A Lightweight Rust Detection Method of Power Equipment Based on DA-MobileNet
Abstract
As a very important part of power system fault detection, corrosion detection of power equipment needs to be quickly and accurately identified. In view of the low efficiency of manual inspection of power equipment, a lightweight corrosion detection method based on DA-MobileNet is proposed in this paper. Firstly, the algorithm model is greatly compressed by MobileNet optimization method based on dense connection. Secondly, an up-sampling feature fusion strategy based on the dual attention model is proposed to compensate the loss of precision caused by the reduced model structure. Finally, based on the standard SSD and the deep separable convolution of MobileNet, an improved lightweight neural network model is constructed by combining dual attention mechanism. The experimental results show that compared with the standard SSD, the proposed algorithm can reduce the number of parameters by 63.6%, and improve the accuracy by 10.47% and average accuracy by 5.99% besides a speed increase of 46.7%, which can meet the requirements of rapid and accurate identification of rust detection of power equipment.