差分遗传算法优化BP神经网络的双目相机标定

In this paper, a method using a differential genetic algorithm to optimize backpropagation (BP) neural networks and complete binocular camera calibration is proposed. This method solves the problems of large computation and complex calibration process in traditional binocular camera calibration. The energy growth and ray scanning matching algorithms are used to detect and match the corner points with the same name and extract pixel coordinates. The selection and crossover operator of the genetic algorithm are improved, and the mutation operator of the genetic algorithm is improved by the differential genetic algorithm. The differential genetic algorithm is used to optimize BP neural networks for binocular camera calibration. The experimental results show that the root mean square error of the calibration of the binocular camera based on the proposed method is 0.038 mm. The traditional calibration methods based on OpenCV and Matlab are 0.155 mm and 0.417 mm respectively. By comparison, the calibration accuracy is improved by 75% and 90%, respectively. At the same time, the average time needed to calibrate the binocular camera using the method proposed in this paper is 26.3 s. Therefore, the BP network optimized by differential genetic algorithm simplifies the calibration process of binocular camera and achieves good results, which meets the requirements of binocular camera calibration.