An Image Enhancement Processing Algorithm for Robotic Dogs in Concrete Surface Defect Inspection

The robotic dog, equipped with multi-type visual perception equipment, can perform precise detection of surface defects on concrete structures such as bridges and tunnels. However, during the mobile detection process, issues such as variable shooting perspectives, uneven environmental lighting, and self-vibration lead to poor quality in the captured images, which can even cause false and missed detections. To address this problem, an image enhancement processing method tailored for the robotic dog platform is proposed. Firstly, the image color space is converted from BGR to LAB, and CLAHE technology is applied to enhance image contrast, thereby highlighting edge details. Secondly, a combination of Wiener filtering and wavelet transform is used to remove strip noise from the images. On this basis, images with dynamic blur are divided into blocks, and the Particle Swarm Optimization (PSO) algorithm is employed to globally estimate the blur kernel. Subsequently, a local and global weighted algorithm is utilized for image restoration. Finally, the effectiveness of the proposed method is validated using actual image data. The results demonstrate that the proposed method accentuates the texture and edge features of defects, thereby improving the detection accuracy of surface defects in concrete structures during robotic dog inspections.