Method for positioning mounting point of safety rope of electric power iron tower based on stereoscopic vision of unmanned aerial vehicle

Abstract

The invention relates to the technical field of image analysis, and discloses an unmanned aerial vehicle stereoscopic vision-based electric iron tower safety rope mounting point positioning method, which comprises the following steps of: constructing a mathematical mapping relation between a space plane and image characteristics based on a motion camera observation principle, and identifying a scene plane image of an electric iron tower top end plane observed by an airborne camera, the method comprises the following steps: establishing a ridge line set by taking a straight line segment in a proper scene plane image as a ridge line on which a safety rope can be mounted, and searching to obtain an optimal mounting straight line segment as an optimal mounting point of the safety rope on an electric power iron tower only by traversing the ridge line set. The problem that in the traditional process of achieving safety rope fixing operation through an electric power inspection unmanned aerial vehicle, safety rope mounting points are manually selected, and universality for different iron towers is lacked is solved.

Description

Technical Field

[0001] This invention relates to the field of image analysis technology, specifically to a method for locating the anchoring point of a safety rope on a power transmission tower based on UAV stereo vision. Background Technology

[0002] Power transmission towers are key infrastructure supporting the safe transmission of electricity. They are widely distributed in cities, rural areas, and mountainous regions, and are formed by continuous towers that together create a power network.

[0003] During long-term operation, power transmission towers are affected by various factors such as natural environmental erosion and equipment aging, requiring regular maintenance and inspection by tower workers. When tower workers are carrying out complex work such as maintenance and inspection at heights, they use safety ropes installed by themselves to ensure their own safety from falls.

[0004] However, the traditional manual installation method of tower fall arrest safety ropes has many drawbacks that cannot be ignored. These include the low efficiency of manual installation at high altitudes due to the weight and ease of assembly of the tower fall arrest safety ropes, and the fact that the stability of the tower fall arrest safety rope installation is easily affected by human factors when installed manually.

[0005] In existing technologies, although power line inspection drones are used to automate and secure safety rope installation, the location of the safety rope attachment point during this process is still manually selected. After manual selection, the power line inspection drone is remotely controlled to precisely deploy the safety rope to the top of the tower. Manual selection of the safety rope attachment point relies on accumulated experience, lacks standardized procedures, and is not universally applicable to different power towers.

[0006] This invention provides a method for locating the safety rope attachment point of power transmission towers based on UAV stereo vision, in order to solve the above-mentioned problems. Summary of the Invention

[0007] The purpose of this invention is to provide a method for locating the attachment point of the safety rope on the power tower based on the stereo vision of the UAV, so as to solve the problem that the traditional method of relying on power inspection UAVs to fix the safety rope, which relies on manual selection of the attachment point, lacks universality for different power towers.

[0008] This invention is achieved through the following technical solution:

[0009] The method for locating the safety rope attachment point on a power transmission tower based on UAV stereo vision includes the following steps:

[0010] Step A1: For the scene plane image of the top plane of the power tower observed by the airborne camera, generate the scene plane equation and construct the homography transformation matrix of the same projection point in the inter-frame image, obtain the homogeneous coordinates of the same straight line segment in the scene plane as the time frame changes, and integrate and construct the stereo distance calculation formula from the same straight line segment in the inter-frame image to the camera optical principal plane.

[0011] Step A2: Extract the line segments from each frame image to construct a line segment set, and perform temporal matching on all line segments in the line segment set of the inter-frame images to obtain several sets of line segment pairs, including pairs of line segments.

[0012] Step A3: Calculate and select the straight segment pairs with the minimum stereo distance in each set of straight segment pairs, and use the straight segments in the current frame as the ridges on which safety ropes can be attached. Combine the stereo distance to perform hierarchical clustering to obtain a set of ridges including several straight segments; where the stereo distance is calculated using the stereo distance calculation formula.

[0013] Step A4: Traverse the straight segments in the ridge line set and perform risk assessment. Select the straight segment with the smallest risk assessment value as the optimal mounting straight segment. Generate the optimal mounting point coordinates of the safety rope from the geometric midpoint coordinates of the optimal mounting straight segment.

[0014] Furthermore, in step A1, the constructed homography transformation matrix The expression is:

[0015]

[0016] In the formula, Let be the translation vector in the airborne camera coordinate system. , Let x be the translation vector along the x-axis. Let be the translation vector along the y-axis. Let z be the translation vector along the z-axis. It is a real number; The rotation matrix in the airborne camera coordinate system. , It is a three-dimensional special orthogonal group, and the translation vector and rotation matrix are obtained in real time by the RTK system of the power line inspection drone;

[0017] Let be the intrinsic parameter matrix of the airborne camera, and d be the distance from the scene plane to the principal plane of the airborne camera. Let be the scene plane normal vector, when the optical axis of the airborne camera is perpendicular to the scene plane. .

[0018] Furthermore, in step A1, the formula for calculating the stereo distance between the same straight line segment between frames is as follows:

[0019]

[0020] In the formula, To utilize Time segment and straight line segments exist The line segment corresponding to the time Calculate the line segment Euclidean distance to the camera's principal plane;

[0021] a straight line segment The homogeneous coordinates are Line segment of time exist The time segment is obtained through time sequence matching. The homogeneous coordinates are: One moment corresponds to one time frame;

[0022] a straight line segment Parameterized points on, , , operator This means converting a three-dimensional vector in homogeneous coordinates to a two-dimensional vector in Euclidean coordinates. The inverse matrix of the camera intrinsic parameters is transposed. , operator This represents the vector cross product operation.

[0023] Furthermore, in step A2, during the temporal matching of all line segments in the inter-frame images, a temporal matching optimization algorithm is constructed to integrate the direction similarity filtering results, length similarity filtering results, and position similarity filtering results of the inter-frame line segments, obtain the joint probability of temporal matching of line segments in the inter-frame images, establish a probability threshold, exclude line segment pairs to be matched whose joint probability is less than the probability threshold, thereby selecting the final effective matching line segment pairs and establishing a set of line segment pairs.

[0024] Furthermore, the direction similarity filtering result is obtained based on the direction similarity calculation of inter-frame straight line segments. During the direction similarity calculation process, an angle difference tolerance for direction similarity is set. ;

[0025] The length similarity filtering result is obtained based on the length similarity calculation of line segments between frames. In the length similarity calculation process, the length ratio threshold is set to 0.8.

[0026] The position similarity filtering result is obtained based on the position similarity calculation of line segments between frames. During the position similarity calculation process, the spatial constraint radius of position similarity is set to 80 pixels.

[0027] Furthermore, in step A3, a distance threshold is established to filter the straight line segments corresponding to the ridge. The established distance threshold is 50 pixels and the depth tolerance is 20%.

[0028] Furthermore, in step A4, during the traversal of the ridge set, a risk assessment function based on a multi-objective optimization model is established to assess the risk of straight segments. The multi-objective optimization model is equipped with multiple engineering constraints, and the risk assessment value of the straight segments is calculated based on the joint calculation of multiple engineering constraints.

[0029] Furthermore, the process of assessing the risk of the straight line segment using a mount point risk assessment function based on a multi-objective optimization model, and finding the optimal mount point, includes:

[0030] Step B1: After preprocessing the ridge line set as input, obtain the line segments to be evaluated;

[0031] Step B2: Construct a risk assessment function for the mounting point based on a multi-objective optimization model. The multi-objective optimization model incorporates multiple engineering constraints. The risk assessment value of the straight line segment to be assessed is calculated jointly based on these multiple engineering constraints.

[0032] Step B3: Based on the risk assessment value of each straight line segment, construct the optimal attachment point decision function, traverse the ridge line set, and select the straight line segment with the smallest assessment value as the optimal attachment straight line segment. Generate the optimal attachment point coordinates of the safety rope from the geometric midpoint coordinates of the optimal attachment straight line segment.

[0033] Furthermore, in step B2, the constructed mount point risk assessment function for:

[0034]

[0035] In the formula, For the proximity term of the climbing ladder, it represents the straight line segment. The coordinates of the midpoint and the distance to the climbing ladder;

[0036] The term "interference count for inclined steel" indicates the interference on the straight section. The number of inclined steel bars present in the vicinity;

[0037] The length adaptation term represents a straight line segment. The length adaptation value;

[0038] This represents the interference penalty coefficient for inclined steel. , Indicates the length-adaptive weights. .

[0039] Furthermore, the expression for the optimal mount point decision function is as follows:

[0040]

[0041] In the formula, The line segment corresponding to the minimum evaluation value, which is also the optimal mounting line segment. For the ridge line set, For mount point risk assessment function;

[0042] The optimal attachment point coordinates of the safety rope are based on the optimal attachment straight segment. Geometric midpoint Coordinate generation, the expression is as follows:

[0043]

[0044] In the formula, The optimal mount point coordinates; For the optimal mounting straight segment homogeneous coordinates, operators This means converting a three-dimensional vector in homogeneous coordinates into a two-dimensional vector in Euclidean coordinates.

[0045] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0046] 1. This invention is based on the observation principle of motion cameras. It constructs a mathematical mapping relationship between spatial planes and image features. By identifying the scene plane image of the top plane of the power tower obtained by the airborne camera, the straight line segments in the suitable scene plane image are used as ridges for attaching safety ropes. A set of ridges is established. Then, by traversing the set of ridges, the optimal straight line segment for attaching the safety rope can be found as the optimal attachment point for the safety rope on the power tower. This solves the problem that in the traditional process of using power inspection drones to fix safety ropes, the attachment point for the safety rope is not universally applicable to different towers due to the reliance on manual selection.

[0047] 2. After obtaining the ridge line suitable for attaching the safety rope, this invention constructs a multi-objective optimization model to accurately locate the safety rope attachment point. The optimal straight segment is the ridge line that satisfies the following constraints: accessibility (distance from the climbing ladder), safety (number of adjacent inclined steel beams), and reliability (structural stability). The optimal straight segment is the one that meets the minimum constraint condition (i.e., the minimum risk assessment value). The optimal attachment point coordinates of the safety rope are generated from the geometric midpoint coordinates of the optimal attachment straight segment, thus completing the location of the safety rope attachment point on the power tower. The multi-objective optimization model ensures that the final selected optimal straight segment is located on the edge of the top plane. Furthermore, the multi-objective optimization model ensures that this invention conforms to the "high attachment, low use" principle and meets the following standard: GB 30862-2014. Attached Figure Description

[0048] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:

[0049] Figure 1 This is a flowchart of a method for locating the safety rope attachment point of a power transmission tower based on UAV stereo vision, as proposed in an embodiment of this application. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0051] A method for locating safety rope attachment points on power transmission towers based on UAV stereo vision, implementing, for example... Figure 1 As shown, it includes the following steps:

[0052] Step A1: For the scene plane image of the top plane of the power tower observed by the airborne camera, generate the scene plane equation and construct the homography transformation matrix of the same projection point in the inter-frame image, obtain the homogeneous coordinates of the same straight line segment in the scene plane as time frames change, and integrate and construct the stereo distance calculation formula from the same straight line segment in the inter-frame image to the camera's optical principal plane.

[0053] Specifically, the construction of the homography transformation matrix includes the following process:

[0054] by At any given time, the coordinate system of the onboard camera of the power inspection drone is used as the reference coordinate system. For a scene plane observed by the onboard camera, such as the top surface and part of the side plane of the power tower, the following scene plane equation is generated:

[0055]

[0056] In the formula, For the scene plane normal vector, This represents any point on the scene plane when the optical axis of the airborne camera is perpendicular to the scene plane. , This is the distance from the scene plane to the main plane of the airborne camera.

[0057] In this embodiment, an orthogonal coordinate system conforming to the right-hand rule is used as the airborne camera coordinate system. The axis system of the constructed airborne camera coordinate system is specifically as follows:

[0058] Origin: The position of the camera's optical center;

[0059] Axis: Parallel to the lateral movement direction of the drone;

[0060] Axis: Determined by the directions of the x and z axes, approximately pointing in the opposite direction to the drone's nose;

[0061] Axis: Pointing towards the observed target along the optical axis. Airborne camera intrinsic parameter matrix. The expression is: In the formula, Focal length The coordinates of the main point.

[0062] In this embodiment, one moment corresponds to one time frame. The projection point of the scene plane on the image at any moment ,to Corresponding projection points on the time image Construct the homography transformation matrix as follows. To describe the mapping relationship of the same projection point in consecutive frames:

[0063]

[0064] In the formula, Let be the translation vector in the airborne camera coordinate system. , Let X be the translation vector along the X-axis. Let be the translation vector along the Y-axis. Let Z be the translation vector. It is a real number; The rotation matrix in the airborne camera coordinate system. , It is a three-dimensional special orthogonal group, where the translation vector and rotation matrix are real-time motion parameters, and are all acquired in real time by the RTK system of the power line inspection drone.

[0065] The mapping relationship between the corresponding projection points based on the homography transformation matrix is ​​as follows: .

[0066] Based on the established homography transformation matrix, the scene plane... Line segment of time Obtaining through time-series matching The corresponding straight line segment at time ,in, , , , All are homogeneous coordinates. , via a straight segment to straight line segment Based on the changes in the image plane, the following formula is established for calculating the distance from a spatial line segment to the camera's principal optical plane (i.e., a plane parallel to the image plane and containing the camera's center (origin)):

[0067]

[0068] In the formula, To utilize Time segment and straight line segments exist The line segment corresponding to the time Calculate the line segment Euclidean distance to the camera's principal plane;

[0069] a straight line segment Parameterized points on, , , operator This operator converts a 3D vector in homogeneous coordinates to a 2D vector in Euclidean coordinates. This means converting a two-dimensional vector in Euclidean coordinates to a vector in homogeneous coordinates. The inverse matrix of the camera intrinsic parameters is transposed. , operator This represents the vector cross product operation.

[0070] Step A2: Extract the line segments from each frame image to construct a line segment set, and perform temporal matching on all line segments in the line segment set of the inter-frame images to obtain several sets of line segment pairs, including pairs of line segments.

[0071] In this embodiment, the EDLines line detection algorithm based on Edge Drawing is used to process the images of the airborne camera to obtain the set of line segments of each frame image at consecutive time points. During the detection process of the EDLines line detection algorithm, the parameters are set as gradient threshold = 25° and minimum line segment length = 15 pixels.

[0072] For time t, the constructed set of line segments The expression is as follows:

[0073]

[0074] In the formula, Let be the coordinates of the endpoints of the line segment, and be a point in the homogeneous coordinate system. , .

[0075] In this embodiment, during the temporal matching of line segments in inter-frame images, a temporal matching optimization algorithm is constructed, which incorporates a three-layer filtering mechanism: direction similarity filtering, length similarity filtering, and position similarity filtering. The temporal matching optimization algorithm integrates the direction similarity filtering results, length similarity filtering results, and position similarity filtering results for inter-frame line segments to obtain the joint probability of temporal matching of line segments in the inter-frame images, and establishes a probability threshold. The matching line segment pairs that have a joint probability less than the probability threshold are excluded, thereby filtering out the final valid matching line segment pairs and establishing a set of line segment pairs.

[0076] The set of line segments at time t and the set of line segments at time t+1. Taking the temporal matching of line segments as an example, the three-layer filtering mechanism in the temporal matching optimization algorithm includes direction similarity filtering, length similarity filtering, and position similarity filtering. The specific details of the three-layer filtering mechanism are as follows:

[0077] In the process of calculating directional similarity, the angle difference of directional similarity is set. Tolerance is For a straight line segment and line segments Their directional similarity The calculation formula is as follows:

[0078]

[0079] In the formula, To calculate the direction vector function of line segment p, To calculate the direction vector function of line segment q;

[0080] In the length similarity calculation process, the length ratio threshold is set to... =0.8, for a straight line segment and line segments ,make and They represent line segments respectively. and line segments The length of the similarity is calculated as follows:

[0081]

[0082] In the process of calculating positional similarity, the spatial constraint radius of positional similarity is set. For a straight line segment, the value is 80 pixels. and line segments The formula for calculating their positional similarity is as follows:

[0083]

[0084] In the formula, a straight line segment The midpoint coordinates, a straight line segment The coordinates of the midpoint.

[0085] Finally, by integrating directional similarity, length similarity, and positional similarity, the joint probability of temporal matching of line segments in inter-frame images is obtained. The expression is as follows:

[0086]

[0087] After completing time-series matching, and after completing probability threshold-based matching... After filtering, a set of line segment pairs is obtained. The expression is as follows:

[0088]

[0089] Finally, temporal matching is performed on all line segments in the set of line segments in the inter-frame images to obtain several sets of paired line segment pairs.

[0090] Step A3: Calculate and select the pair of straight segments with the minimum stereo distance from the set of each pair of straight segments (a pair of straight segments refers to a pair of straight segments at time t and t+1), and use the straight segments in the current frame as the ridges on which safety ropes can be attached. Combine the stereo distance to perform hierarchical clustering to obtain a set of ridges including several straight segments; where the stereo distance is calculated by the stereo distance calculation formula.

[0091] Specifically, after calculating the 3D distances between all pairs of line segments, the following set of 3D distances is constructed:

[0092]

[0093] From the set of three-dimensional distances In the middle, select the straight line segment with the smallest solid distance according to the following formula:

[0094]

[0095] Finally, using hierarchical clustering, with straight line segments as ridges, a set of ridges is constructed as follows:

[0096]

[0097] In the formula, Distance threshold Pixels, depth tolerance 20%.

[0098] Step A4: Traverse the set of ridge lines, perform risk assessment on each straight line segment, and select the straight line segment with the smallest risk assessment value as the optimal mounting straight line segment. Generate the optimal mounting point coordinates of the safety rope from the geometric midpoint coordinates of the optimal mounting straight line segment.

[0099] In this embodiment, during the traversal of the ridge set, a risk assessment function based on a multi-objective optimization model is established to assess the risk of straight line segments. The multi-objective optimization model incorporates multiple engineering constraints.

[0100] Specifically, the selection of the safety rope attachment point must meet the following three engineering constraints:

[0101] Accessibility constraints require that the attachment point be close to the climbing ladder, meaning that attachment points close to the climbing ladder should be selected first.

[0102] Safety constraints dictate that the mounting point must avoid the inclined steel of the power tower; that is, the selected mounting point must avoid the interference area of ​​the inclined steel.

[0103] Due to reliability constraints, the mounting point must have a stable structure, meaning that the mounting point must be selected from crossbars with a stable structure.

[0104] The process of traversing the ridge set, assessing the risk of straight segments using a mount point risk assessment function based on a multi-objective optimization model, and finding the optimal mount point includes:

[0105] Step B1: After preprocessing the ridge line set as input, obtain the line segments to be evaluated.

[0106] The preprocessing of the input ridge set includes:

[0107] Endpoint normalization: Traversing the set of ridges For each endpoint of a line segment, the coordinates of the endpoints are converted from homogeneous coordinates to Euclidean coordinates, as shown in the following expression:

[0108]

[0109] In the formula, , This refers to projective space.

[0110] Length calculation: Calculate the set of ridges Each straight line segment The Euclidean length is expressed as follows:

[0111]

[0112] In the formula, Represents a straight line segment The second point, Represents a straight line segment The first point, Represents the magnitude of a vector;

[0113] Median benchmark: Calculating the set of ridgeable lines The median of the Euclidean length of each line segment is used as the reference length for the analysis of ridge line segments. The calculation formula is as follows:

[0114]

[0115] In the formula, This indicates taking the median of the input sequence. Represents the set of ridges The k-th line segment in;

[0116] Step B2: Construct a risk assessment function for the mounting point based on a multi-objective optimization model. This model incorporates multiple engineering constraints. The risk assessment value for each straight line segment to be evaluated is calculated jointly based on these constraints. , Based on the risk assessment value, the suitability of the straight segment as the mounting point can be evaluated, ensuring that the mounting point is located on a standard-length crossbar with structural stability. A smaller risk assessment value indicates greater suitability and structural stability. In this embodiment, the constructed mounting point risk assessment function... for:

[0117]

[0118] In the formula, For the proximity term of the climbing ladder, it represents the straight line segment. The distance between the midpoint of the ladder and the climbing ladder is expressed as follows:

[0119]

[0120] in, Let the coordinates be the midpoint of the line segment. , Preset ladder positions for users;

[0121] In the formula, The term "interference count for inclined steel" indicates the interference on the straight section. The number of inclined steel bars existing in the vicinity of the given value is expressed as follows:

[0122]

[0123] in, Represents the nearest neighbor determination function. , In this embodiment, the distance threshold is represented. Set to 30 pixels;

[0124] In the formula, The length adaptation term represents a straight line segment. The length adaptation value is expressed as follows:

[0125]

[0126] In the formula, This represents the interference penalty coefficient for inclined steel. , Indicates the length-adaptive weights. .

[0127] Step B3: Based on the risk assessment value of each straight line segment, construct the optimal attachment point decision function, traverse the ridge line set, and select the straight line segment with the smallest assessment value as the optimal attachment straight line segment. Generate the optimal attachment point coordinates of the safety rope from the geometric midpoint coordinates of the optimal attachment straight line segment.

[0128] Specifically, during the traversal of the ridge set, the expression for the optimal mount point decision function is as follows:

[0129]

[0130] In the formula, The line segment corresponding to the minimum evaluation value, which is also the optimal mounting line segment. For the ridge line set, For mount point risk assessment function;

[0131] Traversing the ridge set based on the optimal mount point decision function All straight line segments are compared and their evaluation values ​​are determined. Select the line segment corresponding to the minimum evaluation value. As the optimal mounting straight segment, and based on the optimal mounting straight segment Geometric midpoint The coordinates for generating the optimal attachment point of the safety rope are expressed as follows:

[0132]

[0133] In the formula, The optimal mount point coordinates; For the optimal mounting straight segment homogeneous coordinates, operators This means converting a three-dimensional vector in homogeneous coordinates into a two-dimensional vector in Euclidean coordinates.

[0134] Finally, the location of the safety rope attachment point on the power tower was completed, and the safety rope was attached to the attachment point by the power inspection drone.

[0135] This invention is based on the observation principle of motion cameras. It constructs a mathematical mapping relationship between spatial planes and image features. By identifying the scene plane image of the top plane of the power tower obtained by the airborne camera, the straight line segments in the suitable scene plane image are used as ridges for attaching safety ropes. A set of ridges is established. Then, by traversing the set of ridges, the optimal straight line segment for attachment can be found as the optimal attachment point for the safety rope on the power tower.

[0136] In this invention, the established distance calculation formula from a straight line segment to the scene plane breaks through the spatial mapping bottleneck of traditional detection methods. By introducing motion geometry constraints, it enables two-dimensional image features to acquire three-dimensional distance measurement capabilities. Through spatiotemporal joint optimization, it overcomes problems such as foreground occlusion and background interference in single-frame images, providing reliable input for mount point decision-making. Simultaneously, the distance parameter... The precise calculations provided crucial spatial prior information for subsequent ridgeline determination.

[0137] After obtaining the ridge line suitable for attaching the safety rope, this invention constructs a multi-objective optimization model to accurately locate the safety rope attachment point. The optimal straight segment is the ridge line that satisfies the following constraints: accessibility (distance from the climbing ladder), safety (number of adjacent inclined steel beams), and reliability (structural stability). The optimal straight segment is the one that meets the minimum constraint condition (i.e., the minimum risk assessment value). The optimal attachment point coordinates of the safety rope are generated from the geometric midpoint coordinates of the optimal attachment straight segment, thus completing the location of the safety rope attachment point on the power transmission tower. The multi-objective optimization model ensures that the final selected optimal straight segment is located on the edge of the top plane. Furthermore, the multi-objective optimization model ensures that this invention conforms to the "high attachment, low use" principle and meets the following standard: GB 30862-2014.

[0138] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for positioning the hanging point of a safety rope of a power tower based on the stereovision of a UAV, characterized in that, Comprising the following steps: Step A1: generating a scene plane equation and constructing a homography matrix of the same projection point in the inter-frame image for the scene plane image of the power tower top plane observed by the on-board camera, obtaining the homogeneous coordinates of the same line segment in the scene plane changing over time frames, and integrating to construct a stereo distance calculation formula of the same line segment to the camera optical principal plane; Step A2: extracting the line segments of each frame image to construct a line segment set respectively, and performing time sequence matching on all line segments in the line segment set of the inter-frame image to obtain a plurality of line segment pair sets including pairs of line segments; Step A3: calculating and selecting the line segment pair with the smallest stereo distance in each line segment pair set, and taking the line segment under the current frame as the ridge line of the mountable safety rope, and performing hierarchical clustering based on the stereo distance to obtain a ridge line set including a plurality of line segments; Step A4: traversing the line segments in the ridge line set and performing risk assessment, and taking the line segment with the smallest risk assessment value as the optimal mounting line segment, and generating the optimal mounting point coordinates of the safety rope from the geometric midpoint coordinates of the optimal mounting line segment.

2. The method for positioning the power tower safety line attachment point based on the UAV stereo vision of claim 1, wherein, In step A1, the constructed single transformation matrix has the expression: wherein, is a translation vector in the onboard camera coordinate system, , is a translation vector of the x-axis, is a translation vector of the y-axis, is a translation vector of the z-axis, is a real number; is a rotation matrix in the onboard camera coordinate system, , is a three-dimensional special orthogonal group, and the translation vector and the rotation matrix are both obtained in real time by an RTK system of the power inspection unmanned aerial vehicle. is the intrinsic matrix of the on-board camera, d is the distance from the scene plane to the principal plane of the on-board camera, is the normal vector of the scene plane, when the optical axis of the on-board camera is perpendicular to the scene plane, . 3.The method of positioning the power tower safety line attachment point based on the UAV stereo vision of claim 2, wherein, In step A1, the stereo distance calculation formula between the same line segments in the inter-frame is as follows: In the formula, To use The straight line segment at the moment And the straight line segment At The straight line segment corresponding to the moment , Calculate the Euclidean distance of the straight line segment To the camera main plane; the homogeneous coordinates of the straight line segment at time t the homogeneous coordinates of the straight line segment at time t the homogeneous coordinates of the straight line segment at time t the homogeneous coordinates of the straight line segment at time t obtained through time matching at time t the homogeneous coordinates of the straight line segment at time t wherein one time corresponds to one time frame; is a parametric point on a straight line segment , , , the operator denotes the conversion of the homogeneous coordinate system three-dimensional vector in it to a Euclidean coordinate system two-dimensional vector, is the transpose of the inverse camera intrinsic matrix, , the operator denotes the vector cross product operation.

4. The method of claim 2, wherein, In step A2, in the process of time sequence matching of all line segments in the inter-frame image, a time sequence matching optimization algorithm is constructed to integrate the direction similarity filtering result, the length similarity filtering result and the position similarity filtering result of the inter-frame line segments, obtain the time sequence matching joint probability of the line segments of the inter-frame image, establish a probability threshold, exclude the to-be-matched line segment pairs with a joint probability less than the probability threshold, thereby screening out the final effective matching line segment pairs, and establish a line segment pair set.

5. The method of positioning a power tower safety line attachment point based on UAV stereo vision of claim 4, wherein, The direction similarity filtering result is obtained based on direction similarity calculation of inter-frame straight line segments, and an angle difference tolerance of the direction similarity is set to ; The length similarity filtering result is obtained based on the length similarity calculation of the inter-frame line segments, and in the length similarity calculation process, the length ratio threshold is set to 0.8; The position similarity filtering result is obtained based on the position similarity calculation of the inter-frame line segments, and in the position similarity calculation process, the spatial constraint radius of the position similarity is set to 80 pixels.

6. The method of claim 3, wherein the method further comprises: In step A3, a distance threshold is established to screen the line segments corresponding to the ridge line, the established distance threshold is 50 pixels, and the depth tolerance is 20%.

7. The method of claim 1, wherein, In step A4, in the process of traversing the ridge line set, a mounting point risk assessment function based on a multi-objective optimization model is established to perform risk assessment on the line segments, and the multi-objective optimization model is loaded with multiple engineering constraint conditions to jointly calculate the risk assessment value of the line segments.

8. The method of claim 7, wherein, The process of performing risk assessment on the line segments by the mounting point risk assessment function based on the multi-objective optimization model and finding the optimal mounting point includes: Step B1: taking the ridge line set as input and obtaining the to-be-evaluated line segment after preprocessing; Step B2: constructing a mounting point risk assessment function based on a multi-objective optimization model, the multi-objective optimization model is loaded with multiple engineering constraint conditions, and the risk assessment value of the to-be-evaluated line segment is jointly calculated based on the multiple engineering constraints; Step B3: Based on the risk assessment value of each straight line segment, construct the optimal attachment point decision function, traverse the ridge line set, and select the straight line segment with the smallest assessment value as the optimal attachment straight line segment. Generate the optimal attachment point coordinates of the safety rope from the geometric midpoint coordinates of the optimal attachment straight line segment.

9. The method of claim 8, wherein, In step B2, the constructed mount point risk assessment function is: wherein is a ladder proximity term representing the distance of the midpoint coordinate of the straight line segment to the ladder. For the diagonal steel interference count item, indicates the number of diagonal steels existing in the adjacent range of the straight line segment ​ L is the length adaptation term, representing the length adaptation value of the straight segment . denotes the slant steel interference penalty coefficient, , denotes the length adaptability weight, .

10. The method of claim 8, wherein, The expression for the optimal mount point decision function is as follows: In the formula, is the straight line segment corresponding to the minimum evaluation value, that is, the optimal mounting straight line segment, is the ridge line set, is the mounting point risk evaluation function; The optimal attachment point coordinates of the safety line are based on the geometric midpoint of the optimal attachment straight line segment Coordinates are generated by the expression:​ wherein is the optimal mounting point coordinate; is the optimal mounting straight line segment homogeneous coordinate, the operator denotes the conversion of the homogeneous coordinate system three-dimensional vector into the Euclidean coordinate system two-dimensional vector.

Images