Artificial intelligence system and method serving power robot

Abstract

The invention discloses an artificial intelligence system and method serving a power robot, and the system comprises an embedded CPU module which is configured to send a data scheduling instruction to a data processing module; the data processing module which is configured to perform intelligent processing on the received data based on the received data scheduling instruction, wherein the data processing module comprises an environment perception unit which is configured to realize intelligent construction and obstacle identification of a three-dimensional semantic map of a target area based on multi-sensor fusion information; a path planning unit which is configured to realize path planning of the power robot based on the constructed three-dimensional semantic map; and a data analysis unit which is configured to realize defect identification and position detection of target equipment based on the acquired image information of power equipment. Based on a lightweight deep network model, intelligence of sensing, planning, operation and decision making of the power robot is realized, and the intelligent operation level of the power robot is effectively improved.

Description

technical field [0001] The invention relates to the technical field of electric robots, in particular to an artificial intelligence system and method serving electric robots. Background technique [0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art. [0003] Since the 1970s, research on electric robots has been carried out at home and abroad, and relevant achievements have been made in the fields of power generation, power transmission, power transformation, power distribution, and power consumption, especially in the field of power transformation, substation inspection robots, Indoor track inspection robots have been popularized and applied on a large scale, effectively ensuring the safe and stable operation of power grid equipment. With the continuous deepening of the application of electric robots, problems such as weak intelligence, poor environmental adaptability, and ...

AI Technical Summary

Problems solved by technology

[0005] (1) The existing intelligent systems of different types of electric robots need to be customized and developed, the workload of development and maintenance is large, there are many types of inspection services for electric equipment, various inspection data collection methods, and different inspection data analysis methods, resulting in various electric robots. The phenomenon of scattered, repetitive, and disorderly development of artificial intelligence functions among systems has caused a waste of resources and objectively limits the improvement of the intelligence of the electric robot system

[0006] (2) In terms of inspection data analysis, the existing electric robots have weak front-end video and image processing capabilities, and most of the image and video data are currently analyzed and processed by the back-end server through the network. Delay, unable to meet the needs of application scenarios with high real-time requirements such as robot navigation, visual servo, and timely defect detection

[0007] (3) The robot carries the manipulator to work, and most of the operation servo systems based on binocular vision guide the work of the manipulator. This kind of manipulator operation system is affected by the positioning accuracy and mechanical error of the binocular image equipment, causing the end of the manipulator to stop. An error occurs between the position and the actual working position, which affects the quality of the manipulator's work; the traditional binocular visual servo system does not obtain the semantic information of the equipment in the image, and usually relies on prior information such as equipment type and structure to confirm the type and position of the equipment to be operated

Although the research on human behavior recognition has been going on for a long time, and phased results have been obtained, the current human behavior recognition method has a high recognition accuracy rate on the public database, which can reach more than 90%, but the inventors found that for relatively complex The accuracy of human behavior recognition in electric power scenarios is much lower, even in a laboratory with harsh conditions, it can only reach 70%

[0010] (6) At present, in the field of electric power, the commonly used image semantic analysis technology is image detection. Due to the particularity of the application scene, the inventors found that the traditional semantic analysis method is easily affected by image noise, resulting in uneven distribution of feature points, resulting in classification errors; volume The product neural network method is limited by the small scale of the target and the shooting angle, resulting in a high false detection rate and affecting the practical application

Images