CN-121977551-A - Intelligent identification and navigation method for full-automatic oil extraction equipment of electric power device

CN121977551ACN 121977551 ACN121977551 ACN 121977551ACN-121977551-A

Abstract

The embodiment of the application provides an intelligent identification and navigation method of full-automatic oil extraction equipment of an electric device, which comprises the following steps of S1, performing global path navigation, controlling equipment to automatically move to an area where a target electric device is located based on preset position coordinates and a navigation map of the electric device, S2, performing image acquisition on the electric device by using a carried visual identification device by using the equipment, identifying the oil extraction valve by using an improved target detection model, and acquiring position information and posture information of the oil extraction valve, and S3, performing automatic oil extraction operation by using the equipment, and controlling a mechanical arm to finish oil extraction operation based on the position information and the posture information. The method realizes the full-flow automation and the intellectualization of the oil extraction operation of the electric power device, and can effectively improve the recognition precision and the pose judgment accuracy of the oil extraction valve under the complex environment, thereby remarkably improving the success rate and the reliability of the oil extraction operation of the mechanical arm and avoiding the potential safety hazard of manual operation.

Inventors

GUO SHAOWEI
YANG CHUNYU
Su Zexin
LIU HUABEI
XU HENGMING

Assignees

苏州壹伏安科技有限公司

Dates

Publication Date: 20260505
Application Date: 20251116

Claims (8)

1. An intelligent identification and navigation method for full-automatic oil extraction equipment of an electric device is characterized by comprising the following steps: s1, global path navigation, wherein the equipment is controlled to autonomously move to an area where a target power device is located based on a preset power device position coordinate and a navigation map; S2, intelligent identification and positioning of an oil extraction valve, wherein the equipment utilizes a carried visual identification device to acquire images of the electric power device, adopts an improved target detection model to identify the oil extraction valve, and acquires position information and attitude information of the oil extraction valve; And S3, automatically taking oil, wherein the equipment controls the mechanical arm to finish the oil taking operation based on the position information and the gesture information.
2. The method of claim 1, wherein the retrofit-objective-detection model is a retrofit YOLOv model, the retrofit measures including embedding an oil-extraction-valve-attention module in a neck module of the YOLOv model and a detection-head-increased-rotation-angle-prediction branch of the YOLOv model.
3. The method of claim 2, wherein the oil extraction valve attention module comprises: The edge contour feature extraction branch comprises a multi-direction gradient convolution layer and a weighted fusion convolution layer which are connected in series, wherein the multi-direction gradient convolution module comprises 4 convolution blocks with fixed parameters, and the weighted fusion convolution layer comprises 1 trainable convolution block and is used for fusing feature images of the multi-direction gradient convolution layer to obtain an edge strengthening feature image; The central feature extraction branch comprises a cavity convolution layer, a channel attention mechanism and a spatial attention mechanism which are connected in series.
4. A method according to claim 3, wherein the rotation angle prediction branch is parallel to the original bounding box regression branch and the category prediction branch, and is used for jointly regressing the deflection angle of the target oil extraction valve.
5. The method according to claim 4, wherein the step S2 comprises: S201, the mechanical arm acquires images of the oil taking valve area of the electric power device by utilizing the visual identification device; S202, inputting the image into the improved target detection model for recognition, and acquiring the position information and the posture information of the oil extraction valve; S203, acquiring actual position coordinates and three-dimensional attitude information of the oil extraction valve based on the position information and the attitude information.
6. The method according to claim 5, wherein the step S3 includes: and if the lost target or the detected deviation of the target position and the expected position exceeds the tolerance range in the tracking process, the mechanical arm pauses the movement and the step S2 is re-executed.
7. The method according to claim 6, wherein the step S1 comprises: controlling the equipment to move based on a navigation map which is constructed in advance through a laser SLAM technology, wherein the navigation map is marked with preset station coordinates of each target power device; the device comprises a laser radar capable of sensing the surrounding environment in real time, and sensing data of the laser radar are matched with the navigation map to plan a path to a target station.
8. The method according to claim 7, wherein step S3 further comprises verifying the operation result by the mechanical arm through the visual recognition device after each automatic oil extraction sub-operation is completed, and executing the subsequent sub-operation only after the verification is passed.

Description

Intelligent identification and navigation method for full-automatic oil extraction equipment of electric power device Technical Field The application relates to the field of operation and maintenance of power equipment, in particular to an intelligent identification and navigation method for full-automatic oil extraction equipment of a power device. Background Along with the continuous improvement of the intelligent operation and maintenance requirements of the power system, the automation of the oil extraction operation of the power device has become an important development direction. At present, the field still generally depends on manual operation or an automation technology with certain limitation, so that the efficiency is low, and the safety of personnel is threatened in a high-voltage and high-risk power equipment environment. The technical scheme of the prior art faces a core bottleneck when full-automatic oil extraction is realized, on one hand, accurate navigation and positioning of equipment are easy to be interfered in a complex transformer substation environment, and on the other hand, for oil extraction valves with exquisite structures and different postures, a conventional vision system is difficult to realize high-precision real-time identification and pose estimation, so that a mechanical arm cannot reliably finish grabbing and butting. These deficiencies severely limit the success rate and universality of the operation of the automatic oil extraction equipment. Therefore, there is an urgent need in the art for a method for intelligent recognition and navigation of a fully automatic oil extraction device of an electric device to overcome the above problems. Disclosure of Invention The application provides an intelligent recognition and navigation method for full-automatic oil extraction equipment of an electric power device, which realizes the full-flow automation and the intellectualization of the oil extraction operation of the electric power device through intelligent navigation and accurate positioning based on an improved visual recognition algorithm, and can effectively improve the recognition precision and the pose judgment accuracy of an oil extraction valve under a complex environment, thereby remarkably improving the success rate and the reliability of the oil extraction operation of a mechanical arm, avoiding the potential safety hazard of manual operation, promoting the unmanned process of the electric power operation and maintenance operation, and having important significance for guaranteeing the safety of personnel and improving the operation efficiency. In a first aspect, a method for intelligent identification and navigation of full-automatic oil extraction equipment of an electric device is provided, the method comprising: s1, global path navigation, wherein the equipment is controlled to autonomously move to an area where a target power device is located based on a preset power device position coordinate and a navigation map; S2, intelligent identification and positioning of an oil extraction valve, wherein the equipment utilizes a carried visual identification device to acquire images of the electric power device, adopts an improved target detection model to identify the oil extraction valve, and acquires position information and attitude information of the oil extraction valve; And S3, automatically taking oil, wherein the equipment controls the mechanical arm to finish the oil taking operation based on the position information and the gesture information. It is understood that through global path navigation and intelligent recognition and positioning of the oil extraction valve, the scheme can enable equipment to accurately reach a target station under the condition of no human intervention, and the accurate position and the gesture of the oil extraction valve can be reliably obtained by utilizing a visual recognition technology, so that the mechanical arm is finally guided to successfully complete oil extraction operation, and the operation efficiency, the operation accuracy and the overall automation level are effectively improved. With reference to the first aspect, in certain implementations of the first aspect, the improvement target detection model is a improvement YOLOv model, and the improvement measure includes embedding an oil extraction valve attention module in a neck module of the YOLOv model and a detection head of the YOLOv model to increase a rotation angle prediction branch. It should be understood that the position information refers to two-dimensional pixel coordinates of the oil extraction valve and corresponding three-dimensional space coordinates thereof, including the position and depth information of the center point of the bounding box, which are identified by improving YOLOv model, while the gesture information refers to the deflection angle of the oil extraction valve around the center point thereof, which is regressed by the rotation angle