Algorithm for 3D reconstruction of agriculture field pests based on binocular stereo vision

In order to solve the shortcomings of existing autodiagnostics system of agricultural insect pests and the pest image database, a new algorithm for 3D reconstruction of pests based on binocular stereo vision is proposed in this paper. The whole approach consists of four components. First, a novel image acquisition device based on PS-15 II lamp and bionic cameras is used to collect the pest images, and the images are preprocessed further. Then, the camera calibration and the distortion correction are completed with the cuboid in the pest images. Next, a background difference algorithm is used to extract the pest object from the images. Finally, based on the detection and the matching of feature points in the binocular object images, the 3D reconstruction of the pest data is achieved. Experimental results showed that the algorithm is efficient and accurate, the average error of 3D reconstruction result is less than 0.5mm, and the percentage error is less than 1.5%.