Monocular detection and positioning method for power transmission line inspection robot

Abstract

The invention provides a power transmission line inspection robot monocular detection and positioning method, which comprises the following steps: (1) collecting color images and depth images of different types of line obstacles, marking an obstacle area in the color image as a target area, and constructing a data set 1; (2) preprocessing the first data set, and sending the preprocessed first dataset to a YOLOV3 network for training; (3) obtaining a target obstacle area according to an output result of the YOLOV3 network, finding the same area in the corresponding depth image to obtain depthinformation, and obtaining an output tensor through the YOLOV3 network and the corresponding depth information to form a data set II; (4) sending the second data set into a linear regression model fortraining; and (5) using the trained YOLOV3 network and a linear regression model to carry out obstacle identification and positioning test. Compared with a traditional method, a deep learning and machine learning method is utilized, the accuracy rate and the real-time performance are improved, the cost is greatly reduced, and a better detection effect and better detection performance are obtained.

Description

technical field [0001] The invention belongs to the technical field of pattern recognition and artificial intelligence, and relates to a monocular detection and positioning method of a power transmission line inspection robot. Background technique [0002] The intelligent high-voltage line inspection robot is a research hotspot, because a necessary function of the intelligent high-voltage line inspection robot is to be able to climb over various obstacles, and what kind of obstacle-crossing operation to perform on the possible obstacles ahead is a very important The premise is to know the specific type and location of the obstacle. For the problem of obstacle detection and positioning on transmission lines, accuracy and real-time performance are the two most important indicators, and the two indicators should be considered comprehensively in practical applications. [0003] At present, there are two basic methods for obstacle detection and recognition: obstacle recognition ...

AI Technical Summary

Problems solved by technology

However, due to the complex operating environment of the line, it is easily affected by the natural environment such as light, which makes it difficult for traditional obstacle detection and recognition methods to meet the requirements of accuracy and real-time performance. The positioning function of obstacles is not perfect, and they rely more on powerful functions The sensor makes the production cost higher

Images