Unmanned aerial vehicle measurement-based traffic accident scene record map generation method and system

Abstract

The invention discloses a traffic accident scene record map generation method and system based on unmanned aerial vehicle measurement, and the method comprises the steps: enabling an unmanned aerial vehicle to be attached to the ground, enabling the focusing center line of a camera to be perpendicular to the ground, determining the shooting height, and placing a standard ruler control point O1 of a corresponding measurement grade at an unshielded part, a panoramic image T1 with the height F1 is shot, and a three-dimensional photogrammetry range T1 is selected. And if the sheltered area exists and the condition is allowed, lowering the unmanned aerial vehicle to F2 to shoot an image T2 of the sheltered area. And identifying image control points and standard ruler size data, fusing T1 and T2 to generate a base map of a record map with a real size, and marking the size to complete the generation of the record map. If a large-range detail image needs to be shot, a control point range is calculated in a frame selection area of T1, control points are arranged after falling to F3, and after shooting and splicing are conducted to form T3, the T3 and T1 are aligned in proportion to generate a complete detail image. The panorama and detail images of the accident scene can be quickly obtained, and a reliable basis is provided for accident handling and judicial expertise.

Description

Technical Field

[0001] The present invention relates to the technical field of road traffic accident handling, and more specifically, to a method and system for generating a traffic accident scene record diagram based on drone measurement. Background Art

[0002] In the work of road traffic accident handling, the evidence obtained from on-site investigation is crucial for accident determination, liability division, and judicial adjudication. The traditional means of on-site investigation of road traffic accidents mainly rely on manual measurement and drawing of on-site diagrams, and at the same time, photography and videography techniques are used to fix evidence. This method has significant defects: the manual operation process is cumbersome, and it takes a lot of time from on-site measurement to the completion of the on-site record diagram. On sections with heavy traffic flow, long-term on-site investigation will seriously affect traffic order, increasing the risk of traffic congestion and secondary accidents; moreover, manual measurement is easily affected by subjective factors, and it is difficult to guarantee the accuracy and integrity of data. In a complex accident scene, key evidence may also be missed, hindering subsequent accident handling and forensic appraisal work.

[0003] With the development of drone technology, its application in on-site investigation of road traffic accidents has gradually attracted attention. Drones can flexibly obtain images of the accident scene, take pictures from different angles and heights, with relatively high image resolution and a wide shooting range. Compared with traditional manual drawing, using the images taken by drones as the base map of the accident scene can reduce the work intensity of traffic police on-site drawing. However, there are still many problems in the actual application of current drone photogrammetry technology: during the shooting process, it is difficult to balance the relationship between the aerial shooting range, elevation, and resolution of the drone, and it is impossible to meet the requirements of a large shooting range, high resolution, and appropriate shooting elevation at the same time; in addition, obstacles such as trees and buildings around the accident scene will interfere with the shooting, resulting in the inability to obtain a complete image of the accident scene and affecting the comprehensive analysis of the accident scene.

[0004] At the same time, although the existing three-dimensional reconstruction technology based on drone photogrammetry can realize the virtual reconstruction of the accident scene, which is conducive to evidence retention, the reconstruction process takes a long time, and after the on-site evacuation, the authenticity of the reconstruction data is difficult to verify, and the key dimension data cannot be confirmed in time on-site, which is likely to cause subsequent disputes. Therefore, the current technical means cannot meet the requirements of rapid, accurate, and complete on-site investigation for road traffic accident handling, and a new technical solution is urgently needed to solve these problems. Summary of the Invention

[0005] This invention aims to address the challenges of road accident scene investigation. Addressing the low efficiency, high risk, and insufficient data integrity of traditional investigation methods, as well as the challenges of drone photogrammetry, this method uses drone-based imagery, combined with algorithms to achieve image recognition, dimension verification, and stitching correction, generating orthographic projections of the actual scene. This approach shortens investigation time, reduces personnel risks, improves data accuracy and efficiency, promotes automated and intelligent accident scene investigation, and provides a reliable basis for accident handling and forensic identification.

[0006] In response to the above-mentioned defects or improvement needs of the prior art, the present invention provides a method for generating a traffic accident scene recording image based on drone measurement, comprising:

[0007] S1. Keep the drone close to the ground and adjust the camera's focus centerline to be perpendicular to the ground.

[0008] S2. Determine the shooting height and place at least one standard ruler control point of the corresponding measurement level suitable for the height resolution in an unobstructed position, denoted as O1;

[0009] S3. Take a clear panoramic image T1 of the traffic accident area at a height F1, and the operator selects the three-dimensional photogrammetry shooting range Ω1 in T1;

[0010] S4. If there is an obstructed area and conditions permit, the operator lowers the drone to a height F2 where it is not obstructed and can capture the obstructed area and the location of control point O1; the shooting range is selected and recorded as D x ×D y , and take an image T2 of the area;

[0011] S5. Identify the control points in the image and confirm the specific size data in the standard ruler. Based on this, the images T1 and T2 taken at heights F1 and F2 are proportionally fused to generate an orthographic projection of the road traffic accident scene with real-size data information and no obstructions. The operator can select points in the map to measure and mark the size, thereby completing the generation of the road traffic accident scene record map;

[0012] S6. If a large-scale detailed image is needed, the operator selects the area to be photographed D in image T1. x ×D y , calculate the control point layout range Z pq , lower the drone to altitude F3 and set up control point O pq , drone takes and automatically stitches a series of partial images under F3 S mn Create image T3, align T3 and T1 proportionally to generate region D x ×D y Full detail drawing with true to size data.

[0013] Furthermore, the method for generating a traffic accident scene record diagram based on UAV measurement further includes:

[0014] The UAV automatically performs three-dimensional photogrammetry within the range Ω1, and the measurement data is used for the three-dimensional reconstruction of the traffic accident scene in the later stage; and after the shooting is completed, all the control points arranged are retrieved.

[0015] Furthermore, the method for placing at least one standard scale control point suitable for the corresponding measurement level of this height resolution in S2 is:

[0016] Use a fusion size comparison standard scale as the standard scale;

[0017] Arrange centimeter-level or decimeter-level measurement standard scales for shooting at heights above 100m;

[0018] Arrange millimeter-level or centimeter-level measurement standard scales for shooting at heights of 10 - 100m;

[0019] Arrange 0.1 millimeter-level or millimeter-level measurement standard scales for shooting at heights of 0 - 10m.

[0020] Furthermore, the fusion size comparison standard scale specifically is:

[0021] Use contrasting colors to fill the standard width size as the size comparison standard scale, and at the same time use it as the control point for image stitching and correction. To distinguish and facilitate the identification of different control points, fill in different numbers, colors, and shapes corresponding to different measurements. The size values under different measurements are greater than 1 unit value of this measurement level and are as close as possible to the value of the higher measurement level for easy identification; the numbers, graphics, and colors of different measurement levels under the same number are the same, and the size ratio is the same as the measurement level ratio.

[0022] Furthermore, the specific method for proportionally fusing the images T1 and T2 taken at the heights F1 and F2 in S5 is:

[0023] Record the measurement level that can be clearly identified by the resolution at the elevation F1 as centimeters, and on the standard scale The corresponding imaging size is The focal length of the picture taken at the elevation F1 is f1. Record the measurement level that can be clearly identified by the resolution at the elevation F2 as millimeters, and on the standard scale The corresponding imaging size is The focal length of the picture taken at the elevation F2 is f2. Record any size on the ground as D, the imaging size of the size D at the elevation F1 as d1, and the imaging size at the elevation F2 as d2. Then there is:

[0024]

[0025] Taking the control point O1 as the image stitching reference point and the completely overlapping of the scaled control point pattern as the image stitching and calibration standard, the occluded area of the image taken at elevation F1 can be filled with the image taken at elevation F2, and a synthetic image with relatively accurate relevant dimensions can be obtained.

[0026] Further, in step S6, the method for taking a large-scale detailed image is as follows:

[0027] The shooting range of a digital camera is limited by the sensor image plane size w×h, the focal length f, and the object distance, that is, the elevation F; the object space field angles are respectively recorded as 2α and 2β, and the size ranges W and H that can be shot along the x-axis are respectively:

[0028]

[0029] For an unmanned aerial vehicle (UAV) photography camera, any object distance, that is, the elevation F, can be obtained from any dimension D on the ground standard scale obtained by photography at this elevation 0 and the imaged dimension d 0 and can be used to obtain the elevation F from any distinguishable dimension on the standard scale:

[0030] F = f * D 0 / d 0

[0031] Substituting the formula for F into the formulas for W and H, the values of W and H can be obtained;

[0032] For the occluded area C of the image taken at elevation F1, the area D that needs to be taken at an unoccluded elevation and includes the area C and the control point O1 is framed x × D y , D x , D y The corresponding dimensions d 1x , d 1y in the image taken at elevation F1 can be measured in the image, so

[0033] D x = d 1x × D1 / d1

[0034] D y = d 1y × D1 / d1

[0035] Assume that the UAV can clearly and unobstructedly capture the ground traces at elevation F3, and any dimension on the ground standard scale obtained by photography The imaged dimension The shooting range of a single photo is W3 × H3, then W3 and H3 are:

[0036]

[0037] In addition,

[0038]

[0039] Due to the small F3 value, in this case, the range that can be captured in a single photo is limited due to the field of view, and the area D needs to be captured. x ×D y Divide into m×n partitions and calculate the maximum overlap rate C x , C y ;

[0040] Take the upper left corner o of the shooting range as the origin, and

[0041] Point (pW3-(mW3-D x ), qH3-(nH3-D y )),

[0042] Point (pW3, qH3-(nH3-D y )),

[0043] Point (pW3, qH3),

[0044] Point (pW3-(mW3-D x ), qH3) enclosed by the square area Z pq Set a series of control points O within (1≤p≤m-1,1≤q≤n-1) pq And the corresponding area S mn Photography can meet the needs of later image stitching and correction.

[0045] Furthermore, in S5, the area D needs to be photographed. x ×D y Divide into m×n partitions and calculate the maximum overlap rate C x , C y The specific method is:

[0046] Considering the need for stitching and correction of adjacent partition images and the highest possible precision of the synthesized image, the values of m and n should satisfy the following requirements:

[0047]

[0048] Among them, roundup means rounding up;

[0049] Therefore, the maximum overlap ratio C between the partitioned image and the photographed area in the x and y directions is x , C y for:

[0050]

[0051]

[0052] For the convenience of calculation, the dimensions x and y of the maximum overlapping area satisfy: x = m * W3 - D x , y = m * H3 - D y .

[0053] Furthermore, the specific method for post - image stitching and correction is as follows:

[0054] The image is recognized by the YOLOv5 model to accurately locate the dimension map area, and the STR scene text recognition method is used to recognize the text information in the image to confirm the specific dimension data;

[0055] The SIFT feature extraction and matching algorithm is used to extract and match feature points of the collected images, establish the geometric relationship between the images, and the image fusion method is used for stitching and fusion to generate an orthographic projection accident scene real - view record map base map without occlusion.

[0056] As the second aspect of the present invention, a traffic accident scene record map generation system based on unmanned aerial vehicle (UAV) measurement is provided, including:

[0057] The UAV attitude adjustment unit is used to make the UAV close to the ground and adjust the focus center line of the photographic camera to be perpendicular to the ground;

[0058] The shooting height determination and control point placement unit is used to determine the shooting height and place at least one standard scale control point O1 corresponding to the measurement level suitable for the height resolution at an unobstructed position;

[0059] The panoramic image shooting and range selection unit is used to shoot a clear panoramic image T1 of the traffic accident range at a height F1, and the operator selects the three - dimensional photogrammetry shooting range Ω1 in T1;

[0060] The occluded area processing unit is used to, if there is an occluded area and conditions permit, the operator lowers the UAV to a height F2 where it is not occluded and can shoot the area including the occluded area and the position where the control point O1 is located; the shooting range is selected and denoted as D x ×D y , and shoot the image T2 of this area;

[0061] The panoramic map base map generation unit is used to identify the control points in the images, confirm the specific dimension data in the standard scale, and fuse the images T1 and T2 taken at heights F1 and F2 in proportion to generate an orthographic projection road traffic accident scene real - view record map base map with real dimension data information and without occlusion; the operator can select points in the map for dimension measurement and annotation, and thus complete the generation of the road traffic accident scene real - view record map.

[0062] Process a large-range detailed image unit. If it is necessary to capture a large-range detailed image, the operator frames the area D to be captured within the image T1. x ×D y , calculate the layout range Z of the control points. pq , lower the UAV to the height F3 to deploy the control points O. pq , the UAV captures and automatically stitches a series of local images S under F3. mn to form the image T3, and align T3 and T1 proportionally to generate the area D. x ×D y A complete detailed map with real-size data information.

[0063] As a third aspect of the present invention, there is also provided a computer-readable storage medium, on which a computer program is stored, and the computer program is executed by a processor to implement the method for generating a traffic accident scene record map based on UAV measurement described in any one of the above.

[0064] Generally speaking, compared with the prior art through the above technical solutions conceived by the present invention, the following beneficial effects can be achieved:

[0065] 1. The method for generating a traffic accident scene record map based on UAV measurement of the present invention adopts an innovative design of control points and standard rulers, improving the accuracy of image stitching and the accuracy of size measurement. The control point pattern integrating the function of the size comparison standard ruler fills the standard width size with a contrasting color as the size comparison standard ruler, and also has the function of control points for image stitching and correction. Control points of different measurement levels are applicable to the UAV photography resolution requirements at different elevations, and high-precision image stitching and correction are achieved through the image coincidence of control points with the same number. In practical applications, this design reduces the size error of the generated road traffic accident real-scene record map, and also improves the recognition rate of the standard ruler image recognition and the size text recognition, effectively guaranteeing the reliability and accuracy of the data in accident handling and forensic identification.

[0066] 2. The method for generating a traffic accident scene record map based on UAV measurement of the present invention effectively solves the problem of image acquisition in occluded areas and areas with unclear details by adopting a targeted secondary acquisition strategy. When encountering an occluded area, the method will set control points with standard rulers near the occluded area. The UAV first takes a panoramic view at a higher elevation F1. If there is occlusion, it will descend to an appropriate elevation F2 to capture the occluded area and the control points. Through image scaling based on the standard ruler and stitching correction based on the control points, a precise composite image is obtained. For unclear detail areas, the operator selects an area in the initial panoramic image, lowers the UAV to an appropriate height F3, deploys control points according to the projection range and captures a series of local images. These images are processed and fused with the original image, making the final real-scene record map complete and clear, providing an accurate basis for accident handling.

[0067] 3. The method for generating a traffic accident scene record diagram based on UAV measurement of the present invention realizes the high efficiency and accuracy of traffic accident scene investigation by adopting a multi-technology collaborative operation mode. In the image acquisition stage, the flexible shooting characteristics of the UAV are utilized to obtain high-resolution images, providing a rich data basis for subsequent processing. Then, with the help of the YOLOv5 model and the STR scene text recognition algorithm, the size diagram area can be accurately located and the text information can be recognized, and then the specific size data can be confirmed, greatly improving the accuracy of data acquisition. The SIFT feature extraction and matching algorithm is used for image stitching and correction, effectively solving problems such as the shooting range, resolution, and occlusions during UAV aerial photography, and finally generating an orthophoto accident scene real-scene record diagram without occlusions, providing a reliable basis for accident handling and forensic identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a flowchart of the method for generating a traffic accident scene record diagram based on UAV measurement according to an embodiment of the present invention;

[0069] Figure 2 It is a schematic diagram of the control point pattern integrating the function of the size comparison standard scale according to an embodiment of the present invention;

[0070] Figure 3 It is a schematic diagram of the control point numbering of the same measurement level standard scale according to an embodiment of the present invention (taking centimeter level / decimeter level as an example);

[0071] Figure 4 It is a schematic diagram of the image taken at a certain elevation F1 according to an embodiment of the present invention;

[0072] Figure 5 It is a schematic diagram of the image near the occluded area C in the image taken at a certain elevation F2 of the image taken at elevation F1 according to an embodiment of the present invention;

[0073] Figure 6 It is a schematic diagram of the field of view angle and shooting range according to an embodiment of the present invention;

[0074] Figure 7 It is the area D of an embodiment of the present invention x ×D y Schematic diagram;

[0075] Figure 8 It is a schematic diagram of the overlapping area according to an embodiment of the present invention;

[0076] Figure 9 It is a schematic diagram of the image area before and after stitching according to an embodiment of the present invention;

[0077] Figure 10 It is a schematic diagram of Picture 1 before stitching according to an embodiment of the present invention;

[0078] Figure 11 Schematic diagram of the pre - splicing picture 2 of the embodiment of the present invention;

[0079] Figure 12 Schematic diagram of the post - splicing picture of the embodiment of the present invention;

[0080] Figure 13 Unit diagram of the traffic accident scene recording map generation system based on unmanned aerial vehicle (UAV) measurement of the embodiment of the present invention. Detailed implementation manners

[0081] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0082] Embodiment 1

[0083] Embodiment 1 of the present invention provides a method for generating a traffic accident scene recording map based on UAV measurement, including the following steps:

[0084] S1. Make the UAV close to the ground and adjust the focus center line of the photographic camera to be perpendicular to the ground;

[0085] S2. Determine the shooting height, and place at least one standard scale control point suitable for the corresponding measurement level of the height resolution at an unobstructed position, denoted as O1;

[0086] S3. Take a clear panoramic image T1 of the traffic accident area at height F1, and the operator frames the three - dimensional photogrammetric shooting range Ω1 in T1;

[0087] S4. If there is an occluded area and conditions permit, the operator lowers the UAV to a height F2 that is unobstructed and can photograph the occluded area and the position where the control point O1 is located; frame the shooting range and denote it as D x ×D y , and take an image T2 of this area;

[0088] S5. Identify the control points in the images and confirm the specific dimension data in the standard scale. Accordingly, fuse the images T1 and T2 taken at heights F1 and F2 in proportion to generate an orthophoto road traffic accident scene real - view record map base map with real - size data information and no occlusion; the operator can select points in the map for dimension measurement and annotation, and thus complete the generation of the road traffic accident scene real - view record map;

[0089] S6. If it is necessary to capture a large-range detailed image, the operator frames the area D to be captured in the image T1 x ×D y , calculates the layout range Z of the control points pq , lowers the UAV to the height F3 to layout the control points O pq , the UAV captures and automatically stitches a series of local images S under F3 mn into the image T3, and aligns T3 and T1 proportionally to generate the area D x ×D y A complete detailed map with real-size data information.

[0090] The specific process will be elaborated in the following embodiments:

[0091] (1) UAV two-dimensional photogrammetry principle

[0092] The principle of UAV two-dimensional photogrammetry is that when the camera takes a picture perpendicular to the ground, the ratio of the size D of the ground being photographed to its imaging size d is equal to the ratio of the distance (i.e., object distance or elevation) F between the camera and the ground to the camera focal length f. Therefore, if there is a standard ruler on the ground, taking any size D on it 0 , the imaging size is d 0 , for a certain measured size D on the ground, the imaging size is d, then there is:

[0093]

[0094] Therefore, D can be calculated through D 0 , d 0 , d.

[0095] (2) Standard ruler and control point setting

[0096] Traditional standard rulers usually use ordinary invar rulers, and their scales are difficult to distinguish when the UAV is at a high altitude from the ground. Please refer to Figure 2 , in this paper, a control point pattern integrating the function of a size comparison standard ruler is designed as shown in Figure 2 , using contrasting colors to fill the standard width size as a size comparison standard ruler, and at the same time serving as control points for image stitching and correction. Please refer to Figure 3 , to distinguish and facilitate the identification of different control points, different numbers, colors and shapes corresponding to different measures are filled in it (the size values under different measures are greater than 1 unit value of this level of measure and are as close as possible to the value of the higher level of measure for easy identification; the numbers, graphics and colors of different measure levels under the same number are the same, and the size ratio is consistent with the measure level ratio).

[0097] On the one hand, it is more convenient to layout the control points and the standard ruler simultaneously. On the other hand, different measurement levels correspond to the resolution requirements suitable for photography at different elevations of the UAV. Based on the directionality of the image design, when the images of the control points with the same number completely overlap, high-precision image stitching and calibration are completed.

[0098] (3) Principle of image stitching and calibration at different elevations

[0099] Please refer to Figure 4 , Figure 4 As shown in the actual scene diagram of the traffic accident between vehicle A and vehicle B photographed at a certain elevation F1, an occluded area C was found during the photography. Therefore, please refer to Figure 5 , place the control point O1 with a standard ruler at a position near the occluded area that does not block the accident traces. The placement position and direction (not limited, but keep still after placement) are as shown in Figure 5 .

[0100] Record the measurement level that can be clearly identified at the resolution at elevation F1 as centimeters, and the corresponding imaging size on the standard ruler is The focal length of the picture taken at elevation F1 is f1. Record the measurement level that can be clearly identified at the resolution at elevation F2 as millimeters, and the corresponding imaging size on the standard ruler is The focal length of the picture taken at elevation F2 is f2. Record any size on the ground as D, the imaging size d1 of size D at elevation F1, and the imaging size d2 at elevation F2. Then there is: The corresponding imaging size is Taking the control point O1 as the image stitching reference point and using the completely overlapping of the scaled control point patterns as the image stitching and calibration standard, the occluded area of the image taken at elevation F1 can be filled with the image taken at elevation F2, and a relatively accurate synthetic image of the relevant size can be obtained.

[0101]

[0102] Taking the control point O1 as the image stitching reference point and using the completely overlapping of the scaled control point patterns as the image stitching and calibration standard, the occluded area of the image taken at elevation F1 can be filled with the image taken at elevation F2, and a relatively accurate synthetic image of the relevant size can be obtained.

[0103] (4) Principle of image partition acquisition, stitching and calibration at the same elevation

[0104] Please refer to Figure 6 , the shooting range of a digital camera is limited by the sensor image plane size w×h, focal length f, and object distance, that is, elevation F; record the object space field angles as 2α and 2β respectively, and the size ranges W and H that can be shot along the x-axis are:

[0105]

[0106] For the UAV photography camera, any object distance, that is, elevation F, can be obtained from any size D on the ground standard ruler obtained by photography at this elevation 0, the size d of the imaged object 0 is obtained, and the elevation F can be obtained from any resolvable size on the standard scale:

[0107] F = f * D 0 / d 0

[0108] Substituting the formula for F into the formulas for W and H, the values of W and H can be obtained;

[0109] The occluded area C shown in the image taken at elevation F1 ( Figure 4 ) is manually or intelligently identified and framed to include the area D that needs to be photographed at an unoccluded elevation, which contains the area C and the control point O1 x × D y , D x , D y The corresponding sizes d 1x , d 1y in the image taken at elevation F1 can be measured in the image. Therefore

[0110] D x = d 1x × D1 / d1

[0111] D y = d 1y × D1 / d1

[0112] Assume that the drone can clearly and unoccludedly photograph the ground traces at elevation F3. Any size on the ground standard scale obtained by photography The range that can be photographed in a single photo is W3 × H3. Then W3 and H3 are:

[0113]

[0114] In addition,

[0115]

[0116] Since the value of F3 is generally small, in this case, affected by the field of view angle, the range that can be photographed in a single photo is limited. The area D to be photographed x × D y is divided into m × n sub - regions. Considering the need for adjacent sub - region image stitching and correction and the highest possible accuracy of the synthesized image, the values of m and n should satisfy:

[0117]

[0118] where roundup means rounding up;

[0119] Therefore, the maximum overlap rate C between the sub - region photographed images and the photographed range in the x and y directionsx , C y for:

[0120]

[0121] For ease of calculation, please refer to Figure 8 ,Pick Figure 8 For the overlapping area scheme shown, the maximum overlapping area dimensions x and y satisfy:

[0122] x=m*W3-Dx

[0123] y=m*H3-D y

[0124] remember Figure 8 In the right figure, the upper left corner o of the shooting range is taken as the origin, and the

[0125] Point (pW3-(mW3-D x ), qH3-(nH3-D y )),

[0126] Point (pW3, qH3-(nH3-D y )),

[0127] Point (pW3, qH3),

[0128] Point (pW3-(mW3-D x ), qH3) enclosed by the square area Z pq Set a series of control points O within (1≤p≤m-1,1≤q≤n-1) pq And the corresponding area S mn (like Figure 9 Photography within the shadow range shown in the figure can effectively meet the needs of later image stitching and correction.

[0129] Please refer to Figure 9 , taking the four images taken near the origin O as an example, in the square area Z 11 Control point O placed inside 11 The clearly distinguishable measurement level is 0.1mm, and 4 images around it are taken separately S 11 、S 21 、S 12 、S 22 , then the control point O 11 Complete the stitching and correction of the four images. Let the size of any ground to be measured in the stitched and corrected image be D, and the imaging size of the size at elevation F3 be d3, then we have

[0130]

[0131] D can be calculated.

[0132] (5) Image acquisition and recognition

[0133] Use a drone to take pictures according to a preset path and altitude, obtain high-resolution accident scene images, and identify the images through the YOLOv5 model, accurately locate the dimension map area, use the STR scene text recognition algorithm to recognize the text information in the images, and confirm the specific dimension data.

[0134] (6) Image stitching and correction

[0135] Design a control point pattern that integrates the function of a dimension comparison standard scale, fill the standard width dimension with a contrast color as the dimension comparison standard scale, and at the same time use it as a control point for image stitching and correction. To distinguish and facilitate the identification of different control points, fill them with different numbers, colors, and shapes corresponding to different measures (the dimension values under different measures are greater than 1 unit value of this level of measure and are as close as possible to the value of the higher level of measure for easy identification; the numbers, graphics, and colors of different measure levels under the same number are the same, and the size ratio is consistent with the measure level ratio).

[0136] Use the SIFT feature extraction and matching algorithm to extract and match feature points of the collected images, establish the geometric relationship between the images, and use image fusion technology for stitching and fusion to generate an orthophoto accident scene real scene record map base map without occlusion. Please refer to Figures 10 - 12 , Figure 10 , Figure 11 , Figure 12 The pictures before and after image stitching and the actual taken pictures are shown as follows.

[0137] According to the foregoing ideas and solutions for key links, the steps for generating an orthophoto road traffic accident scene real scene record map by drone photogrammetry in this embodiment are as follows:

[0138] (1) The drone adjusts its attitude angle close to the ground so that the focus center line of the photographic camera is perpendicular to the ground;

[0139] (2) The drone ascends to a height F1 at which a panoramic orthophoto of the accident area can be taken. The operator observes the situation of taking the panoramic orthophoto at this height through a synchronous computer, and places a standard scale control point corresponding to the measure level suitable for the resolution at this height at a suitable unobstructed position (if there is an obstructed area, then near the adjacent obstructed area) (i.e., O1 in the foregoing text) (generally, a centimeter-level / decimeter-level measure standard scale is arranged for shooting at a height above 100m, a millimeter-level / centimeter-level measure standard scale is arranged for shooting at a height of 10 - 100m, and a 0.1 millimeter-level / millimeter-level measure standard scale is arranged for shooting at a height of 0 - 10m);

[0140] (3) Take a clear panoramic image T of the traffic accident area at height F1 1，The operator outlines the three-dimensional photogrammetry shooting range Ω1 in T1;

[0141] (4) If there is an occluded area and conditions permit, the operator lowers the UAV to a height F2 where it is not occluded and can capture the occluded area and the location of the control point O1. The operator manually or through intelligent recognition on the synchronization computer outlines the shooting range (i.e., the shooting area D including the occluded area C and the control point O1 at the height F2 as mentioned above x ×D y ), and captures the image T2 of this area at the shooting height F2;

[0142] (5) The system automatically fuses the images T1 and T2 captured at the heights F1 and F2 in proportion to generate an orthophoto road traffic accident scene real - scene record map base with real - size data information and no occlusion. The operator can select points on the map for dimension measurement and annotation, thus completing the generation of the road traffic accident scene real - scene record map.

[0143] (6) If it is necessary to capture detailed images of a larger area (such as tire marks), the operator outlines the area D to be captured in the image T1 x ×D y ; the system calculates the layout range Z of each control point pq , then lowers the UAV to a height F3 where detailed images can be clearly captured, and the operator arranges the control points O according to the UAV projection range prompt pq ; after the arrangement is completed, the UAV captures a series of local images S at F3 mn , and automatically stitches them into the image T3 of this area at the height F3. The system aligns T3 and T1 in proportion and generates a complete detailed map of area D x ×D y with real - size data information.

[0144] (7) The UAV automatically performs three - dimensional photogrammetry within the range Ω1 and is used for the later three - dimensional reconstruction of the traffic accident scene;

[0145] (8) After the shooting is completed, all the arranged control points are retrieved.

[0146] This embodiment meets the requirements of rapid evidence collection. The shooting duration of the UAV in this method is within three minutes, and the entire real - scene record map generation duration is within two minutes. After forming an accident scene investigation equipment integrating UAV + on - site pad + computer + back - end cloud support in the later stage, the generation time will be shortened to within 30 seconds. While quickly completing the on - site drawing, it can provide support for on - site photo - taking evidence collection work, such as completing accident overview photography, element photography, and detail photography, and providing a basis for the reproduction of the positional relationship between various elements at the later scene.

[0147] Embodiment 2

[0148] Embodiment 2 provides a traffic accident scene recording map generation system based on UAV measurement, including:

[0149] A UAV attitude adjustment unit, configured to make the UAV close to the ground and adjust the focus center line of the photographic camera to be perpendicular to the ground;

[0150] A determining shooting height and placing control point unit, configured to determine the shooting height, and place at least one standard scale control point corresponding to the measurement level suitable for the resolution of this height at an unobstructed position, denoted as O1;

[0151] A shooting panoramic image and framing range unit, configured to shoot a clear panoramic image T1 of the traffic accident range at a height F1, and an operator frames the three-dimensional photogrammetry shooting range Ω1 in T1;

[0152] An occluded area processing unit, configured to if there is an occluded area and conditions permit, the operator lowers the UAV to a height F2 where it is not occluded and can shoot the area including the occluded area and the position where the control point O1 is located; frame the shooting range and denote it as D x ×D y , and shoot the image T2 of this area;

[0153] A panoramic map base map generation unit, configured to identify the control points in the image, confirm the specific dimension data in the standard scale, and fuse the images T1 and T2 taken at heights F1 and F2 according to the ratio, to generate an orthophoto road traffic accident scene real scene recording map base map with real dimension data information and no occlusion; the operator can select points in the map for dimension measurement and annotation, and then complete the generation of the road traffic accident scene real scene recording map.

[0154] A large-scale detailed image processing unit, configured to if it is necessary to shoot large-scale detailed images, the operator frames the area to be shot D in the image T1 x ×D y , calculates the control point layout range Z pq , lowers the UAV to height F3 to layout the control point O pq , the UAV shoots and automatically stitches a series of local images S at F3 mn into the image T3, and aligns T3 with T1 according to the ratio to generate the area D x ×D y complete detailed map with real dimension data information.

[0155] Embodiment 3

[0156] Embodiment 3 also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, any step of the traffic accident scene recording map generation method based on UAV measurement can be implemented.

[0157] The computer-readable storage medium may include: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs.

[0158] Those skilled in the art can easily understand that the above description is only a preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for generating a traffic accident scene record diagram based on UAV measurement, characterized in that, It includes the following steps: S1. Make the drone close to the ground and adjust the focus center line of the photographic camera to be perpendicular to the ground; S2. Determine the shooting height. At an unobstructed position, place at least one standard scale control point suitable for the corresponding measurement level of this height resolution, denoted as O1; S3. Take a clear panoramic image T1 of the traffic accident area at height F1. The operator frames the three-dimensional photogrammetry shooting range Ω1 in T1; S4. If there is an occluded area and conditions permit, the operator will lower the UAV to a height F2 where it is not occluded and can capture the occluded area and the location of the control point O1; frame the shooting range and denote it as D x ×D y , and capture an image T2 of this area; S5. Identify the control points in the image and confirm the specific dimension data in the standard scale. Accordingly, fuse the images T1 and T2 taken at heights F1 and F2 in proportion to generate an orthophoto road traffic accident scene real-scene record map base map with real dimension data information and no occlusion. The operator can select points in the map for dimension measurement and annotation, and then complete the generation of the road traffic accident scene real-scene record map; S6. If it is necessary to capture a large-scale detailed image, the operator selects the area D to be captured within the image T1 x ×D y , calculates the control point layout range Z pq , lowers the UAV to the height F3 to deploy the control points O pq , the UAV captures and automatically stitches a series of local images S under F3 mn into the image T3, aligns T3 and T1 proportionally to generate the area D x ×D y A complete detailed map with real-size data information.

2. The method for generating a traffic accident scene record diagram based on UAV measurement according to claim 1, wherein, The method for generating a traffic accident scene record map based on drone measurement further includes: The drone automatically performs three-dimensional photogrammetry within the range Ω1, and the measurement data is used for the later three-dimensional reconstruction of the traffic accident scene. After the shooting is completed, all the arranged control points are retrieved.

3. The method for generating a traffic accident scene record diagram based on UAV measurement according to claim 1, wherein The method for placing at least one standard scale control point suitable for the corresponding measurement level of this height resolution in S2 is: Use a fused dimension comparison standard scale as the standard scale; Place a centimeter-level or decimeter-level measurement standard scale at a height above 100m; Place a millimeter-level or centimeter-level measurement standard scale at a height of 10 - 100m; Place a 0.1-millimeter-level or millimeter-level measurement standard scale at a height of 0 - 10m.

4. The method for generating a traffic accident scene record diagram based on UAV measurement according to claim 3, wherein The fused dimension comparison standard scale is specifically: Use contrasting colors to fill the standard width dimension as the dimension comparison standard scale, and at the same time use it as the control point for image stitching and calibration. To distinguish and facilitate the identification of different control points, fill in different numbers, colors, and shapes corresponding to different measurements. The dimension values under different measurements are greater than 1 unit value of this level of measurement and are as close as possible to the value of the higher level of measurement for easy identification. The numbers, graphics, and colors of different measurement levels under the same number are the same, and the size ratio is the same as the measurement level ratio.

5. The method for generating a traffic accident scene record diagram based on UAV measurement according to claim 1, wherein, The specific method for fusing the images T1 and T2 taken at heights F1 and F2 in proportion in S5 is: The measurement level that can be clearly identified at elevation F1 is in centimeters. On the standard scale The corresponding imaging size is The focal length of the picture taken at elevation F1 is f1. The measurement level that can be clearly identified at elevation F2 is in millimeters. On the standard scale The corresponding imaging size is The focal length of the picture taken at elevation F2 is f2. Denote any size on the ground as D, the imaging size of size D at elevation F1 as d1, and the imaging size at elevation F2 as d2. Then we have: Use the control point O1 as the image stitching reference point, and use the completely overlapping scaled control point pattern as the image stitching and calibration standard, that is, the image taken at elevation F2 can fill the occluded area of the image taken at elevation F1 to obtain a synthetic image with relatively accurate relevant dimensions.

6. The method for generating a traffic accident scene record diagram based on UAV measurement according to claim 1, characterized in that, The specific method for taking large-range detailed images in S6 is: The shooting range of the digital camera is restricted by the sensor image plane size w×h, focal length f, and object distance, that is, elevation F. Denote the object space field angles as 2α and 2β respectively. The size ranges W and H that can be shot along the x-axis are: For a drone photography camera, for any object distance, i.e., elevation F, it can be obtained from any dimension D on the ground standard scale obtained by photography at this elevation 0 , the imaged dimension d 0 and thus the elevation F can be obtained from any resolvable dimension on the standard scale: F = f * D 0 / d 0 Substitute formula F into formula W and formula H to obtain the values of W and H; The occluded area C in the image taken at elevation F1, and the area D to be photographed at an unoccluded elevation that encloses area C and the control point O1 x ×D y , D x 、D y The corresponding dimension d in the image taken at elevation F1 1x 、d 1y can be measured in the image, so D x = d 1x × D1 / d 1 D y = d 1y × D1 / d1 Assume that the UAV can clearly capture the ground traces without occlusion at the elevation F3, and any dimension on the ground standard scale obtained by photography The imaged dimension The shooting range of a single photo is W3×H3, then W3 and H3 are: Another, Since the value of F3 is small, in this case, affected by the field of view angle, the shooting range of a single photo is limited, and the area D to be photographed x ×D y is divided into m×n partitions, and the maximum overlap rate C x , C y ; Denote the upper left corner point o of the shooting range as the origin, and in the object plane with Point (pW3-(mW3-D x ),qH3-(nH3-D y )) Point (pW3, qH3 - (nH3 - D y )) point (pW3, qH3), The square area Z enclosed by the points (pW3 - (mW3 - D x ), qH3), where (1 ≤ p ≤ m - 1, 1 ≤ q ≤ n - 1), is provided with a series of control points O pq (1 ≤ p ≤ m - 1, 1 ≤ q ≤ n - 1), and corresponding area S pq is photographed, which can meet the needs of later image stitching and correction. mn ​ 7. The method for generating a traffic accident scene record diagram based on drone measurement according to claim 6, wherein The area D to be photographed x ×D y is divided into m×n sub-regions, and the maximum overlap rate C x , C y is calculated. The specific method is as follows: Considering the need for adjacent partition image stitching and calibration and the highest possible accuracy of the synthetic image, the values of m and n should satisfy: where roundup means rounding up; Therefore, the maximum overlap rate C between the partitioned captured images in the x and y directions and the captured range x , C y is as follows: For the convenience of calculation, the maximum overlapping region sizes x and y satisfy: x = m * W3 - D x , y = m * H3 - D y .

8. The method for generating a traffic accident scene record diagram based on UAV measurement according to claim 6, characterized in that, The specific method for later image stitching and calibration is: The image is recognized by the YOLOv5 model to accurately locate the dimension map area, and the STR scene text recognition method is used to recognize the text information in the image to confirm the specific dimension data; Using the SIFT feature extraction and matching algorithm, feature points of the collected images are extracted and matched to establish the geometric relationship between the images, and the image fusion method is used for stitching and fusion to generate an orthophoto accident scene real - time record map base map without occlusion.

9. A traffic accident scene record diagram generation system based on UAV measurement, characterized in that, Including: The UAV attitude adjustment unit is used to make the UAV close to the ground and adjust the focus center line of the photographic camera to be perpendicular to the ground; The unit for determining the shooting height and placing control points is used to determine the shooting height, and at an unobstructed position, at least one standard scale control point suitable for the corresponding measurement level of the height resolution is placed, denoted as O1; The unit for shooting panoramic images and selecting the range is used to shoot a panoramic image T1 of the clear traffic accident range at a height F1, and the operator selects the three - dimensional photogrammetry shooting range Ω1 in T1; Occlusion area processing unit. If there is an occlusion area and conditions permit, the operator will lower the UAV to a height F2 where it is not occluded and can capture the occlusion area and the location of the control point O1; select the shooting range and denote it as D x ×D y , and capture the image T2 of this area; The unit for generating the panoramic map base map is used to identify the control points in the image and confirm the specific dimension data in the standard scale, and accordingly, the images T1 and T2 taken at heights F1 and F2 are fused proportionally to generate an orthophoto road traffic accident scene real - time record map base map with real dimension data information and no occlusion; the operator can select points in the map for dimension measurement and annotation, and thus complete the generation of the road traffic accident scene real - time record map; Process a large-range detailed image unit. For taking a large-range detailed image if needed, the operator frames the area D to be photographed in image T1 x ×D y , calculate the layout range Z of the control points pq , lower the UAV to height F3 to layout the control points O pq , the UAV takes and automatically stitches a series of local images S under F3 mn to form image T3, align T3 and T1 proportionally to generate area D x ×D y A complete detailed map with real-size data information.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program is executed by a processor to implement the method for generating a traffic accident scene record map based on UAV measurement according to any one of claims 1 - 8.

Images