Real Time Automatic Detection Algorithm for Loose Railway Track Fasteners under Uneven Lighting Conditions

When the lighting is uneven, the local direction of the light strip in the railway track fastener image output by column traversal will be affected by the curvature value of the light strip, resulting in discontinuity and inability to effectively obtain the loose state of the fastener. Therefore, a real-time automatic detection algorithm for railway track fastener looseness under uneven lighting conditions is proposed. After Gaussian filtering denoising, the neighborhood of the image center point is described using an eight neighborhood chain code. Insert coarse positioning points between the centers of intermittent light strips based on neighborhood grayscale. Determine the direction of the local light band by traversing the output vertically, horizontally, diagonally 45 ° to the left, and diagonally 45 ° to the right. After aggregating the local direction of the light strip to obtain complete contour information, calculate the gap between the clip and the elastic strip, and use it as the final looseness detection result. The experimental results show that under extreme lighting conditions, the AP value of this method for detecting fastener looseness is above 0.92, and the actual detection is stable within 130 ms, which can meet the requirements of practical applications.