A computer vision-assisted method for detecting the inclination of buildings around foundation pits

Abstract

The present invention relates to the field of posture measurement technology, and specifically to a method for detecting the inclination of buildings around a foundation pit with the assistance of computer vision. After determining the spot area of ​​each target, the method determines the movement characteristics of the light plate area between adjacent inspection processes, and the abnormal positioning targets corresponding to the abnormal movement characteristics can be screened out by analyzing the deviation of the movement characteristics between targets. The positioning coordinates of the abnormal positioning targets can be further corrected in combination with the difference in movement characteristics between adjacent normal targets. The degree of inclination of the building is determined by the movement change trend of the positioning coordinates of the targets in each longitudinal target monitoring chain. The present invention ensures the accuracy of the positioning information by correcting the positioning coordinates of the targets, thereby obtaining more accurate inclination detection results.

Description

Technical Field

[0001] The present invention relates to the technical field of posture measurement, and in particular to a method for detecting the inclination of buildings around a foundation pit with the aid of computer vision. Background Art

[0002] During the implementation of the foundation pit project, the surrounding soil will be disturbed to varying degrees, which may cause the surrounding land to sink or surface tension cracks to appear, which may cause the buildings around the foundation pit to tilt. As the construction progresses, the degree of land subsidence and the severity of cracks will change to a certain extent, and the impact on surrounding buildings will also change accordingly.

[0003] When using computer vision to assist in monitoring the tilt of surrounding buildings during foundation pit construction, the existing technology usually uses a camera network posture measurement system to analyze the change in the position of the building's target light spot, thereby obtaining the building's tilt. However, in actual operation, due to the heavy dust in the foundation pit construction environment and the influence of factors such as light, the target light spot may be blurred or missing, resulting in errors in the obtained light spot position information, resulting in obvious error accumulation when the camera network finally integrates all information, and it is impossible to accurately analyze the tilt of the building. Summary of the invention

[0004] In order to solve the technical problem that when using a camera network to detect the tilt of a building, the target light spot is affected by noise, resulting in errors in the coordinate information, which in turn leads to inaccurate analysis of the tilt degree, the purpose of the present invention is to provide a computer vision-assisted method for detecting the tilt of buildings around a foundation pit. The technical solution adopted is as follows: The present invention proposes a method for detecting the inclination of buildings around a foundation pit with the aid of computer vision, the method comprising: A plurality of luminous targets are arranged in the longitudinal direction of the building to form a plurality of longitudinal target monitoring chains; a shooting unit is deployed in each longitudinal target monitoring chain, and the image acquisition direction of the shooting unit is always perpendicular to the ground; the shooting unit is used to obtain a monitoring image containing each target light spot area in each inspection; For each target, obtain the movement characteristics of the spot area between adjacent inspection processes; obtain the positioning abnormality of each target according to the movement characteristic deviation between the targets in the longitudinal target monitoring chain and screen out the targets with abnormal positioning and normal targets; For each abnormally positioned target, according to the difference in moving direction between the abnormally positioned target and an adjacent normal target, and the degree of abnormal positioning of the abnormally positioned target, a positioning adjustment weight of each abnormally positioned target is obtained; and the positioning coordinates of each abnormally positioned target are adjusted according to the positioning adjustment weight; According to the movement change trend of the positioning coordinates of each target in each longitudinal target monitoring chain during different inspection processes, the severity of the building's tilt is obtained.

[0005] Furthermore, the shooting unit is connected to the bracket via a hinge; one end of the bracket is connected to the hinge, and the other end is installed on the surface of the building.

[0006] Furthermore, the shooting unit is a camera with two viewing angles, upper and lower. In a longitudinal target monitoring chain, there is a shooting unit between every two targets, and each target can obtain monitoring images from two viewing angles, upper and lower.

[0007] Furthermore, the method for acquiring the light spot area includes: For two monitoring images of a target, a threshold segmentation algorithm is used to obtain an initial light spot area in each monitoring image, and an intersection area of ​​the initial light spot areas of the two monitoring images is used as the light spot area.

[0008] Furthermore, the method for acquiring the movement feature includes: For each target, the lateral offset and longitudinal offset of the center point of the light spot area between adjacent inspection processes and the movement vector of the center point are obtained; the lateral offset, longitudinal offset and movement direction constitute the movement feature.

[0009] Furthermore, the method for obtaining the positioning abnormality degree includes: In a longitudinal target monitoring chain, the movement vectors of each target are summed to obtain an overall movement vector; For each target, the average lateral offset and the average longitudinal offset of all targets in the same longitudinal target monitoring chain are obtained, the difference between the lateral offset of each target and the average lateral offset is taken as the lateral offset deviation, and the difference between the longitudinal offset of each target and the average longitudinal offset is taken as the longitudinal offset deviation; the product of the lateral offset deviation and the longitudinal offset deviation is taken as the offset feature; The angle value between the movement vector and the corresponding overall movement vector is used as the degree of direction difference; and the product of the offset feature and the degree of direction difference is used as the positioning abnormality.

[0010] Furthermore, the method for obtaining the positioning adjustment weight includes: For each positioning abnormal target, in the same longitudinal target monitoring chain, the average positioning abnormality of the upper-level normal target of the positioning abnormal target is used as reference data, and the sum vector of the movement vectors of the upper-level normal target of the positioning abnormal target is used as a reference vector; Obtain the positioning anomaly degree of the positioning anomaly target and the positioning anomaly degree difference between the reference data; use the angle value between the moving vector of the positioning anomaly target and the reference vector as the moving direction anomaly degree; normalize the product of the positioning anomaly degree difference and the moving direction anomaly degree difference to obtain the positioning adjustment weight.

[0011] Further, adjusting the positioning coordinates of each abnormally positioned target according to the positioning adjustment weight includes: For each abnormal positioning target, if the difference between the lateral offset of the abnormal positioning target and the average lateral offset of the normal targets in the same longitudinal target monitoring chain is greater than a preset threshold, the product of the lateral offset of the center point of the spot area of ​​the abnormal positioning target between adjacent inspection processes and the positioning adjustment weight is used as the lateral adjustment amount, and the center point of the spot area of ​​the abnormal positioning target is moved in the reverse direction according to the lateral adjustment amount; otherwise, the center point of the spot area of ​​the abnormal positioning target is moved forward according to the lateral adjustment amount; similarly, the longitudinal coordinate of the center point of the spot area of ​​the abnormal positioning target is adjusted in the same way, and the center point of the spot area corresponding to the adjusted abnormal positioning target is used as the positioning coordinate of the adjusted abnormal positioning target.

[0012] Furthermore, the method for obtaining the severity of the tilt includes: After each inspection, the longitudinal target monitoring chain where the target with the largest horizontal coordinate movement of the positioning coordinate is located is used as the transverse reference chain, and the longitudinal target monitoring chain where the target with the largest vertical coordinate movement of the positioning coordinate is located is used as the longitudinal reference chain; For the lateral reference chain, the average moving distance of all targets between adjacent inspection processes is taken as the overall lateral displacement feature between adjacent inspection processes, and the overall lateral displacement features of all inspection processes are counted to obtain the overall lateral displacement feature sequence; according to the change trend of the overall lateral displacement feature sequence and the size of the elements in the sequence, the lateral displacement degree of the building is obtained; Similarly, for the longitudinal reference chain, the degree of longitudinal displacement of the building is obtained; The maximum moving direction angle of the target on the building during all inspections is taken as the maximum tilt angle; The severity of the tilt is obtained according to the lateral displacement degree, the longitudinal displacement degree and the maximum tilt angle.

[0013] Furthermore, the method for obtaining the lateral displacement degree includes: Obtain the accumulated value of the elements in the overall lateral displacement characteristic sequence; obtain the differential sequence of the overall lateral displacement characteristic sequence, and use the average value of the elements in the differential sequence as the displacement increment degree; and use the product of the accumulated value and the displacement increment degree as the lateral displacement degree.

[0014] The present invention has the following beneficial effects: The embodiment of the present invention takes into account that the targets on the same target monitoring chain should have similar displacement characteristics. Therefore, after determining the spot area of ​​each target, the present invention determines the movement characteristics of the light plate area between adjacent inspection processes, and the positioning abnormal target corresponding to the abnormal movement characteristics can be screened out by analyzing the deviation of the movement characteristics between targets. The positioning coordinates of the positioning abnormal target can be further corrected by combining the difference in movement characteristics between adjacent normal targets, wherein the abnormal degree of the current positioning abnormal target is determined by the information of the two dimensions of movement direction difference and positioning abnormality, that is, the greater the abnormal degree, the more adjustment is required, so the positioning abnormal target can be effectively corrected by the positioning adjustment weight. Then, the tilt degree of the building is determined by the movement change trend of the positioning coordinates of the target in each longitudinal target monitoring chain. The present invention ensures the accuracy of the positioning information by correcting the positioning coordinates of the target, thereby obtaining a more accurate tilt detection result. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] 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.

[0016] Figure 1 A schematic diagram of a method of longitudinally installing a target on a building provided by an embodiment of the present invention; Figure 2 A flow chart of a method for detecting the inclination of buildings around a foundation pit assisted by computer vision provided by an embodiment of the present invention; Figure 3 A schematic diagram of the installation of a photographing unit provided by an embodiment of the present invention; Figure 4 A schematic diagram of a shooting unit operation provided by an embodiment of the present invention; Figure 5 A monitoring image schematic diagram provided by an embodiment of the present invention; Figure 6 A schematic diagram of a threshold segmentation result provided by an embodiment of the present invention; Figure 7 A schematic diagram of the displacement of a light spot area corresponding to a target provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0017] 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 specific implementation method, structure, features and effects of a computer vision-assisted foundation pit surrounding building tilt detection method proposed by the present invention in combination 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.

[0018] 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.

[0019] The embodiment of the present invention is aimed at detecting the tilt of a building in a construction site. It mainly determines the tilt information of the building by evenly setting targets on the building and monitoring the positioning information of the targets through computer vision technology. Figure 1 , which shows a schematic diagram of longitudinally installing targets on a building provided by an embodiment of the present invention. In the embodiment of the present invention, evenly spaced targets are vertically installed along the contour line of the longitudinal direction of the building.

[0020] The following is a detailed description of a computer vision-assisted method for detecting the inclination of buildings around a foundation pit provided by the present invention in conjunction with the accompanying drawings.

[0021] See also Figure 2 , which shows a flow chart of a method for detecting the inclination of buildings around a foundation pit assisted by computer vision according to an embodiment of the present invention, the method comprising: Step S1: multiple luminous targets are set in the longitudinal direction of the building to form multiple longitudinal target monitoring chains; a shooting unit is deployed in each longitudinal target monitoring chain, and the image acquisition direction of the shooting unit is always perpendicular to the ground; in each inspection, the shooting unit is used to obtain a monitoring image containing each target spot area.

[0022] In the embodiment of the present invention, a shooting unit is deployed in the longitudinal target monitoring chain so that the shooting unit can capture the image of each target. In order to detect the building by using the movement of the target, it is necessary to ensure that the shooting unit and the target are not in a relatively static state. If the building is tilted, the target will also tilt. Therefore, the embodiment of the present invention needs to ensure that the image acquisition direction of the shooting unit is always perpendicular to the ground, so that when the building is tilted, the spot area of ​​the target captured by the shooting unit will produce a displacement change in the image field of view.

[0023] In one embodiment of the present invention, considering that the tilt of a building is not a rapid process, a daily inspection is set, that is, a monitoring image of the target is collected once a day at the same time point for tilt analysis. After obtaining the detection image, in order to facilitate tilt detection, it is also necessary to calibrate each coordinate in the image, complete the calibration of the image coordinate system, the world coordinate system and other coordinate systems, and determine the coordinate information of each position in the image. In some embodiments of the present invention, after the image is collected, the image quality of the monitoring image can be improved by high-frequency acquisition and image fusion technology.

[0024] Preferably, in the embodiment of the present invention, in order to ensure that the image acquisition direction of the shooting unit is always perpendicular to the ground, the shooting unit is connected to the bracket through a hinge, one end of the bracket is connected to the hinge, and the other end is installed on the surface of the building, so that the shooting unit can be affected by the gravity and ensure that it is always perpendicular to the ground. Figure 3 , which shows a schematic diagram of the installation of a shooting unit provided by an embodiment of the present invention.

[0025] Preferably, in the embodiment of the present invention, the shooting unit is a top and bottom dual-view camera device, see Figure 4 , which shows a schematic diagram of the working of a shooting unit provided by an embodiment of the present invention. In a longitudinal target monitoring chain, there is a shooting unit between every two targets, and each target can obtain monitoring images from two viewing angles, upper and lower. In the subsequent process, the spot area can be jointly determined by the monitoring images from two viewing angles, reducing the analysis error from a single viewing angle. Figure 4 As shown, the dual-lens camera device can capture monitoring images of the target from two perspectives. Figure 4 The parallelogram in the image is the image field of view in the camera. The image fields between the two perspectives are the same, and both have a light spot area corresponding to the target. It should be noted that a laser is also installed on the shooting unit. The function of the laser is to emit laser to the target, so that the target reflects the laser to form a bright spot. Figure 5 As shown, Figure 5 A schematic diagram of a monitoring image provided by an embodiment of the present invention.

[0026] Furthermore, in the embodiment of the present invention, for two monitoring images of a target, a threshold segmentation algorithm is used for each monitoring image to obtain the initial spot area in each monitoring image, and the intersection area of ​​the initial spot areas of the two monitoring images is used as the spot area. Figure 5 It can be seen that through the irradiation of the laser, the target will form an obvious bright spot area in the image field of the shooting unit, so the spot area can be directly segmented by threshold segmentation technology, please refer to Figure 6 , which shows a schematic diagram of a threshold segmentation result provided by an embodiment of the present invention, wherein Figure 6 It is a binary image. Since the embodiment of the present invention mainly analyzes coordinate information and does not involve pixel value information, the binary image can clearly represent the area of ​​the light spot. The coordinate information of the light spot area can be determined by further coordinate fusion of the initial light spot area represented by the binary image.

[0027] Step S2: For each target, obtain the movement characteristics of the spot area between adjacent inspection processes; obtain the positioning abnormality of each target according to the movement characteristic deviation between the targets in the longitudinal target monitoring chain and screen out the positioning abnormal targets and normal targets.

[0028] If the building is tilted, the target will move with the tilt of the building, and the position of the spot area in the monitoring image collected by the shooting unit will change in the image field of view. Figure 7 , which shows a schematic diagram of the displacement of a spot area corresponding to a target provided by an embodiment of the present invention, Figure 7 The middle circle is the spot area, the dotted circle is the moving position, and the moving distance is d. For a longitudinal target monitoring chain, because the targets on the monitoring chain are all arranged vertically in a regular manner, if the building is tilted, all targets on a longitudinal target monitoring chain will have a similar degree of tilt relative to the horizontal plane. Therefore, if there is an abnormal positioning of the target spot area due to factors such as environmental noise, there will be a significant deviation from the position characteristics of other normal targets in the same longitudinal target monitoring chain. Therefore, the embodiment of the present invention first obtains the moving characteristics of the spot area between adjacent inspection processes for each target, that is, according to the position of the spot area obtained in this inspection and the position of the spot area obtained in the previous inspection, the moving characteristics are obtained, wherein the direction and distance between the position of the spot area in the previous inspection and the position of the spot area in this inspection can be used as the moving characteristics. Further, according to the deviation of the moving characteristics between the targets in the longitudinal target monitoring chain, the positioning abnormality of each target can be obtained, and then according to the positioning abnormality, the abnormal targets and normal targets on a longitudinal target monitoring chain can be screened out.

[0029] Preferably, in an embodiment of the present invention, the method for acquiring movement features includes: The monitoring image in the embodiment of the present invention has been calibrated by the coordinate system, so the center point of the spot area can be used as a positioning point to analyze the movement characteristics in adjacent inspection processes. For each target, the lateral offset and longitudinal offset of the center point of the spot area between adjacent inspection processes, as well as the movement vector of the center point are obtained; the lateral offset, longitudinal offset and movement vector constitute the movement feature. That is, in the embodiment of the present invention, the movement between the center points is analyzed separately from the lateral and longitudinal dimensions, which helps to more accurately analyze the deviation of the movement characteristics between targets later. It should be noted that the movement vector is the vector formed by the center point of the previous inspection process pointing to the center point of this inspection process; the lateral offset and longitudinal offset are the projection lengths of the movement vector in the horizontal and vertical directions.

[0030] Furthermore, the method for obtaining the positioning abnormality includes: First, it is necessary to determine the overall movement characteristics in a longitudinal target monitoring chain, and based on the overall movement characteristics, the movement characteristic deviation of each target can be analyzed.

[0031] The embodiment of the present invention analyzes the overall characteristics in the moving direction, and in a longitudinal target monitoring chain, adds up the moving vector of each target to obtain the overall moving vector.

[0032] Further analyze the offset caused by the displacement. For each target, obtain the average lateral offset and average longitudinal offset of all targets in the same longitudinal target monitoring chain. The difference between the lateral offset of each target and the average lateral offset is taken as the lateral offset deviation, and the difference between the longitudinal offset of each target and the average longitudinal offset is taken as the longitudinal offset deviation. Further fuse the deviation information obtained in the two dimensions, and take the product of the lateral offset deviation and the longitudinal offset deviation as the offset feature. The larger the offset feature, the greater the difference in the degree of displacement between the target and the entire monitoring chain. The more likely the target is a positioning abnormality target, and the greater the positioning abnormality should be.

[0033] The angle value between the moving vector and the corresponding overall moving vector is taken as the degree of direction difference. That is, the larger the angle, the greater the difference between the moving direction of the target and the entire monitoring chain, and the greater the degree of direction difference. Therefore, the product of the offset feature and the degree of direction difference is taken as the positioning anomaly.

[0034] In the embodiment of the present invention, the positioning abnormality is normalized, and the abnormality threshold is set to 0.5. If the positioning abnormality of the target is greater than the abnormality threshold, it is judged as a positioning abnormal target, otherwise it is a normal target.

[0035] Step S3: For each positioning abnormal target, obtain the positioning adjustment weight of each positioning abnormal target according to the difference in moving direction between the positioning abnormal target and the adjacent normal target, and the positioning abnormality degree of the positioning abnormal target; and adjust the positioning coordinates of each positioning abnormal target according to the positioning adjustment weight.

[0036] For the abnormal positioning target, its positioning coordinates need to be corrected. If the center point of the spot area of ​​the abnormal target has a large difference in movement direction with the adjacent normal target during the current inspection process, it means that the abnormal target needs more adjustment; and because the positioning abnormality also takes into account the offset deviation between the abnormal positioning target and the same monitoring chain, the positioning adjustment weight of each abnormal positioning target can be determined in combination with the positioning abnormality. That is, the larger the positioning adjustment weight, the greater the degree of adjustment required for the abnormal positioning target.

[0037] Preferably, in one embodiment of the present invention, the method for obtaining the positioning adjustment weight includes: Similar to the above-mentioned deviation acquisition process, the embodiment of the present invention first determines a reference quantity, and determines the relative size of the positioning anomaly degree based on the difference between the reference quantity and the positioning anomaly degree of the positioning anomaly target; and further determines a reference vector as the directional comparison data. For each positioning anomaly target, in the same longitudinal target monitoring chain, the average positioning anomaly degree of the superior normal target of the positioning anomaly target is used as the reference data, and the sum vector of the movement vectors of the superior normal target of the positioning anomaly target is used as the reference vector.

[0038] The positioning anomaly degree difference between the positioning anomaly degree of the positioning anomaly target and the reference data is obtained, that is, the difference between the two is the positioning anomaly degree difference. Because the positioning anomaly degree of the positioning anomaly target must be larger data, and the reference data is the data determined by the normal target, the difference between the two must be greater than 0.

[0039] The angle value between the moving vector of the abnormally positioned target and the reference vector is taken as the abnormality of the moving direction.

[0040] The product of the positioning abnormality difference and the moving direction abnormality difference is normalized to obtain the positioning adjustment weight.

[0041] It should be noted that the data about the angle in the embodiments of the present invention are all numerical values ​​of the angle, that is, there is no unit of angle. The normalization method is also to eliminate the dimension and limit the data range. In the embodiments of the present invention, the normalization method of range standardization is adopted to limit the data between 0 and 1.

[0042] Preferably, in an embodiment of the present invention, adjusting the positioning coordinates of each abnormally positioned target according to the positioning adjustment weight includes: For each abnormal positioning target, if the difference between the lateral offset of the abnormal positioning target and the average lateral offset of the normal target in the same longitudinal target monitoring chain is greater than the preset threshold, it means that the abnormal positioning target at this time has a relatively large displacement error and needs to be reversed. Therefore, the product of the lateral offset of the center point of the spot area of ​​the abnormal positioning target between adjacent inspection processes and the positioning adjustment weight is used as the lateral adjustment amount, and the center point of the spot area of ​​the abnormal positioning target is reversely moved and adjusted according to the lateral adjustment amount. It should be noted that the reverse movement adjustment is to move in the opposite direction of the movement vector of the center point of the adjacent inspection process. Similarly, the forward movement adjustment is to move in the direction of the movement vector.

[0043] On the contrary, if the difference between the lateral offset of the abnormally positioned target and the average lateral offset of the normal targets in the same longitudinal target monitoring chain is not greater than the preset threshold, the center point of the spot area of ​​the abnormally positioned target is moved forward according to the lateral adjustment amount.

[0044] Similarly, the vertical coordinate of the center point of the spot area of ​​the positioning abnormal target is adjusted in the same way, and the center point of the spot area corresponding to the adjusted positioning abnormal target is used as the positioning coordinate of the adjusted positioning abnormal target.

[0045] Step S4: Obtain the severity of the tilt of the building according to the movement change trend of the positioning coordinates of each target in each longitudinal target monitoring chain during different inspection processes.

[0046] After the above steps to adjust the positioning coordinates of the positioning abnormal targets, the positioning coordinates of each target can be regarded as effective information without errors and can be used for tilt detection. Then, in each longitudinal target monitoring chain, the severity of the tilt of the building can be obtained according to the movement change trend of the positioning coordinates of each target in different inspection processes. That is, the greater the degree of change of the positioning coordinates of the target in different inspection processes, and the existence of an increasing trend, the greater the severity of the tilt of the building.

[0047] Preferably, in one embodiment of the present invention, considering that there are multiple longitudinal target monitoring chains on the building, a group of representative monitoring chains can be selected to determine the longitudinal tilt characteristics and transverse tilt characteristics of the building. Therefore, after each inspection, the longitudinal target monitoring chain where the target with the largest horizontal coordinate movement of the positioning coordinate is located is used as the transverse reference chain, and the longitudinal target monitoring chain where the target with the largest vertical coordinate movement of the positioning coordinate is located is used as the longitudinal reference chain.

[0048] For the lateral reference chain, the average moving distance of all targets between adjacent inspection processes is taken as the overall lateral displacement feature between adjacent inspection processes, that is, the larger the overall lateral displacement feature, the larger the displacement of the building. The overall lateral displacement features of all inspection processes are further statistically analyzed to obtain the overall lateral displacement feature sequence. The change trend of the building displacement can be analyzed by analyzing the change trend of the elements in the overall lateral displacement feature sequence. That is, the more drastic the change, and the larger the overall lateral displacement feature, the greater the lateral displacement degree of the building in the lateral direction. Therefore, according to the change trend of the overall lateral displacement feature sequence and the size of the elements in the sequence, the lateral displacement degree of the building is obtained.

[0049] Similarly, the longitudinal displacement degree of the building can be obtained for the longitudinal reference chain.

[0050] Furthermore, the value of the maximum moving direction angle of the target on the building during all inspections is taken as the maximum tilt angle; the tilt severity is obtained according to the lateral displacement degree, the longitudinal displacement degree and the maximum tilt angle. That is, the larger the values ​​of the lateral displacement degree, the longitudinal displacement degree and the maximum tilt angle, the greater the tilt severity. Therefore, in an embodiment of the present invention, the lateral displacement degree, the longitudinal displacement degree and the maximum tilt angle are multiplied and normalized to obtain the tilt severity. The normalization method and the like have been described in the above description of the embodiment of the present invention and will not be repeated here.

[0051] Preferably, in an embodiment of the present invention, the method for obtaining the lateral displacement degree includes: Obtain the cumulative value of the elements in the overall lateral displacement characteristic sequence; obtain the differential sequence of the overall lateral displacement characteristic sequence, that is, the sequence obtained by subtracting the previous element from the next element in the overall lateral displacement characteristic sequence is used as the differential sequence. The average value of the elements in the differential sequence is used as the displacement increment; the product of the cumulative value and the displacement increment is used as the lateral displacement degree. That is, the larger the cumulative value of the elements in the overall lateral displacement characteristic sequence, the more serious the accumulated lateral tilt of the building, and the larger the displacement increment, the worse the lateral tilt change, and the greater the lateral displacement degree. Similarly, the longitudinal displacement degree can be obtained.

[0052] In an embodiment of the present invention, tilt detection is required after each inspection process to determine the tilt severity. After the tilt severity is normalized, a tilt severity threshold is set to 0.5, and buildings with a tilt severity greater than the tilt severity threshold are identified as dangerous buildings. The foundation pit is repaired during the construction process, and the inspection process can be reset after the repair is completed, that is, each subsequent inspection will use this updated process to continuously perform inspections to ensure the safety of buildings around the foundation pit during the entire construction process.

[0053] In summary, after the embodiment of the present invention determines the spot area of ​​each target, it determines the movement characteristics of the light panel area between adjacent inspection processes, and the abnormal positioning targets corresponding to the abnormal movement characteristics can be screened out by analyzing the deviation of the movement characteristics between targets. The positioning coordinates of the abnormal positioning targets can be further corrected by combining the difference in movement characteristics between adjacent normal targets. The degree of inclination of the building is determined by the movement change trend of the positioning coordinates of the targets in each longitudinal target monitoring chain. The present invention ensures the accuracy of the positioning information by correcting the positioning coordinates of the targets, thereby obtaining more accurate inclination detection results.

[0054] 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.

[0055] 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 computer vision-assisted method for detecting the inclination of buildings around a foundation pit, characterized in that: The method comprises: A plurality of luminous targets are arranged in the longitudinal direction of the building to form a plurality of longitudinal target monitoring chains; a shooting unit is deployed in each longitudinal target monitoring chain, and the image acquisition direction of the shooting unit is always perpendicular to the ground; the shooting unit is used to obtain a monitoring image containing each target light spot area in each inspection; For each target, obtain the movement characteristics of the spot area between adjacent inspection processes; obtain the positioning abnormality of each target according to the movement characteristic deviation between the targets in the longitudinal target monitoring chain and screen out the targets with abnormal positioning and normal targets; For each abnormally positioned target, according to the difference in moving direction between the abnormally positioned target and an adjacent normal target, and the degree of abnormal positioning of the abnormally positioned target, a positioning adjustment weight of each abnormally positioned target is obtained; and the positioning coordinates of each abnormally positioned target are adjusted according to the positioning adjustment weight; According to the movement change trend of the positioning coordinates of each target in each longitudinal target monitoring chain during different inspection processes, the severity of the building's tilt is obtained.

2. According to the computer vision-assisted method for detecting the inclination of buildings around a foundation pit according to claim 1, it is characterized in that: The shooting unit is connected to the bracket via a hinge; one end of the bracket is connected to the hinge, and the other end is installed on the surface of the building.

3. The computer vision-assisted method for detecting the inclination of buildings around a foundation pit according to claim 1, characterized in that: The shooting unit is a camera with two viewing angles, upper and lower. In a longitudinal target monitoring chain, there is a shooting unit between every two targets, and each target can obtain monitoring images from two viewing angles, upper and lower.

4. The computer vision-assisted method for detecting the inclination of buildings around a foundation pit according to claim 3, characterized in that: The method for acquiring the light spot area comprises: For two monitoring images of a target, a threshold segmentation algorithm is used to obtain an initial light spot area in each monitoring image, and an intersection area of ​​the initial light spot areas of the two monitoring images is used as the light spot area.

5. The computer vision-assisted method for detecting the inclination of buildings around a foundation pit according to claim 1, characterized in that: The method for acquiring the movement feature includes: For each target, the lateral offset and longitudinal offset of the center point of the light spot area between adjacent inspection processes and the movement vector of the center point are obtained; the lateral offset, longitudinal offset and movement direction constitute the movement feature.

6. The computer vision-assisted method for detecting the inclination of buildings around a foundation pit according to claim 5, characterized in that: The method for obtaining the positioning abnormality degree includes: In a longitudinal target monitoring chain, the movement vectors of each target are summed to obtain an overall movement vector; For each target, the average lateral offset and the average longitudinal offset of all targets in the same longitudinal target monitoring chain are obtained, the difference between the lateral offset of each target and the average lateral offset is taken as the lateral offset deviation, and the difference between the longitudinal offset of each target and the average longitudinal offset is taken as the longitudinal offset deviation; the product of the lateral offset deviation and the longitudinal offset deviation is taken as the offset feature; The angle value between the movement vector and the corresponding overall movement vector is used as the degree of direction difference; and the product of the offset feature and the degree of direction difference is used as the positioning abnormality.

7. The computer vision-assisted method for detecting the inclination of buildings around a foundation pit according to claim 5, characterized in that: The method for obtaining the positioning adjustment weight includes: For each positioning abnormal target, in the same longitudinal target monitoring chain, the average positioning abnormality of the upper-level normal target of the positioning abnormal target is used as reference data, and the sum vector of the movement vectors of the upper-level normal target of the positioning abnormal target is used as a reference vector; Obtain the positioning anomaly degree of the positioning anomaly target and the positioning anomaly degree difference between the reference data; use the angle value between the moving vector of the positioning anomaly target and the reference vector as the moving direction anomaly degree; normalize the product of the positioning anomaly degree difference and the moving direction anomaly degree difference to obtain the positioning adjustment weight.

8. The computer vision-assisted method for detecting the inclination of buildings around a foundation pit according to claim 5, characterized in that: The step of adjusting the positioning coordinates of each abnormally positioned target according to the positioning adjustment weight includes: For each abnormal positioning target, if the difference between the lateral offset of the abnormal positioning target and the average lateral offset of the normal targets in the same longitudinal target monitoring chain is greater than a preset threshold, the product of the lateral offset of the center point of the spot area of ​​the abnormal positioning target between adjacent inspection processes and the positioning adjustment weight is used as the lateral adjustment amount, and the center point of the spot area of ​​the abnormal positioning target is moved in the reverse direction according to the lateral adjustment amount; otherwise, the center point of the spot area of ​​the abnormal positioning target is moved forward according to the lateral adjustment amount; similarly, the longitudinal coordinate of the center point of the spot area of ​​the abnormal positioning target is adjusted in the same way, and the center point of the spot area corresponding to the adjusted abnormal positioning target is used as the positioning coordinate of the adjusted abnormal positioning target.

9. The computer vision-assisted method for detecting the inclination of buildings around a foundation pit according to claim 1, characterized in that: The method for obtaining the tilt severity comprises: After each inspection, the longitudinal target monitoring chain where the target with the largest horizontal coordinate movement of the positioning coordinate is located is used as the transverse reference chain, and the longitudinal target monitoring chain where the target with the largest vertical coordinate movement of the positioning coordinate is located is used as the longitudinal reference chain; For the lateral reference chain, the average moving distance of all targets between adjacent inspection processes is taken as the overall lateral displacement feature between adjacent inspection processes, and the overall lateral displacement features of all inspection processes are counted to obtain the overall lateral displacement feature sequence; according to the change trend of the overall lateral displacement feature sequence and the size of the elements in the sequence, the lateral displacement degree of the building is obtained; Similarly, for the longitudinal reference chain, the degree of longitudinal displacement of the building is obtained; The maximum moving direction angle of the target on the building during all inspections is taken as the maximum tilt angle; The severity of the tilt is obtained according to the lateral displacement degree, the longitudinal displacement degree and the maximum tilt angle.

10. The computer vision-assisted method for detecting the inclination of buildings around a foundation pit according to claim 9, characterized in that: The method for obtaining the lateral displacement degree includes: Obtain the accumulated value of the elements in the overall lateral displacement characteristic sequence; obtain the differential sequence of the overall lateral displacement characteristic sequence, and use the average value of the elements in the differential sequence as the displacement increment degree; and use the product of the accumulated value and the displacement increment degree as the lateral displacement degree.

Images