Underwater Thruster Stripe Detection based on Principal Component Analysis and Deep learning

This article proposed a stripe detection method of underwater thruster image based on principal component analysis (PCA) and deep learning. Firstly, the principal component analysis method is used to denoise underwater thruster image, and the denoised image is taken as the input of the trimmed convolutional neural network similar to VGGNet. Then, the output value is obtained after network training by using the depth learning method, and the stripe are obtained. The experimental results have shown that this method can effectively extract underwater thruster stripe features, and the stripe detection results are clear and coherent.This method can be used to detect faults in underwater thruster.