CN-115937720-B - Automatic navigation point generation method based on substation target detection cluster

CN115937720BCN 115937720 BCN115937720 BCN 115937720BCN-115937720-B

Abstract

The invention discloses a method for automatically generating waypoints based on a substation target detection cluster, which particularly relates to the field of unmanned aerial vehicle inspection, and comprises the steps of establishing a substation three-dimensional image, collecting RGB images of equipment to be detected, calculating the geometric centers of all the equipment on a plane by utilizing Graph-Detect3D, obtaining coordinate information of targets to be inspected of all the equipment according to equipment types, orientations and the geometric centers of the equipment, numbering the targets to be inspected of all the equipment, clustering the geometric centers of the equipment by adopting a density clustering algorithm DBSCAN, and taking the geometric centers of all the equipment centers in the same cluster as waypoints, wherein the waypoints are generated according to the geometric center coordinates of each class, so that the quantity of the waypoints cannot be redundant, can exactly cover each equipment, and the pictures of all the equipment of the substation can be collected by the least waypoints, thereby reducing the redundancy of collected data.

Inventors

WANG HAILEI
WANG CHANGJUN
HE ZIGUO
LI HUI
LI DONGDONG
LI HU

Assignees

安徽优航海图科技有限公司

Dates

Publication Date: 20260508
Application Date: 20230103

Claims (2)

1. The automatic generating method of the waypoints based on the substation target detection cluster is characterized by comprising the following steps of: (1) Shooting a bird's eye view of a transformer substation by using an unmanned aerial vehicle, establishing a three-dimensional map of the transformer substation, and marking position information of each device; (2) Setting parameters of a camera focal length, a shooting distance and a cradle head angle, and shooting all equipment to be detected by using the unmanned aerial vehicle; (3) Taking the RGB image acquired in the step (2) as the input of a Graph-Detect3D (three-dimensional) of a 3D target detection depth network based on a Graph, and obtaining the type of equipment in a transformer substation, a 3D boundary frame of a target to be patrolled and examined on the equipment and the orientation of the 3D boundary frame; (4) Calculating the geometric center of each device on the plane according to the coordinate information of the vertex of the 3D boundary frame of the device in the target detection aerial view output; (5) According to the type, the direction and the geometric center of the equipment, coordinate information of targets to be inspected of all the equipment is obtained, and the targets to be inspected of all the equipment are numbered; (6) Clustering the geometric center of the equipment by adopting a density clustering algorithm DBSCAN, and taking the shooting distance of the unmanned aerial vehicle as a radius to obtain a plurality of clustering clusters; (7) Connecting all equipment centers in the same cluster to obtain a polygon, and calculating the geometric center of the polygon as a waypoint; (8) Judging whether targets to be detected on the equipment can be detected by navigation points in a cluster where the targets are located, and marking all numbers of targets to be patrolled and examined which are not in a detection range; (9) Defining a circular range by taking the maximum detection distance of the unmanned aerial vehicle as a radius and taking the position of the center of the 3D surrounding frame of the equipment as a circle center, and recording all waypoint numbers in the circular range as alternative waypoints of the equipment; (10) Judging which alternative waypoints can be detected by the undetected target to be patrolled on the equipment; (11) Binding the equipment with all the screened waypoints one to many, and storing the equipment as a binding file; (12) Checking whether each device has missing target points to be detected, if the missing target points have one place, directly adding waypoints, if the missing target points have two places, setting new waypoints at the connecting center of the two target points, and if the missing target points have multiple places, setting one or more new waypoints according to the relative distance so as to ensure that no waypoints are missing; (13) Updating and storing the binding relation between the equipment and the waypoints.
2. The automatic generating method of waypoints based on the substation target detection cluster according to claim 1, wherein in the step (6), a density clustering algorithm DBSCAN is adopted to cluster the geometric center of the equipment, and the method is specifically as follows: The DBSCAN algorithm firstly selects one core object in the transformer substation, namely equipment in the transformer substation as a seed, creates a cluster and finds out all the core objects, and finds out equipment with reachable density of the combined core objects until all the core objects are accessed, wherein the cluster of the DBSCAN at least comprises one core object, if only one core object exists, other non-core objects fall in epsilon-neighborhood of the core object, if a plurality of core objects exist, at least one other core object exists in epsilon-neighborhood of any one core object, otherwise, the two core objects cannot be reachable in density.

Description

Automatic navigation point generation method based on substation target detection cluster Technical field: the invention relates to the field of unmanned aerial vehicle inspection, in particular to an automatic navigation point generation method based on a substation target detection cluster. The background technology is as follows: The unmanned aerial vehicle inspection is used as an important mode of substation inspection, equipment running conditions and surrounding environment changes can be mastered, equipment defects and potential safety hazards can be found in time, accidents are prevented, and accordingly safety of power transmission and transformation equipment and stability of a power system are guaranteed. However, before inspection, the unmanned aerial vehicle needs to collect waypoints and route planning, the collection of the waypoints of the existing transformer substation is mainly based on field exploration and manual drawing, and the method has subjectivity, generally causes incomplete coverage of the transformer substation area, redundancy of the waypoints and very consumes manpower and time. Based on the method, the invention aims to provide the automatic generation method of the waypoints based on the substation target detection cluster, and the problems of insufficient coverage of the waypoints to the substation area and redundancy of the waypoints are solved. The invention comprises the following steps: In order to overcome the defects of the prior art, the invention aims to provide the automatic navigation point generation method based on the substation target detection cluster, which is used for automatically generating the patrol navigation points through the equipment cluster, so that the efficiency of navigation point generation is high, the number of navigation points cannot be redundant, the coordinate information of targets to be patrol of all equipment is obtained according to the equipment type, the orientation and the geometric center of the equipment, the targets to be patrol of all equipment are numbered, and the problem that the targets to be patrol are not covered can be prevented. The technical scheme of the invention is as follows: a method for automatically generating waypoints based on a substation target detection cluster comprises the following steps: (1) Shooting a bird's eye view of a transformer substation by using an unmanned aerial vehicle, establishing a three-dimensional map of the transformer substation, and marking position information of each device; (2) Setting parameters of a camera focal length, a shooting distance and a cradle head angle, and shooting all equipment to be detected by using the unmanned aerial vehicle; (3) Taking the RGB image acquired in the step (2) as the input of a Graph-Detect3D (three-dimensional) of a 3D target detection depth network based on a Graph, and obtaining the type of equipment in a transformer substation, a 3D boundary frame of a target to be patrolled and examined on the equipment and the orientation of the 3D boundary frame; (4) Calculating the geometric center of each device on the plane according to the coordinate information of the vertex of the 3D boundary frame of the device in the target detection aerial view image output; (5) According to the type, the direction and the geometric center of the equipment, coordinate information of targets to be inspected of all the equipment is obtained, and the targets to be inspected of all the equipment are numbered; (6) Clustering the geometric center of the equipment by adopting a density clustering algorithm DBSCAN, and taking the shooting distance of the unmanned aerial vehicle as a radius to obtain a plurality of clustering clusters; (7) Connecting all equipment centers in the same cluster to obtain a polygon, and calculating the geometric center of the polygon as a waypoint; (8) Judging whether targets to be detected on the equipment can be detected by navigation points in a cluster where the targets are located, and marking all numbers of targets to be patrolled and examined which are not in a detection range; (9) Defining a circular range by taking the maximum detection distance of the unmanned aerial vehicle as a radius and taking the position of the center of the 3D surrounding frame of the equipment as a circle center, and recording all waypoint numbers in the circular range as alternative waypoints of the equipment; (10) Judging which alternative waypoints can be detected by the undetected target to be patrolled on the equipment; (11) Binding the equipment with all the screened waypoints one to many, and storing the equipment as a binding file; (12) Checking whether each device has missing target points to be detected, if the missing target points have one position, directly adding the waypoints, if the missing target points have two positions, setting new waypoints at the connecting center of the two target points, and if the missing target points have multiple