A New Camera Calibration Based on Neural Network with Tunable Activation Function in Intelligent Space

Abstract

In order to solve the camera calibration in intelligent space of mobile robot, a new calibration method based on neural network with tunable activation function (TAF) is presented. In the TAF model, the inner product mode is adopted in the calculation of output signal in synapse model and the S function is adopted in base function. Taking the coordinate in image coordinate system as the network input, and the coordinate in world coordinate system as the network output, the weight matrix, threshold matrix and activation function parameter are achieved firstly through the network training based on sample data, then the calibration test are carried out using the TAF network which is trained. The experiment results show that, compared with the test results of BP network, the proposed calibration method is characterized by high detection precision and quick detection speed, which verifies the validity of TAF network in camera calibration.