Blur Edge Detection using 2-D Entropy for Pantograph Slide Plate Wear Estimation

The thickness monitoring of pantograph slide plate is particularly important for the safe and stable operation of electric locomotive. The results of the existing edge detection algorithms detecting wear edge is incomplete in most cases. This paper proposed an improved Canny operator based on longitudinal two-dimensional entropy, which adjust the threshold of Canny operator to adapt different brightness of wear curve. Longitudinal 2D entropy is used to evaluate the edge information in the sliding window which reflects the brightness of the wear edge in the sliding window. The thresholds of Canny operator are determined by the brightness of the edge evaluated by longitudinal 2D entropy. The detection result of improved Canny operator is compared with the other four edge detection methods. The experimental and evaluation results show that, when the time consumption of the algorithm is allowed, the longitudinal two-dimensional entropy and Canny algorithm can achieve a much better result.