Using a vision system to locate a target is a necessary step for its three-dimensional detection of the target. The traditional single-aperture imaging system can only obtain the geometric image information of the target. A compound eye vision system has the advantages of large field of view, large depth of field, multi-channel imaging, and can obtain the depth information of the target and be sensitive to fast moving targets. At present, a common visual positioning method is to use the binocular vision system to locate the target based on the parallax between two cameras. However, because the binocular vision system has only one set of constraints, and the baseline is fixed, the binocular vision system has low positioning accuracy in the long distance, while the compound vision system has more constraints because of the number of sub-eyes. In the long distance, the positioning accuracy is higher than the binocular vision system. It has aroused a wide attention of researchers. This paper uses the bionic curved compound eye camera developed in the laboratory to carry out the research of 3D positioning and 3D reconstruction. The compound eye vision system consists of a curved compound eye, an optical relay image conversion subsystem and a high-definition image sensor. In this paper, CAL Tag calibration board and MATLAB stereo calibration toolbox is used to calibrate the internal parameter matrix of the compound eye camera and the rotation matrix and translation vector between the sub-eye and the world coordinate system. Based on the principle of binocular vision positioning, a mathematical model for multi eye positioning is established on a compound eye vision system developed in the laboratory, and positioning experiments are conducted. The experimental system includes a laser rangefinder, black cardboard, and a compound eye vision system. The laser spot is used as a positioning target. Because the shape of the sub-eye is circular, the hough circle transformation algorithm is used to detect the sub-eye of the compound eye system, and the sub-eye number is determined according to the center coordinates and radius of the circle. Because this experiment is carried out under dark conditions, the background gray value is low and the spot gray value is high, so the gray centroid method is used to locate the centroid of the spot and obtain the centroid of the spot taken by different sub-eyes. The three-dimensional coordinates of the centroid of the spot are obtained from the coordinates of the centroid of the spot in the camera pixel coordinate system according to the corresponding relationship between the pixel coordinate system and the world coordinate system. The linear equations of several sub-eyes are combined to form the overdetermined equations and the optimal solution is obtained by the least square method. The distance measurement experiment results show that the distance measurement error of the compound eye camera is less than 2% within a range of at least 4 meters. The experimental results show that the bionic curved compound eye camera prepared in the laboratory could carry out more accurate three-dimensional positioning of objects in space. The error caused by the laser jitter and the size change of the light spot with the distance change on the positioning result is analyzed in detail. In the aspect of target 3D reconstruction, the sift algorithm is used to detect and match the feature points of the target images of different sub-eyes, and the RANSAC algorithm is used to remove the wrong matching points, to obtain the accurate feature point matching of the target captured by different sub-eyes. Then, according to the corresponding relationship between the pixel coordinate system obtained by camera calibration and the world coordinate system, the three-dimensional coordinates of the feature point in the world coordinate system are calculated from the coordinates of the sub-eye pixel coordinate system, and the complete reconstructed point cloud of the target is obtained through point cloud stitching. The 3D reconstruction experiment is carried out by the reconstruction algorithm. The experiment takes the cube covered with speckles as the reconstruction target. The cube is photographed at about 0.6 meters from the camera, and a relatively complete 3D reconstruction point cloud is obtained. The research results in this paper show that the bionic curved compound eye camera has great development potential and application prospects in the fields of 3D positioning, 3D reconstruction and optical navigation.