An approach for rapid, accurate and automatic 3D reconstruction of maize ear based on computer vision was presented. Firstly, we rotate the maize ear in a proper angle interval to acquire images in different views, and then calculate out points cloud of maize ear surface with binocular stereovision. Secondly, we eliminate outliers according to the threshold of reprojection error, find out 2D matching points in two neighboring images, determine the 3D matching points set of points clouds of maize ear surfaces by the 2D matching points, calculate the rotation matrix and translation vector of the matching points between two neighboring views, and test the consistency of the 3D registration model by RANSAC method. Finally, by rotating and translating each point cloud of different views to stitch the whole maize ear surface, eliminating the redundant points, simplifying the mesh, and mapping the texture, we get the final 3D shape of maize ear. The experiment results show that the volume of the 3D reconstruction maize ear has no significant difference from the measurement value, and the method proposed can meet the need of 3D reconstruction of maize ear.