Evolutionary Optimization of Stereo Matching Parameters for Plant Shape Estimation

Plant growth is one of the most important components in production environment control. In recent years, the development of stereo imaging techniques and, hardware such as stereo camera allows for precise measurement of the 3D structure of an object. However, in order to monitor plant growth at multiple points in space and time, it is necessary to construct an affordable system from a practical viewpoint in agriculture. Therefore, in this study, we focus on parameter optimization in parallax estimation to achieve highly accurate plant height measurement even with inexpensive hardware. In particular, this study applies an optimization method, namely a Genetic algorithm, to several parameters of the basic algorithm for disparity estimation considering objective functions. In this study, we set our objective functions such as Root Mean Squared Error (RMSE), Peak Signal to Noise Ratio (PSNR), and Structural SIMilarity (SSIM). The results of this study show that the optimization methods can improve the accuracy of disparity estimation for plant shape.