论文部分内容阅读
为探索激光雷达在农业机器人环境理解和导航中的应用,研究一种基于改进DBSCAN算法的果园树干检测算法。该算法使用自适应密度阈值和聚类半径对不同距离处数据点进行聚类和整合,以克服DBSCAN算法对全局变量值敏感的缺点。针对激光雷达可能扫到地面造成机器人误检的问题,采用机器人航位推算模型计算当前帧数据中待定类的距离,通过与前一帧数据中对应类距离的比较判定待定类的类别,进而对地面干扰类进行排除。试验结果表明:1)机器人正常行走时本算法能够排除噪声准确识别树干类点;2)存在果树分枝或地面干扰时,有少量漏检,平均误判果树数目为-0.13棵,能够区分出地面类和果树类。该研究可以应用到农业机器人果园环境理解和导航中。
In order to explore the application of lidar in agricultural robot environment understanding and navigation, an orchard trunk detection algorithm based on improved DBSCAN algorithm was studied. The algorithm uses adaptive density thresholds and clustering radii to cluster and integrate data points at different distances to overcome the shortcomings of the DBSCAN algorithm being sensitive to global variable values. Aiming at the problem that the laser radar may swept to the ground to cause false detection of the robot, the robot dead reckoning model is used to calculate the distance to be determined in the current frame data and the class of the to-be determined is compared with the corresponding class distance in the previous frame data, Ground interference category to be excluded. The results show that: 1) When the robot is walking normally, the algorithm can eliminate the noise and accurately recognize the trunk-like points; 2) There is a small amount of undetected false positives when there are branches or ground disturbances, and the average number of misclassified trees is -0.13, Ground and fruit trees. The research can be applied to agricultural robot orchard environmental understanding and navigation.