In order to solve the problem that the branches and leaves were obscured in the visual recognition of tomato fruit size and posture during the grasping process of traditional agricultural robots, a method of tomato size and posture analysis based on visual and tactile perception was proposed. In the process of fruit grasping, the local point cloud information of fruit contour contact was obtained by visual and tactile sensors, and then the point cloud information under different sensor coordinate systems was transformed to the same base coordinate system by camera parameter calibration and finger joint transformation matrix, and then the size and posture information of fruit was analyzed by point cloud improved PCA algorithm and ICP algorithm. In order to evaluate the performance of the proposed analytical method, tomato size and posture tests were performed in the laboratory environment. The real values of tomato fruit size and posture were obtained by vernier caliper measurement and depth camera scanning, and compared with the presented analytical results. The test results showed that the average errors of transverse and longitudinal dimensions of tomato obtained by this method were 8.66% and 11.08%, and the average errors of horizontal angle and vertical deflection angle of tomato fruit axis and projection plane were 10.03% and 14.02%. The size and posture information of tomato fruit analyzed by the proposed method can be applied to the posture regulation of tomato fruit grasping process, so as to improve the reliability of tomato fruit grasping and picking.