An image enhancement method for cable tunnel inspection robot

Influenced by multiple factors such as low light intensity, dispersed light, excessive dust, and excessive ambient noise inside cable tunnels, the images captured by cable tunnel inspection robots have shortcomings such as low contrast, low pixel values, and high noise. To improve the image quality, this paper proposes an image enhancement method suitable for cable tunnel inspection robot. First, in this paper, a bivariate hybrid optimization module using the alternating direction multiplier method and the adaptive learning rate acceleration SVRG algorithm is constructed to achieve image pre-processing. Second, a feature extraction module combining a U-Net network and a coordinate attention mechanism is constructed to extract features from the original image and the pre-processed image. Third, the progressive feature fusion module and the image recovery module are constructed to fuse the above features and are combined with the original image to obtain the enhanced image. Finally, the pixel compensation module is constructed to compensate for the features of the enhanced image and recover the lost texture details.