Research On Target Identification And Picking Methods Of Kiwifruit Harvesting Robot

The current kiwifruit harvesting process still relies on manual harvesting,which is timeconsuming,labor-intensive and inefficient.Therefore,the application of agricultural robots to kiwifruit harvesting operations is of great importance to reduce labor intensity,save production costs,and improve production efficiency.Among the various components of agricultural picking robots,the kiwifruit picking robot vision system and end-effector have an important role and are one of the core technologies of agricultural picking robots.Therefore,it is of great theoretical significance and practical value to carry out research on kiwifruit identification methods and end-effector picking methods applicable to the orchard environment to improve the accurate identification,harvesting effect,and intelligent picking of kiwifruit harvesting robots.To this end,a kiwifruit harvesting robot with integrated vision system,flexible end-effector and robotic arm has been developed and field trials have been conducted.The main contents and conclusions of the study are as follows.(1)A kiwifruit picking robot vision system was built.In this study,a lightweight Ghost NetYOLOv4 kiwifruit fruit detection algorithm is proposed.The implementations of the method are as follows: The original CSP-Darknet53 backbone network model was replaced by Ghost Net,a feature layer facilitating small object detection was introduced in the feature fusion layer,and part of the ordinary convolution was replaced by a combination of 1 × 1 convolution and depth-separable convolution to reduce the computational pressure caused by the fused feature layer.The parameters of the new network are reduced,and the generalization ability of the model is improved by loading pre-training weights and freezing some layers.The trained model was tested,and the results showed that the detection performances were better than that of the original YOLOv4 network.The F1 value,map,and precision were improved on the test set,which were 92%,93.07%,and 90.62%,respectively.The size of weight parameters was reduced to 1/6 of the original YOLOv4 network.Therefore,the method proposed in this study shows the features of light weight parameters,and high recognition accuracy,which can provide technical support for vision systems of kiwifruit picking robots.(2)Research on picking methods of end-effector.Based on the physical characteristics and growth features of kiwifruit,a four-finger flexible end-effector was proposed that takes a bottom-up envelope gripping approach.In addition,the ease of fruit picking and the force required to separate the fruit stalks are closely related to the design of the picking action.In this study,a dynamic picking data acquisition system was developed,and two simplified manual picking methods were proposed and tested for the design of the kiwifruit picking robot picking action.The test results showed that the average values of the maximum peak forces required for the "bend-pull" and "tilt-pull" picking methods were 10.9 N and 15.1 N,respectively.The picking method required less force to separate the kiwifruit.Finally,the picking parameters of the kiwifruit picking robot were optimized by using slip as the target value.The factors affecting the picking performance of the end-effector are,in primary and secondary order,the bending angle,the clamping force and the pulling force.The optimal picking angle in the test range was 60.3°,corresponding to a clamping force and a pull force of 7.98 N and7.96 N,respectively.(3)Performance trials of a kiwifruit picking robot were conducted.A kiwifruit picking robot with integrated vision system,flexible end-effector and robotic arm was developed and field trials were conducted to evaluate two picking modes.The indoor test results showed that the recognition accuracy and picking success rate were 97.9% and 92%,respectively.The field trials indicated that 76.78% and60.16% of all kiwifruit were picked successfully when the "bend-pull" and "tilt-pull" motions were used,respectively.The failure of fruit separation was the main reason for the higher success rate of the "bend-pull" than the "tilt-pull".The "tilt-pull" method results in a small angle between the stalk and the inertia axis of the fruit during separation,which increases the force required to separate the stalk,and the end-effector does not provide enough friction,resulting in the fruit falling out of the finger.The "bend-pull" method has a higher picking success rate,requires less friction from the end-effector,and makes it simpler to separate the fruit.