A method for estimating the location of abnormal targets in power transmission channels

Abstract

This invention proposes a method for estimating the location of abnormal targets in power transmission channels. Two identical cameras are placed facing each other on two adjacent power transmission towers. The two cameras are networked with the same edge intelligent computing terminal and perform calculations locally to estimate the location of abnormal targets in power transmission channels. The method uses the ratio of the differences between the images of the abnormal target captured by the two cameras to determine the location of the abnormal target based on the unique correspondence between the ratio and the distance between the two power transmission towers, given the known distance between the two towers.

Description

Technical Field

[0001] This invention relates to the technical fields of power transmission line anomaly monitoring and edge computing, and in particular to a method for estimating the location of anomaly targets in power transmission channels. Background Technology

[0002] The safe and reliable operation of transmission lines is the foundation for the stable operation of the power system. Transmission lines are mainly distributed in the field and are frequently threatened by abnormal targets such as foreign objects, tower cranes, and construction vehicles. To better address the impact of abnormal targets on transmission lines and reduce power outages, the State Grid Corporation of China has been systematically promoting the construction of smart transmission lines in recent years. Various monitoring devices are installed on the towers to achieve regular monitoring and surveillance of the environment surrounding the transmission lines. However, traditional cloud-based abnormal target detection methods are difficult to adapt to the needs of the power grid. This is mainly because transmission lines have a wide coverage area; uploading all the massive amounts of collected images and video data to the cloud would not only put enormous pressure on communication links but also impose a heavy computational and storage burden on the cloud. To solve this problem, some scholars have proposed a solution based on edge intelligence technology. This involves embedding abnormal target detection models into edge intelligent terminals and intelligent visualization devices on transmission lines, enabling rapid on-site detection and identification of abnormal targets, thereby avoiding the migration of massive amounts of images and videos and effectively saving communication and cloud computing and storage resources.

[0003] With the computing power of edge intelligent terminals, it is now possible to identify abnormal targets around power transmission lines in real time at a rate of over 33 frames per second (the minimum frame rate requirement for real-time monitoring of power transmission lines). However, simply identifying targets is insufficient to accurately determine threats, or rather, it is difficult to effectively discern whether abnormal targets have a threatening movement trend towards the power transmission lines. Therefore, it is urgent to overcome the challenge of developing methods for estimating the spatial coordinates of abnormal targets.

[0004] However, if existing target positioning and dynamic identification schemes are added to the current power transmission line monitoring system, it will greatly increase the cost and computing power burden. For example, existing technologies such as lidar 3D point cloud scanning and binocular depth estimation require a lot of additional costs, and the computing power required to implement the corresponding identification algorithms also exceeds the computing power level of the current power transmission line edge intelligent terminals. Summary of the Invention

[0005] Therefore, considering the actual needs of power transmission line anomaly monitoring, as well as cost and modification difficulties, this invention proposes an anomaly target location estimation scheme for power transmission channels. The proposed scheme, based on a dual-camera visual bias algorithm, achieves rapid estimation of the spatial location of anomaly targets without consuming excessive computing power to reconstruct a 3D model of the site, supporting the detection, alarm, and even early warning of threatening anomalies. It can be built upon existing target recognition algorithms and hardware systems, requiring only minor software and hardware upgrades to re-energize existing equipment; simultaneously, it is lightweight and will not affect the frame rate of video information processing by the edge intelligent gateway.

[0006] The present invention specifically adopts the following technical solution:

[0007] A method for estimating the location of abnormal targets in power transmission channels:

[0008] Two identical cameras are positioned facing each other on adjacent power transmission towers. The two cameras are networked with the same edge intelligent computing terminal and perform calculations locally to estimate the location of abnormal targets in the power transmission channel.

[0009] By using the ratio of the differences between the images of the abnormal target captured by two cameras, and given the known distance between the two transmission towers, the location of the abnormal target can be determined by the unique correspondence between the ratio and the distance to the camera.

[0010] Furthermore, the geometric midpoint of the abnormal target recognition box is used to represent the position of the abnormal target. The changes in the position of the abnormal target between two frames are collected by the two cameras respectively. The displacement ratio of the pixel displacement is used to estimate the displacement of the abnormal target in the direction of the power transmission channel.

[0011] Furthermore, the two cameras are positioned at the geometric center of all line mounting points, ensuring that the lowest point of the line sag appears at the near end of the field of view at one-third of the distance.

[0012] Among them, the camera and edge intelligent computing terminal are standard equipment configurations for existing power transmission network monitoring. The improvement of the device in this invention is that the camera's setting position is readjusted to construct a facing cylindrical field of view, so that abnormal targets can be located using a simplified method without further adding or replacing equipment.

[0013] Regarding the identification method, since target type identification is already a standard configuration for existing power transmission line monitoring, this solution transforms the problem of identifying abnormal targets into pixel displacement of points in the detection frame; and cleverly utilizes the symmetry by taking advantage of the differences between the images acquired by the two cameras, thereby achieving relatively accurate identification. Moreover, it only involves simple table lookup or linear calculation, which will not increase the burden on the edge intelligent computing terminal, thus achieving localized processing.

[0014] Furthermore, establish a cylindrical coordinate system with the line connecting the two cameras as the z-axis, and the origin of the z-axis at one of the cameras; let the relative position coordinates of the abnormal target be ( ),in It is the radial distance in cylindrical coordinates. It's the azimuth. It is the coordinate on the z-axis;

[0015] When an abnormal target enters the field of view of one of the cameras, the edge intelligent computing terminal first identifies it and waits for the same target to appear in the field of view of the other camera. Once it is confirmed that the target has appeared in both opposing fields of view, the edge intelligent computing terminal takes the current frame and the next frame of the dual cameras to analyze the movement of the abnormal target.

[0016] This approach considers constructing a "visual fence" using a cylindrical coordinate system. Only abnormal targets entering the fenced area are considered to pose a potential threat to the power transmission line, triggering further monitoring and thus saving computing power.

[0017] Furthermore, by pre-establishing the pixel displacement ratio and the cross-sectional position of the cylindrical coordinate system in the direction of the power transmission channel... A one-to-one correspondence is achieved by using the pixel displacement ratio between the current frame and the next frame. The estimation is then performed; further, the intrinsic parameter matrix of the camera is used to realize the relationship between pixels in the image and... By establishing a one-to-one correspondence between the coordinate points on the plane, the spatial coordinates of the abnormal target can be obtained. ).

[0018] Furthermore, when an abnormal target continues to appear in the field of view of the two cameras, the displacement of the abnormal target in the direction of the power transmission channel is estimated by utilizing the change in the position of the center point of the two sets of images between the two frames and the pixel displacement ratio.

[0019] Furthermore, when estimating the location of multiple abnormal targets:

[0020] For multiple different abnormal targets, the dual cameras identify and pair the targets by recognizing their names.

[0021] For multiple identical anomalous targets, based on multi-frame images from dual cameras, the matching of the same anomalous target bounding boxes is performed by utilizing the property that the inter-frame displacement ratios of the bounding boxes belonging to the same anomalous target are the same.

[0022] Furthermore, if the anomalous target stops moving after entering both sides of the field of view, no ratio calculation or update is performed; only the previous calculation results are used. If the anomalous target moves out of one side of the field of view, position estimation is stopped. If the anomalous target only appears on one side of the field of view, no position estimation is performed.

[0023] This is because the different angles and shapes of abnormal targets may be uncertain and unknown. Only by using the movement of the center of the recognition box in this solution to calculate the proportion can a relatively high accuracy be achieved. Furthermore, targets that have temporarily stopped moving can generally be considered not to pose further risks (an alarm can be triggered once when they enter the dual field of view), so there is no need to invest computing resources for the time being.

[0024] Furthermore, the frame rate of the images acquired during the abnormal target location estimation calculation is adjusted to match the hardware processing capabilities and / or ensure that obvious displacement of the abnormal target is captured.

[0025] This adjustability is designed to match existing target acquisition and identification schemes and to save computing resources as much as possible.

[0026] And, a corresponding power transmission channel abnormal target location estimation device, including: two identical cameras and an edge intelligent computing terminal connected in a network;

[0027] Two cameras are positioned facing each other on two adjacent power transmission towers. By using the ratio of the differences between the images of the abnormal target captured by the two cameras, and given the known distance between the two power transmission towers, the location of the abnormal target can be determined by the unique correspondence between the ratio and the distance to the camera.

[0028] Compared to existing technologies, this invention and its preferred solution enable rapid estimation of the spatial location of abnormal targets in power transmission channels, supporting the detection, alarm, and even early warning of threatening abnormal targets. Built upon existing target recognition algorithms and hardware systems, it requires only minor software and hardware upgrades to re-energize existing equipment; furthermore, it is lightweight and will not affect the frame rate of video information processing by the edge intelligent gateway. Attached Figure Description

[0029] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0030] Figure 1 This is a schematic diagram of the dual-camera setup locations according to an embodiment of the present invention;

[0031] Figure 2 This is a schematic diagram illustrating the position estimation principle of an embodiment of the present invention; wherein, a) is a spatial view; b) is a top view. Detailed Implementation

[0032] In the following, specific embodiments of this application will be described in detail with reference to the accompanying drawings. Based on these detailed descriptions, those skilled in the art will be able to clearly understand and implement this application. Without departing from the principles of this application, features from various embodiments can be combined to obtain new implementations, or certain features from some embodiments can be substituted to obtain other preferred implementations.

[0033] To make the features and advantages of this patent more apparent and understandable, specific embodiments are provided below, along with accompanying drawings, for detailed explanation:

[0034] 1. Dual camera settings

[0035] Transmission towers come in various models. This invention uses a group of cat-head shaped transmission towers as an example to illustrate a general principle for setting up dual cameras, specifically as follows: Figure 1 As shown. To ensure that the parallax information from the cameras can be effectively correlated with the spatial location of abnormal targets, it is necessary to ensure that the two cameras are positioned directly opposite each other. Simultaneously, the general requirements for camera placement in power transmission channel monitoring should be met, namely, the lowest point of the line sag should appear at the near one-third of the field of view. Where conditions permit, the cameras should also be positioned at the geometric center of the plane of all line mounting points, such as... Figure 1 The green dots indicate this. This approach provides more comprehensive security for the line, enabling timely detection and early warning of abnormal targets approaching from all directions. The dual cameras will form a network with the same edge intelligent computing terminal. At this time, the intelligent terminal can utilize the algorithm of this invention and its own computing power to perform correlation analysis on the visual information (video) transmitted back by the dual cameras.

[0036] 2. Position Estimation Principles and Algorithms

[0037] To effectively assess the spatial location of abnormal targets and determine their threat level to the power transmission channel, a cylindrical coordinate system is first established. The z-axis is the line connecting the two opposing cameras, with its origin at the camera on tower n. Subsequent algorithms will use this coordinate system to determine the relative position coordinates of the abnormal targets. The calculation of ), where It is the radial distance. It's the azimuth. It is the coordinate on the z-axis, such as Figure 2 As shown.

[0038] When an unusual target enters the field of view of one of the cameras, the edge intelligent computing terminal will first complete the identification, while waiting for the same target to appear in the field of view of the other camera. Once it is confirmed that the target has appeared in both opposing fields of view, the edge intelligent terminal will take one frame and the next frame from both cameras to analyze the movement of the unusual target.

[0039] The specific analysis methods and algorithmic principles involved are as follows:

[0040] a) Take the geometric midpoint of the abnormal target identification box to represent the position of the abnormal target. This can effectively eliminate the influence of the displacement of the abnormal target in the z-axis direction on the accuracy of spatial position estimation.

[0041] b) Calculate the displacement of the target in the two sides of the field of view by using the changes in the center point position of the two sets of images between the two frames. The displacement distance is measured in pixels, i.e., pixel displacement. It should be noted that if the original video processing frame rate is high, it is necessary to consider selecting images every other frame for analysis so that the abnormal target has enough time to complete a significant displacement, thereby ensuring the accuracy of the estimation algorithm.

[0042] c) In most cases, due to the different distances of the abnormal target from the two cameras, its displacement (pixel displacement) in the two fields of view will differ. However, in reality, the displacement of the abnormal target on the z-axis normal section is consistent. Therefore, this invention establishes a pixel displacement ratio and a cross-sectional position... The one-to-one correspondence can be effectively achieved by utilizing the pixel displacement ratio. The estimation. Specific implementation methods include, but are not limited to, table creation and lookup, linear fitting, etc. The input to the algorithm should be the pixel displacement ratio, and the output should be... Numerical value. Obtained. Then, the intrinsic parameter matrix of the camera can be used to realize the relationship between pixels in the image and... The coordinates of the points on the plane are now in a one-to-one correspondence, meaning the spatial coordinates of the abnormal target have been obtained. ).

[0043] d) By repeating the above steps, dynamic tracking of the location of abnormal targets can be achieved.

[0044] The above is the principle for estimating the spatial location of a single anomalous target.

[0045] To better handle situations where multiple targets enter the field of view simultaneously, the following processing logic can be added to this embodiment:

[0046] 1) Distinguishing between multiple different abnormal targets: When abnormal targets enter the two fields of view one after the other, the names of the targets on both sides will be compared first. After ensuring that all names appear in pairs, the recognition boxes will be grouped based on the names.

[0047] 2) Distinguishing between multiple identical anomalous targets: When a third target with the same name appears in the field of view, the previous estimation results are de-emphasized, and then the system waits for the appearance of a fourth target with the same name. If a fourth target appears, the displacement lengths of the geometric centers of the four target bounding boxes are calculated over the three frames, and the ratio of the displacement of each center point between the first and second frames to the displacement between the second and third frames is calculated, i.e., the inter-frame displacement ratio. At this point, the bounding boxes belonging to the same anomalous target should have the same inter-frame displacement ratio. Therefore, comparing the four calculated ratios can achieve effective matching of the bounding boxes on both sides.

[0048] It should be noted that if an abnormal target stops moving after entering the horizon on both sides, the various ratios will not be calculated or updated, and only the previous calculation results will be used. If the target moves out of the horizon on one side, the position estimation will stop. In addition, if the target only appears on one side of the horizon, the position estimation will not be performed.

[0049] The above description is a preferred embodiment of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

[0050] This patent is not limited to the above-described preferred embodiment. Anyone can derive other forms of methods for estimating the location of abnormal targets in power transmission channels based on the inspiration of this patent. All equivalent changes and modifications made within the scope of this patent application shall fall within the scope of this patent.

Claims

1. A method for estimating the location of abnormal targets in a power transmission channel, characterized in that: Two identical cameras are positioned facing each other on adjacent power transmission towers. The two cameras are networked with the same edge intelligent computing terminal and perform calculations locally to estimate the location of abnormal targets in the power transmission channel. By using the ratio of the differences between the images of the abnormal target captured by two cameras, and given the known distance between the two transmission towers, the location of the abnormal target can be determined by the unique correspondence between the ratio and the distance to the camera. The geometric midpoint of the abnormal target recognition box is used to represent the position of the abnormal target. The change of the abnormal target position between two frames is collected by two cameras respectively. The displacement ratio of the pixel displacement is used to estimate the displacement of the abnormal target in the direction of the power transmission channel. Establish a cylindrical coordinate system with the line connecting the two cameras as the z-axis, and the origin of the z-axis at one of the cameras; let the relative position coordinates of the abnormal target be ( ). ),in It is the radial distance in cylindrical coordinates. It's the azimuth. It is the coordinate on the z-axis; When an abnormal target enters the field of view of one of the cameras, the edge intelligent computing terminal first identifies it and waits for the same target to appear in the field of view of the other camera. After confirming that the target appears in both opposing fields of view, the edge intelligent computing terminal takes the current frame and the next frame of the dual cameras to analyze the movement of the abnormal target. By pre-establishing the pixel displacement ratio and the cross-sectional position of the cylindrical coordinate system along the power transmission channel direction. A one-to-one correspondence is achieved by using the pixel displacement ratio between the current frame and the next frame. The estimation is then performed; further, the intrinsic parameter matrix of the camera is used to realize the relationship between pixels in the image and... By establishing a one-to-one correspondence between the coordinate points on the plane, the spatial coordinates of the abnormal target can be obtained. ); When an abnormal target continues to appear in the field of view of the two cameras, the displacement of the abnormal target in the direction of the power transmission channel is estimated by utilizing the change in the position of the center point of the two sets of images between two frames and the pixel displacement ratio. When performing multi-target anomaly target location estimation: For multiple different abnormal targets, the dual cameras identify and pair the targets by recognizing their names. For multiple identical anomalous targets, based on multi-frame images from dual cameras, the matching of the same anomalous target bounding boxes is performed by utilizing the property that the inter-frame displacement ratios of the bounding boxes belonging to the same anomalous target should be the same. If the anomalous target stops moving after entering the horizon on both sides, no ratio calculation or update will be performed; only the previous calculation results will be used. If the anomalous target moves out of the horizon on one side, position estimation will stop. If the anomalous target only appears on one side of the horizon, no position estimation will be performed.

2. The method for estimating the location of abnormal targets in a power transmission channel according to claim 1, characterized in that: Place the two cameras at the geometric center of all line mounting points, and ensure that the lowest point of the line sag appears at the near one-third of the field of view.

3. The method for estimating the location of abnormal targets in a power transmission channel according to claim 2, characterized in that: By adjusting the frame rate of the images acquired during the abnormal target location estimation calculation, the hardware processing capabilities can be matched and / or the obvious displacement of the abnormal target can be captured.

4. A device for estimating the location of an abnormal target in a power transmission channel, using the method for estimating the location of an abnormal target in a power transmission channel as described in any one of claims 1-3, characterized in that, The network consists of two identical cameras and an edge intelligent computing terminal. Two cameras are positioned facing each other on two adjacent power transmission towers. By using the ratio of the differences between the images of the abnormal target captured by the two cameras, and given the known distance between the two power transmission towers, the location of the abnormal target can be determined by the unique correspondence between the ratio and the distance to the camera.

Images