Locating and Tracking Algorithm of Biomimetic Eye Based on Backpropagation Neural Network

In order to realize rapid locating and tracking of a mobile target in the field of view by a biomimetic eye platform, a algorithm to maintain the target on the center of the video frames of two industrial cameras in real-time through joint rotation of each rotational degree of freedom on the biomimetic eye platform is proposed. A binocular biomimetic eye platform with 5 rotational degrees of freedom is designed based on the structural characteristics of human eyes and head. Backpropagation (BP) neural network is adopted to correct the error of binocular 3D measurement based on the dynamically changing external parameters, and the joint space algorithm is utilized to solve the rotation angle of each rotational degree of freedom. The experimental results show that the mean absolute error of the measured value after BP network correction is 2.1849 mm and the average relative error is 0.0051. The method proposed can improve the accuracy of 3D measurement and calculate the joint rotation angle in real time.