Authors

Abstract

Intelligent agricultural vehicles have been widely used in the process of farming and harvesting in the field, which has brought great convenience to agricultural production. However, there are also safety issues such as accidental collision of agricultural vehicles or other agricultural machinery during operation. The use of sensing technology for the timely and accurate detection and pre-warning of obstacles during the operation of agricultural machinery is critically important for ensuring safety. In this paper, a two-dimensional lidar is used to detect obstacles in front of tractors with the Density-Based Spatial Clustering of Applications with Noise(DBSCAN) algorithm and the Minimum Cost Maximum Flow algorithm(MCMF). A method to judge whether the obstacle is static or dynamic and a classification model of different security warning levels for obstacles in different states is proposed. Actual vehicle tests were conducted, with static obstacles tested repeatedly, and dynamic obstacles tested at different directions and speeds. The results showed that the overall average warning accuracy rate is 89.95%. Prediction results were robust for obstacles in different states, indicating that this system is able to ensure the safety of agricultural vehicles during their operation and promoted the development of agricultural mechanization.

Abstract in Chinese

智能化农业车辆在田间耕作和收获过程中已广泛使用，给农业生产带来了极大地便利。然而，还存在着在作业过程中农业车辆误撞人或其他农业机械等安全问题。利用传感技术，及时、准确地对作业农业车辆周围的障碍物进行检测和预警就有着重要的作用和意义。本文利用二维激光雷达，结合DBSCAN点云聚类算法和最小费用最大流算法对障碍物进行检测和追踪，提出了判断障碍物状态的方法并对不同状态的障碍物提出对应的安全预警级别划分模型。最后进行了实车试验，对静态障碍物进行了多次重复试验，对动态障碍物进行了不同方向不同速度的多次重复试验，整体的预警准确率均值为89.95%。试验结果表明对不同状态的障碍物均取得了良好的预警效果,推动了农业机械化事业的发展。