Fruit recognition and classification based on tactile information of flexible hand
Yangyang Wei,Lihua Cai,Haifeng Fang,Hongyao Chen
DOI: https://doi.org/10.1016/j.sna.2024.115224
IF: 4.291
2024-03-03
Sensors and Actuators A Physical
Abstract:In the field of agricultural production, the sorting of agricultural products is a basic operation performed by robot manipulators. Agricultural products are fragile and easily damaged, and their shape, texture and size are different. In order to ensure grasping compliance and the integrity of agricultural products. In this study, firstly, a three-layer flexible tactile sensor was designed and fabricated. The performance of the sensor was tested. Three functions were used to fit the relationship between the sensor pressure and the output voltage, and the function with the best fitting degree was selected for sensor calibration. The sensor exhibits two kinds of sensitivity under the loading of 0 ~ 4 N.When the applied pressure is 0 ~ 2.75 N, the sensitivity is 1.21 V / N. When the applied pressure is 2.75 ~ 4 N, the sensitivity is 0.24 V / N. Then, the sensor and the flexible gripper are combined to build a tactile information acquisition platform to collect the original data set of the tactile sequence generated when the flexible hand grabs the fruit, and the original data is preprocessed. Finally, we evaluated the performance of Support Vector Machine ( SVM), Random Forest ( RF), K-Nearest Neighbor ( KNN), Naive Bayes ( NB) and Neural Network ( NN) in fruit classification. The results show that the RF classifier achieves the highest accuracy in the training and testing stages, which are 98.64% and 93.06%, respectively, and the accuracy of other classifiers ranges from 66.43% to 95.83%. Our work is helpful to the realization of robotic tactile perception and non-invasive grasping in the field of agriculture, and can be helpful to fruit picking, handling, sorting and other related fields.
engineering, electrical & electronic,instruments & instrumentation