A Novel Intelligent Recognition of Video Objects in Drone Routing Inspection based on Deep Learning
Shuo Zhang, Dandan Wang, Kun Wang, Xinbo Liu, Qin Deng, Zhaoxiang Mi · 2022 7th International Conference on Intelligent Computing and Signal Processing (ICSP) · 2022
In the field of electric power routing inspection, the drone electric power routing inspection method can better break through the limitations of human routing inspection and manned helicopter routing inspection, and can be used for fine routing inspection, fixed-point routing inspection and range routing inspection. The existing drone routing inspection video object recognition has the problem of misjudging the recognition of the same type and the same model of equipment state in different scenes. To address the above problems, this paper proposes a histogram video image feature extraction method based on rotation-invariant HOG (Histogram of Oriented Gradients), and further realizes the accurate recognition of objects in the routing inspection process under deep convolutional neural network. The experiments have shown that the recognition accuracy rate of 14 types of objects such as even pressure ring, anti-vibration hammer, tower end of hanging hammer string, sub conductor spacer bar, damaged part of insulator, insulator string, heavy hammer and attached gold tools commonly found in power routing inspection is not less than 75%, and the recognition accuracy of some objects can reach 97%, which has good practical value.