Skip to Main Content
In order to describe the camera internal geometrical and optical parameters, and the relation between camera and object in three-dimensional space, traditional camera calibration methods usually need to assume lots of priori knowledge, establish precise mathematical model and then decompose and calculate those parameters we need. While artificial neural network has outstanding non-linearity mapping performance, can avoid these process. It only needs to provide some data of input and output for the net, we can imply various parameters into connection weights of the network by certain iterative training. In this paper, a new approach based on Harris corner extraction algorithm and artificial neural network for camera calibration of monocular vision system is presented. Since angles of between the camera and the calibration template are introduced in the calibration process, it only needs a camera to calibration, can achieve the effect similar with the binocular vision system. The experiment results indicated that the method for calibrating is feasible and effective.