Straight line detection via rival penalized competitive learning

In this paper, rival penalized competitive learning (RPCL) algorithms are adopted to solve the straight line detection problem. In order to improve the performance of the distance sensitive RPCL (DSRPCL) algorithm on straight line detection, the cost function is modified. It is demonstrated by the experiments that the improved DSRPCL based algorithm can automatically detect the correct number of actual straight lines in an image, with a better line location in comparison with the frequency sensitive competitive learning (FSCL) as well as the original RPCL based algorithms.