The small agricultural vehicles can effectively solve the problem of pest control. However, in the middle and later stages of corn growing period, leaves crisscross between interrows would severely obstruct the passable region, which would lead great trouble for crop protection vehicles to pass between interrows. A passable region extraction method was proposed for crop protection vehicles，which used a 16line LiDAR installed on the top of the vehicles as the sensing unit to collect the corn interrows information. The maize leaves were nonrigid obstacles. Touching the leaves as the robot travels did not cause crop damage. Through analyzing the threedimensional point cloud data of corn along vehicle forward direction, and studying the distribution law of ground projection of leaves and trunks, the center point of maize point cloud obtained by K-means clustering estimation was taken as the main regional point. Then, the confidence interval was introduced to remove the estimated outlier clustering points in the corn trunk area, and the analysis accuracy was improved. Finally, the central navigation line of the corn crop row under high occlusion environment was analyzed. The experiment was carried out by simulating the real corn field scene. Compared with the trunk position of actual simulated corn, the maximum apparent distance error of the maize position identified by this method was 3.55cm along both sides of the crop line. The average time of the current system perception response was 2s, which satisfied the local positioning requirements of the 60cm autonomous crop protection vehicles.