采摘机器人触觉信息特征提取与处理研究

Research on Feature Extraction and Processing of Tactile Information for Picking Robot

针对采摘机器人对被抓果实感知能力有限、触觉反馈信息不丰富等问题,提出了一种基于主成分分析法的触觉特征提取与处理的方法。首先,将4×6阵列式触觉传感器集成在机械手中,通过控制机械手的闭合来抓持苹果样本,触觉传感器可实时采集接触信息。然后,对采集到的触觉信息进行预处理,并结合主成分分析法特征提取,以获取触觉特征集。最后,为验证特征集数据的有效性,将特征集作为反馈信息用于抓取果实预测分拣。结果表明:基于主成分分析的算法能够成功实现抓取触觉数据集的维数约简,并找到了最能反映接触果实过程的特征。该方法消除了接触数据的冗余,解决了特征提取困难的问题,验证了特征集能够用于采摘机器人对果实特性进行辨识感知。

Aiming at problems that the picking robot has limited ability to perceive the grasped fruit and insufficient tactile feedback information, a method of tactile feature extraction and processing based on principal component analysis was proposed.Firstly, a 4×6 array tactile sensor was integrated into the manipulator, and the apple sample was grasped by controlling the closing of the manipulator, and the tactile sensor could collected contact information in time. Secondly, the collected tactile information was preprocessed, and combined with principal component analysis method feature extraction, to obtain a tactile feature set. Finally, in order to verify the validity of the feature set data, the feature set was used as the feedback information for the prediction and sorted of the grabbing fruit. The results show that the algorithm based on principal component analysis can successfully achieve the dimensionality reduction of the grasping tactile data set, and find the features that can best reflect the process of touching the fruit.This method eliminates the redundancy of contact data, solves the problem of difficulty in feature extraction, and verifies that the feature set can be used for the identification and perception of fruit characteristics by picking robot.