Research on ant colony algorithm based on motion optimization and adaptive threshold for pantograph slider edge detection

The pantograph slide is a part that directly contacts the catenary. Once the fault occurs, it will affect the flow quality of the locomotive and even cause major safety accidents such as scraping bows. Therefore, the online fault detection of pantograph slides is of great significance in ensuring the safe operation of railways. Accurate edge detection is necessary for carbon slide wear detection. Image-based fault diagnosis is an efficient method. Edge detection is a hot topic in image processing with many methods proposed. Ant colony algorithms show promise for image edge detection. They are efficient at identifying edges in images. We propose an edge detection algorithm that combines edge detection technology with an adaptive ant colony optimization algorithm. First, the ant colony layout is optimized based on the gradient. The gradient value is used to determine the position of the ants. The ants are not randomly distributed but are placed on the highest gradient. Then, heuristic information is used in the construction stage. In the improvement stage, the movement probability of the ant colony is changed to optimize the movement path and obtain the optimal solution. Finally, the threshold adaptive algorithm is applied, which is expected to be used for better and faster path discovery optimization, and then the edge detection image is obtained according to the optimal solution. By comparing the traditional ant colony algorithm with other operators and edge detection algorithms, the proposed method has the greatest advantage over other methods in quantitative indicators.Compared with traditional ACO algorithms, we have improved in PSNR, MSE, SSIM, and FSIM indexes, with a minimum improvement of 2.45% and a maximum improvement of 94.2%. The results show that the gradient-based position can improve the accuracy of edge detection, and the edge detection results are optimized by applying the adaptive ant colony algorithm.