Abstract: Obstacle detection is a crucial technology in environment perception, which directly affects the safety and reliability of intelligent agricultural machinery equipment operation. People working in the fields, as nonstructural and nonfixed obstacles, are important factors affecting the safety of agricultural machinery. This paper reviews the application of pedestrian detection sensor technology and pedestrian detection algorithms in China and internationally, focusing on the applicability of various detection technologies and algorithms in complex farmland environments. Firstly, the advantages, disadvantages, and applications ranging from singlesensor detection technology to multisensor fusion technology are summarized. Secondly, the research and applications of pedestrian detection algorithms using traditional methods and deep learning methods to deal with occlusion are analyzed. Finally, it identifies limitations of current technologies, such as relatively limited application of single sensor technology, insufficient stability of multisensor fusion technology, and limited effect of pedestrian detection algorithm in handling occlusion. This paper proposes prospects for the diversification of collected information, UAVassisted obstacle avoidance, and the establishment of a complete farmland environment sensing system.

Key words: pedestrian detection, obstacle detection, sensors, machine vision, farmland environment

摘要： 障碍物检测是环境感知中的关键技术，直接影响到智能农机作业的安全性和可靠性，其中田间行人作为非结构、非固定的障碍物，是影响农机行驶安全的重要因素。遮挡情况是行人检测中的研究重点和难点，也是为满足农机自主作业的避障需求必须考虑的问题。因此对行人检测传感器技术和行人检测算法的国内外应用情况进行综述，重点关注各种检测技术和算法在复杂农田环境中的适用性。首先，梳理总结从单一传感器检测技术到多传感器融合技术的优缺点及其应用情况。其次，分别对利用传统方法和深度学习方法处理优化遮挡情况的行人检测算法的研究应用进行分析。最后，提出现有技术存在单一传感器技术的应用相对局限、多传感器融合技术的稳定性不足、行人检测算法对遮挡问题的处理效果有限等问题，同时对采集信息的多样化、无人机辅助避障、建立完整农田环境感知系统进行展望。

关键词: 行人检测, 障碍物检测, 传感器, 机器视觉, 农田环境