquired road surface image is preprocessed, including but not limited to image denoising and removal of non-road surface areas, so as to reduce noise and interference of non-road surface areas.

[0049] Step S2: Perform threshold segmentation on the road surface image based on the grayscale features of the pixels in each road surface image to obtain suspected defect areas; obtain the cycle length of the road marking based on the repetitive features of the suspected defect areas in the continuous road surface images obtained by the patrol vehicle driving in a straight line at a uniform speed, combined with the uniform speed of the vehicle; obtain the historical cycle time point at the current moment based on the cycle length and the vehicle speed at the current moment.

[0050] Considering that cracks and markings have large grayscale differences from the road surface, the road surface image is first segmented by threshold according to the grayscale characteristics of the pixels in each road surface image to obtain the suspected defect area, narrow the analysis scope, facilitate the regular characteristics of the subsequent suspected defect area, and adjust the acquisition frequency of the road surface image.

[0051] Preferably, in one embodiment of the present invention, considering that the grayscale values ​​of road surface pixels are relatively concentrated, a dual threshold method is used to determine the pixel points of the road surface image that belong to a preset grayscale threshold interval as road surface pixels, that is, background pixels, and the pixel points of the road surface image that do not belong to the preset grayscale threshold interval are marked as suspected pixels, and the connected domain formed by the suspected pixels in the road surface image is used as a suspected defect area.

[0052] It should be noted that, in one embodiment of the present invention, by obtaining the average gray value of the road surface , set the preset gray threshold range to , For example, the preset threshold control parameters are: , , that is, the preset grayscale threshold interval is [100,200].

[0053] In other embodiments of the present invention, the implementer may also set other preset threshold control parameters, or divide the road surface image into blocks, obtain the grayscale threshold range for each block, mark the suspected pixels, and obtain the suspected defect area based on the marking results of all block images.

[0054] In the road surface detection task, since both crack defects and markings have a large grayscale difference from the road surface, it is impossible to distinguish the two. However, on highways, road surface markings and cracks have different distribution patterns. Road surface markings have regular periodicity, while gaps appear randomly. In order to utilize the periodic characteristics of road surface markings, the period length of road surface markings is first determined; considering that when the inspection vehicle is traveling in a straight line at a uniform speed, the regular characteristics of road surface markings in the collected road surface images are most obvious, and the suspected defective areas belonging to the markings are repeated in continuous road surface images. At the same time, the repetition interval is related to the speed of the vehicle's average speed. Therefore, according to the repetitive characteristics of the suspected defective areas in the continuous road surface images obtained by the inspection vehicle traveling in a straight line at a uniform speed, combined with the vehicle's uniform speed, the period length of the road surface markings is obtained, in preparation for the subsequent determination of the historical period time point at the current moment.

[0055] Preferably, in one embodiment of the present invention, the minimum time interval for the markings to recur in continuous road surface images is taken into consideration, that is, the recurrence frequency of the markings at a uniform driving speed, and the period length of the road surface markings can be obtained by combining the recurrence frequency and the average speed;

[0056] Based on this, in the continuous road surface images obtained by the patrol vehicle driving in a straight line at a uniform speed, the minimum time interval between the same suspected defect areas appearing in different road surface images is obtained; the product of the minimum time interval and the uniform speed of the vehicle is taken as the cycle length.

[0057] It should be noted that, in one embodiment of the present invention, the period length of the road marking is used as a basic parameter, and the road image is acquired by having a patrol vehicle drive in a straight line at a uniform speed on a straight road section and acquiring the road image at a fixed acquisition frequency, and the acquisition is performed before the acquisition frequency is dynamically adjusted; the minimum time interval is not greater than a preset empirical time interval, and the preset empirical time interval is set to an empirical value of 5 seconds. When the minimum time interval obtained exceeds 5 seconds, it is determined that the continuous road image used to obtain the period length of the road marking is not ideal, and the acquisition is retested.

[0058] It should be noted that the method of acquiring the speed of the vehicle is already an existing technology, and can be specifically acquired by reading the speed sensor parameters when collecting road surface images, which will not be described in detail here.

[0059] After determining the cycle length of the road markings, the historical cycle time point at the current moment can be obtained based on the cycle length and the vehicle speed at the current moment, so as to compare the road surface image at the current moment with the road surface image at the historical cycle time point, and use the periodic characteristics of the road markings to dynamically adjust the acquisition frequency of the road surface image.

[0060] Preferably, in one embodiment of the present invention, the ratio of the cycle length to the vehicle speed at the current moment is used as the historical cycle interval at the current moment; taking the current moment as the starting point and the historical cycle interval as the time domain length, the historical cycle time points are obtained in the negative direction of the time domain.

[0061] In other embodiments of the present invention, considering the rate fluctuations during vehicle driving, the current moment is used as the upper limit of the integral, and the lower limit of the integral is selected in the negative direction of the time domain, and the corresponding rate data is integrated over time until the integral value is equal to the cycle length. The lower limit of the integral at this time is used as the historical cycle time point. Compared with the method of calculating the historical cycle interval, the method of obtaining the historical cycle time point through integration consumes more computing resources, but improves the accuracy of the historical cycle time point, and the implementer can select it according to the needs of the implementation scenario.

[0062] It should be noted that the negative direction in the time domain is the direction in which the moment decreases, that is, the historical direction.

[0063] Step S3: Based on the suspected defective area in the road surface image at the current moment, the distribution characteristics of the suspected defective area in the road surface image at the historical period time point are similar, combined with the vehicle's driving direction and driving angle, the possibility of marking each suspected defective area in the road surface image at the current moment is obtained.

[0064] The historical cycle time point at the current moment is a cycle-adjacent time point, which represents the time point at which the road surface image at the current moment periodically appears. There is at least one cycle of repeated road markings between the current moment and the historical cycle time point. Therefore, based on the suspected defect area in the road surface image at the current moment and the similar distribution characteristics of the suspected defect area in the road surface image at the historical cycle time point, the periodicity of the suspected defect area is analyzed to determine the possibility of markings in the suspected defect area. Considering that the vehicle's driving direction will affect the distribution of the suspected defect area, it is also necessary to analyze in combination with the vehicle's driving direction to obtain the possibility of markings for each suspected defect area in the road surface image at the current moment.

[0065] Preferably, in one embodiment of the present invention, in order to facilitate the analysis of the distribution similarity characteristics of the suspected defect areas of different images, the road surface images at the current moment and the corresponding historical period time point are mapped to the same rectangular coordinate system; in order to facilitate the analysis one by one, any suspected defect area in the road surface image at the current moment is selected as the area to be analyzed; in order to facilitate the comparison of the suspected defect areas in the road surface images at different moments, the periodic characteristics exhibited by the suspected defect areas are analyzed, and the suspected defect areas in the road surface images at the historical period time point corresponding to the current moment are called comparison areas for easy distinction;

[0066] Taking into account that the position of the markings will shift when the vehicle is traveling on a curve, and the vehicle's driving speed and driving angle will affect the specific offset characteristics, the center point of the area to be analyzed is taken as the center of the neighborhood window, and the neighborhood window of the area to be analyzed is constructed with a preset window size. The distribution similarity characteristics of the area to be analyzed and the comparison area are analyzed within the neighborhood window. According to the relative distribution characteristics of the area to be analyzed and the comparison area in the neighborhood window, combined with the vehicle's driving speed and driving angle, the possibility of the markings in the area to be analyzed is obtained.

[0067] It should be noted that, in one embodiment of the present invention, the lower left corner pixel of each road image is taken as the origin, and the road images at the current time and the corresponding historical period time point are mapped to the same rectangular coordinate system; the preset window size is , taking the centroid of the suspected defect area as the center point, when the center point of the comparison area is within the neighborhood window of the area to be analyzed, it is determined that the comparison area is within the neighborhood window of the area to be analyzed.

[0068] Preferably, in one embodiment of the present invention, firstly, any comparison area in the neighborhood window of the area to be analyzed is selected as the target comparison area, so as to compare the area to be analyzed and the comparison area one by one;

[0069] See also Figure 3 , which shows a schematic diagram of a marking deviation distribution of a vehicle at a curve provided by an embodiment of the present invention; Figure 3 In the example, the predicted markings are the suspected defect areas in the road image at the current moment, and the actual markings are the suspected defect areas in the road image at the historical cycle time point. The vehicle driving direction at the historical cycle time point is taken as the vertical direction. Figure 3 It can be seen that due to the turning of the vehicle, the suspected defect area that overlaps due to periodicity in the ideal is offset. Therefore, it is necessary to analyze the rationality of the offset of the suspected defect area according to the vehicle driving state, so as to measure the possibility of the line marking in the area to be analyzed;

[0070] When the marking is offset, it will be offset in both the horizontal and vertical directions, so the possibility of the marking is measured from two angles: horizontal and vertical. Considering that the turning angle of the vehicle will affect the horizontal offset of the marking, and the driving speed will affect the vertical offset of the marking, it is necessary to combine the vehicle driving angle and the vehicle speed when analyzing the offset characteristics between the center point of the area to be analyzed and the center point of the target comparison area. Considering that the larger the vehicle driving angle, the larger the horizontal offset, the greater the influence weight of the feature parameters analyzed in the horizontal direction on the possibility of the marking, so when fusing the horizontal possibility parameter and the vertical possibility parameter, the weighted weight is obtained according to the vehicle driving angle.

[0071] Based on this, according to the horizontal offset characteristics of the center point of the area to be analyzed and the center point of the target comparison area, combined with the vehicle driving angle at the current moment, the horizontal possibility parameters corresponding to the area to be analyzed and the target comparison area are obtained;

[0072] According to the vertical offset characteristics of the center point of the area to be analyzed and the center point of the target comparison area, combined with the vehicle speed at the current moment, the vertical possibility parameters corresponding to the area to be analyzed and the target comparison area are obtained;

[0073] The weighted weights of the horizontal possibility parameter and the vertical possibility parameter are obtained according to the vehicle driving angle at the current moment; the vehicle driving angle at the current moment is positively correlated with the weighted weight of the horizontal possibility parameter; the vehicle driving angle at the current moment is negatively correlated with the weighted weight of the vertical possibility parameter; the sum of the weighted weight of the horizontal possibility parameter and the weighted weight of the vertical possibility parameter is 1; the weighted sum of the horizontal possibility parameter and the vertical possibility parameter corresponding to the area to be analyzed and the target comparison area is taken as the period rationality corresponding to the area to be analyzed and the target comparison area; the maximum period rationality corresponding to the area to be analyzed is taken as the line marking possibility of the area to be analyzed.

[0074] As an example: The formula for calculating the probability of marking a line includes:

[0075] ;

[0076] in, Indicates the serial number of the area to be analyzed; Indicates the serial number of the target comparison area; Indicates The area to be analyzed and The rationality of the period corresponding to each target comparison area; express Type function, used for normalization; Indicates the vehicle's driving angle; Indicates The area to be analyzed and The horizontal possibility parameter corresponding to the target contrast area; Indicates The area to be analyzed and The vertical possibility parameter corresponding to the target contrast area.

[0077] In the calculation formula for the possibility of line markings, the periodic rationality of the offset of the suspected defective area is measured from both the horizontal and vertical directions. The greater the periodic rationality, the more the offset characteristics of the suspected defective area are consistent with the driving state of the vehicle, and the more likely it is a line marking. Therefore, the maximum periodic rationality corresponding to the area to be analyzed is used as the possibility of line markings in the area to be analyzed, providing a basis for the subsequent dynamic adjustment of the acquisition frequency.

[0078] It should be noted that straight-line driving is a special case of curved driving, so the same method is adopted to obtain the possibility of line markings; when there is no comparison area in the neighborhood window of the area to be analyzed, it means that the area to be analyzed exhibits strong isolation and non-periodicity, and the possibility of line markings in the area to be analyzed is directly set to 0.

[0079] It should be noted that the vehicle's driving direction at the historical cycle time point is taken as the vertical direction, and then the horizontal direction is obtained; through devices such as a gyroscope, the angle of rotation of the vehicle from the historical cycle time point to the current moment is obtained as the vehicle's driving angle.

[0080] Preferably, in one embodiment of the present invention, considering that the more periodic the area to be analyzed and the target comparison area are, the closer the offset angle is to the vehicle driving angle, so the offset angle is obtained according to the horizontal offset distance between the center point of the area to be analyzed and the center point of the target comparison area in the horizontal direction, combined with the distance between the center point of the area to be analyzed and the center point of the target comparison area;

[0081] See also Figure 3 The marked offset angle is the angle between the line connecting the center point of the target comparison area and the center point of the area to be analyzed and the vertical direction, which reflects the angle of rotation from the historical cycle time point to the current moment in vehicle theory. Therefore, the more similar the offset angle is to the actual vehicle driving angle, the stronger the periodicity and regularity shown by the center point of the target comparison area and the center point of the area to be analyzed, and the larger the horizontal possibility parameter. Therefore, the horizontal possibility parameter of the area to be analyzed is obtained based on the similarity characteristics of the vehicle driving angle and the offset angle at the current moment; the similarity characteristics of the vehicle driving angle and the offset angle at the current moment are positively correlated with the horizontal possibility parameter.

[0082] As an example, under the premise of knowing the opposite side and the hypotenuse, it is easy to obtain the offset angle through the inverse trigonometric function, that is, the horizontal offset distance between the center point of the area to be analyzed and the center point of the target comparison area in the horizontal direction is taken as the opposite side, and the distance between the center point of the area to be analyzed and the center point of the target comparison area is taken as the hypotenuse, and the offset angle is obtained through the inverse sine function; the calculation formula of the horizontal possibility parameter includes:

[0083] ;

[0084] In the formula, Indicates the serial number of the area to be analyzed; Indicates the serial number of the target comparison area; Indicates the vehicle's driving angle; Indicates The area to be analyzed and The horizontal possibility parameter corresponding to the target contrast area; No. The area to be analyzed and The offset angle corresponding to the target contrast area; Indicates the preset divide-by-zero positive parameter. In this example, .

[0085] In the calculation formula of the horizontal possibility parameter, the absolute value of the difference between the vehicle driving angle and the offset angle at the current moment is obtained to express the difference characteristics between the vehicle driving angle and the offset angle, and then the reciprocal method is used for negative correlation mapping to express the similar characteristics of the vehicle driving angle and the offset angle, and the denominator is prevented from being zero by using the preset positive parameter of zero division. The offset angle is obtained by analyzing the offset distance in the horizontal direction, and the distribution similarity characteristics of the area to be analyzed and the target comparison area are analyzed through the similar characteristics of the offset angle and the vehicle driving angle, providing a basis for obtaining the possibility of the marking line.

[0086] In other embodiments of the present invention, the implementer may also use a negative correlation mapping function such as Perform negative correlation mapping, Expressed as a natural constant is the exponential function of the base, Represents the independent variable.

[0087] Preferably, in one embodiment of the present invention, considering that the driving angle and vehicle speed of the vehicle will affect the horizontal offset distance of the vehicle, the horizontal theoretical offset distance at the current moment can be obtained by using a pre-trained neural network model according to the offset angle and vehicle speed of the vehicle at the current moment; and then the vertical theoretical distance can be obtained by using the Pythagorean theorem according to the distance between the center point of the area to be analyzed and the center point of the target comparison area and the horizontal theoretical offset distance;

[0088] Considering that the more similar the actual vertical offset distance is to the vertical theoretical distance, the more likely it is that the offset characteristics of the area to be analyzed and the target comparison area in the vertical direction are normal offsets caused by vehicle driving. The more reasonable the offset is, the more obvious the similarity characteristics of the distribution of the area to be analyzed and the target comparison area in the vertical direction are, the stronger the periodicity of the area to be analyzed is, the more likely it is a marking line, and the larger the vertical possibility parameter is. Therefore, the vertical possibility parameter of the area to be analyzed is obtained according to the vertical offset distance between the center point of the area to be analyzed and the center point of the target comparison area in the vertical direction and the similarity characteristics with the vertical theoretical distance; the similarity characteristics of the vertical offset distance and the vertical theoretical distance are positively correlated with the vertical possibility parameter.

[0089] As an example, the calculation formula of the vertical possibility parameter includes:

[0090] ;

[0091] In the formula, Indicates the serial number of the area to be analyzed; Indicates the serial number of the target comparison area; Indicates The area to be analyzed and The vertical possibility parameter corresponding to the target contrast area; Indicates The area to be analyzed and The vertical offset distance corresponding to the target comparison area; Indicates The area to be analyzed and The vertical theoretical distance corresponding to the target comparison area; Indicates the preset divide-by-zero positive parameter. In this example, .

[0092] In the calculation formula of the vertical possibility parameter, the absolute value of the difference between the vertical offset distance and the vertical theoretical distance is obtained to express the difference characteristics between the vertical offset distance and the vertical theoretical distance. Then, the inverse method is used for negative correlation mapping to express the similar characteristics between the vertical offset distance and the vertical theoretical distance. The preset positive division by zero parameter is used to prevent the denominator from being zero. Through the similar characteristics between the vertical offset distance and the vertical theoretical distance, the distribution similarity characteristics of the area to be analyzed and the target comparison area are analyzed to provide a basis for obtaining the possibility of the marking line.

[0093] In other embodiments of the present invention, the implementer may also use a negative correlation mapping function such as Perform negative correlation mapping, Expressed as a natural constant is the exponential function of the base, Represents the independent variable.

[0094] It should be noted that the pre-trained neural network model inputs the driving angle and vehicle speed of the vehicle, and outputs the horizontal theoretical offset distance. The pre-trained neural network model is obtained by obtaining a training set and a verification set in a test section, which is convenient for use in highway crack detection. In one embodiment of the present invention, a convolutional neural network model is trained. In other embodiments of the present invention, the implementer may also train other neural network models. The training process of the neural network model is a technical means well known to those skilled in the art and will not be repeated here.

[0095] It should be noted that, in one embodiment of the present invention, when calculating the distance, it is unified as pixel distance; in other embodiments of the present invention, it can also be unified as meters, centimeters, etc. in actual scenes.

[0096] Step S4: adjusting the acquisition frequency of the road surface image according to the possibility of markings in all suspected defective areas in the road surface image at the current moment.

[0097] After obtaining the marking possibilities of all suspected defective areas in the road surface image at the current moment, the crack defect situation in the road surface image can be evaluated based on the marking possibilities of all suspected defective areas in the road surface image at the current moment, and the acquisition frequency of the road surface image can be dynamically adjusted.

[0098] Preferably, in one embodiment of the present invention, considering that the greater the possibility of line marking in the suspected defect area, the more likely the suspected defect area is to be a normal line marking area, the higher the line marking possibilities of all suspected defect areas in the road surface image at the current moment are concentrated in the area with a higher value, indicating that the road surface image contains less crack defects and the lower the acquisition frequency requirement;

[0099] Based on this, the adjustment coefficient is obtained according to the concentrated features of the marking possibilities of all suspected defective areas in the road surface image at the current moment; the concentrated features of the marking possibilities are negatively correlated with the adjustment coefficient;

[0100] The sampling frequency at the current moment is obtained according to the preset basic sampling frequency and the adjustment coefficient; the preset basic sampling frequency and the adjustment coefficient are both positively correlated with the sampling frequency at the current moment.

[0101] As an example, the calculation formula for the sampling frequency at the current moment includes:

[0102] ;

[0103] in, Indicates the sampling frequency at the current moment; represents the linear normalization function; Indicates the adjustment coefficient corresponding to the current moment; Indicates the average value of the marking probability of all suspected defective areas in the road surface image at the current moment; Indicates the preset divide-by-zero positive parameter. In this example, ; Indicates the preset basic sampling frequency. In this example, ; Indicates rounding down to ensure is an integer.

[0104] The calculation formula of the sampling frequency at the current moment represents the concentrated characteristics of the possibility of markings through the average value. The larger the average value of the possibility of markings of all suspected defective areas in the road surface image, the more likely it is that the suspected defective areas in the road surface image are normal marking areas, the smaller the possibility of containing crack defects, and the smaller the acquisition frequency, which reduces the consumption of memory and computing resources.

[0105] In other embodiments of the present invention, the average value, mode, and median may be used to jointly represent the concentrated features of the marking possibility, and the average value, mode, and median of the marking possibility may be weighted and summed to replace , the weighted weights can be 0.4, 0.3, 0.3; or a negative correlation mapping function such as Perform negative correlation mapping, ; Other preset basic sampling frequencies can also be set, which will not be described here.

[0106] In another embodiment of the present invention, considering that the minimum value of the marking probability of all suspected defective areas in the road surface image at the current moment represents the maximum possibility that the road surface image at the current moment contains crack defects, the sampling frequency at the current moment is adjusted by using the minimum value of the marking probability, such as , Indicates the minimum value of the marking probability of all suspected defect areas in the road surface image at the current moment.

[0107] It should be noted that, compared with the method of adjusting the frequency by the minimum value of the marking possibility and the method of adjusting the frequency by the concentrated feature of the marking possibility, the method of adjusting the frequency by the minimum value of the marking possibility is more sensitive to crack defects, but the shooting frequency is higher and consumes more resources. Implementers can choose it at their own discretion.

[0108] In summary, in order to solve the technical problems of resource waste and low detection efficiency in highway crack detection, the present invention proposes a highway crack detection method. The present invention utilizes the characteristics that road markings are periodic while crack defects are not periodic, analyzes the periodic characteristics of road markings, determines the historical periodic time point at the current moment, and analyzes the suspected defective area in the road image at the current moment in combination with the vehicle's driving direction and driving angle, and the distribution similarity characteristics of the suspected defective area in the road image at the historical periodic time point, determines the marking possibility of each suspected defective area in the road image at the current moment, and finally dynamically adjusts the acquisition frequency of the road image, reduces the waste of resources in highway crack detection, focuses on acquiring road images containing crack defects, and improves detection efficiency.

[0109] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0110] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A method for detecting cracks in a highway, characterized in that: The method comprises: Obtaining images of highway road surfaces through patrol vehicles; According to the grayscale characteristics of the pixel points in each of the road surface images, the road surface images are threshold segmented to obtain suspected defect areas; according to the repetitive characteristics of the suspected defect areas in the continuous road surface images obtained by the patrol vehicle driving at a uniform speed in a straight line, combined with the uniform speed of the vehicle, the cycle length of the road surface marking is obtained; according to the cycle length and the vehicle speed at the current moment, the historical cycle time point at the current moment is obtained; According to the suspected defective area in the road surface image at the current moment, the distribution similarity characteristics of the suspected defective area in the road surface image at the historical period time point, combined with the vehicle driving direction and driving angle, the possibility of the marking of each suspected defective area in the road surface image at the current moment is obtained; adjusting the acquisition frequency of the road surface image according to the possibility of the markings in all suspected defective areas in the road surface image at the current moment; The method for obtaining the possibility of marking lines includes: Mapping the road surface image at the current moment and the corresponding historical period time point to the same rectangular coordinate system; selecting any suspected defect area in the road surface image at the current moment as the area to be analyzed; calling the suspected defect area in the road surface image at the historical period time point corresponding to the current moment as the comparison area; Taking the center point of the area to be analyzed as the center of the neighborhood window, constructing the neighborhood window of the area to be analyzed with a preset window size; obtaining the possibility of the marking line of the area to be analyzed according to the relative distribution characteristics of the area to be analyzed and the comparison area in the neighborhood window, combined with the driving speed and driving angle of the vehicle; The method for adjusting the acquisition frequency of the road surface image includes: Obtaining an adjustment coefficient according to the concentrated features of the marking possibilities of all suspected defective areas in the road surface image at the current moment; the concentrated features of the marking possibilities are negatively correlated with the adjustment coefficient; According to the preset basic sampling frequency and the adjustment coefficient, the sampling frequency at the current moment is obtained; the preset basic sampling frequency and the adjustment coefficient are both positively correlated with the sampling frequency at the current moment; The method for obtaining the suspected defect area includes: Pixels of the road surface image that do not belong to a preset grayscale threshold range are marked as suspected pixels, and a connected domain formed by the suspected pixels in the road surface image is used as a suspected defect area.

2. A highway crack detection method according to claim 1, characterized in that: The method for obtaining the possibility of the marking line of the area to be analyzed includes: Select any of the comparison areas within the neighborhood window of the area to be analyzed as a target comparison area; According to the horizontal offset characteristics of the center point of the area to be analyzed and the center point of the target comparison area, combined with the vehicle driving angle at the current moment, the horizontal possibility parameters corresponding to the area to be analyzed and the target comparison area are obtained; According to the offset characteristics of the center point of the area to be analyzed and the center point of the target comparison area in the vertical direction, obtaining the vertical possibility parameters corresponding to the area to be analyzed and the target comparison area; The weighted weights of the horizontal possibility parameter and the vertical possibility parameter are obtained according to the vehicle driving angle at the current moment; the vehicle driving angle at the current moment is positively correlated with the weighted weight of the horizontal possibility parameter; the vehicle driving angle at the current moment is negatively correlated with the weighted weight of the vertical possibility parameter; the sum of the weighted weight of the horizontal possibility parameter and the weighted weight of the vertical possibility parameter is 1; the weighted sum of the horizontal possibility parameter and the vertical possibility parameter corresponding to the area to be analyzed and the target comparison area is taken as the period rationality corresponding to the area to be analyzed and the target comparison area; the maximum period rationality corresponding to the area to be analyzed is taken as the line marking possibility of the area to be analyzed.

3. A highway crack detection method according to claim 2, characterized in that: The method for obtaining the horizontal possibility parameter includes: Obtaining an offset angle according to a horizontal offset distance between the center point of the area to be analyzed and the center point of the target comparison area in the horizontal direction and a distance between the center point of the area to be analyzed and the center point of the target comparison area; According to the similarity characteristics between the vehicle driving angle at the current moment and the offset angle, the horizontal possibility parameter of the area to be analyzed is obtained; the similarity characteristics between the vehicle driving angle at the current moment and the offset angle are positively correlated with the horizontal possibility parameter.

4. A highway crack detection method according to claim 3, characterized in that: The method for obtaining the vertical possibility parameter includes: According to the offset angle and the vehicle speed of the vehicle at the current moment, the horizontal theoretical offset distance at the current moment is obtained by using the pre-trained neural network model; according to the distance between the center point of the area to be analyzed and the center point of the target comparison area and the horizontal theoretical offset distance, the vertical theoretical distance is obtained; According to the vertical offset distance between the center point of the area to be analyzed and the center point of the target comparison area in the vertical direction, and the similarity characteristics with the vertical theoretical distance, the vertical possibility parameter of the area to be analyzed is obtained; the similarity characteristics between the vertical offset distance and the vertical theoretical distance are positively correlated with the vertical possibility parameter.

5. A highway crack detection method according to claim 1, characterized in that: The method for obtaining the cycle length includes: In the continuous road surface images obtained when the patrol vehicle travels in a straight line at a uniform speed, the minimum time interval between the appearance of the same suspected defect area in different road surface images is obtained; and the product of the minimum time interval and the uniform speed of the vehicle is taken as the cycle length.

6. A highway crack detection method according to claim 5, characterized in that: The method for obtaining the cycle length includes: The ratio of the cycle length to the vehicle speed at the current moment is taken as the historical cycle interval at the current moment; taking the current moment as the starting point and the historical cycle interval as the time domain length, the historical cycle time points are obtained in the negative direction of the time domain.

7. A highway crack detection method according to claim 1, characterized in that: The preset grayscale threshold interval is [100, 200].

Images