An image recognition method for intelligent inspection of power grid equipment

The coverage area of the power system continues to expand, and the grid structure becomes more and more complex, which also brings severe challenges to the inspection work of power personnel. The characteristics of inspection images and factors affecting image quality are summarized. The traditional method has low accuracy and singleness in identifying faults or targets, takes a long time, and is easily affected by factors such as background, form, and materials, and is difficult to apply in practice. In order to solve the above-mentioned shortcomings of traditional methods, deep learning is introduced by using a region-based fully convolutional network, annotated data to train its network, and improved methods such as online difficult sample mining, sample optimization, and soft non-maximum suppression. Experimental results show that the proposed aspect is faster and more accurate in locating targets, and has higher detection accuracy when applied to transmission line inspections to meet the needs of intelligent inspections of transmission lines.