To solve the difficulty of conveyor belt edge detection, an algorithm for detecting conveyor belt edge based improved pulse coupled neural network (PCNN) model and vector method is proposed. There are many structure parameters in PCNN model, and it is difficult to get the adaptive parameters. The model structure was improved, and thus the number of the parameters was reduced. On the basis of this, using vector method to search for conveyor belt edge. Experiments show that the algorithm is reliable and valid.