Pantograph is a key equipment of the electrical system in rail transit vehicles, which is easy to wear during operation. The monitoring of pantograph slider wear is beneficial to the maintenance of electrical system, which is of great significance to the safety and stability of rail transit. In this paper, an image-based strategy for extracting the wear edge of the pantograph slider is proposed. A bi-directional cascade network structure is utilized to detect the edge of pantograph slider. Reciprocal Dice coefficient loss is introduced to enhance the edge detection results. With the design of a post-processing module, the complete framework for extracting the wear edge of pantograph slider is established. This method is validated on the pantograph image data from Beijing Metro Line 6. Experimental results show that the proposed method has better performance than the original bi-directional cascade network with the number of model parameters significantly reduced. ODS, OIS and IMP increased by 1.2%, 1.3% and 2.54%, respectively with the detection speed improved greatly, and the model efficiency is approaching that of a lightweight network.