Based on the structural shortage of an agricultural labor force and the continuous increase in tomato planting scale in China, as well as the limitation of a greenhouse working environment on the development of tomato picking productivity, a motor-driven end-effector for a tomato picking robot based on a hybrid force/position control strategy was designed. First, a hybrid force/position control strategy was applied to control the picking process. Consistent with this strategy, the mechanical structure design of the end-effector was determined. The maximum torque of the finger joint motor was verified by applying a load to the end-effector fingertip under a parabolic linear angular velocity. Second, the D-H method was used to establish the end-effector single-finger and whole-hand coordinate systems and to perform forward and inverse kinematic analysis. The Monte Carlo method was used to analyze the workspace of the end-effector, and an isochronous interpolation algorithm was used to analyze the angular displacement of the motor. A basic algorithm for starting, stopping and accelerating the finger joints was designed to achieve the smooth movement of the end-effector. The control system for the end-effector was designed based on STM32F103ZET6, and the end-effector remote wireless debugging system was designed based on Tiny6410. Finally, a test prototype of the tomato picking end-effector was manufactured and picking tests were conducted, which showed that our tomato picking end-effector moved smoothly. The proposed control algorithm reduced the impact force and recorded the contact force between the end-effector and the tomato in real time, and the end-effector essentially achieved nondestructive picking. Therefore, our tomato picking end-effector demonstrated good utility in practice.

To achieve early recognition of lettuce diseases, this paper combines the technology of image processing and the classifier of support vector machine (SVM) to identify and classify two common diseases of hydroponic lettuce: leafroll and brown blotch disease (BBD). Specifically, the authors designed programs for the acquisition and preprocessing, segmentation, and feature extraction of hydroponic lettuce images, and developed an identification program for hydroponic lettuce diseases based on the SVM. On this basis, the color, shape, and texture features were extracted from these images, and adopted as the training set of the SVM. Then, the identification model for hydroponic lettuce diseases was trained with the radial kernel function as the core, and applied to identify the different types of diseases. In total, 1,800 images were selected as samples, and subjected to denoising, enhancement, segmentation, and feature extraction. The leaf features of hydroponic lettuce were extracted, and used to establish the SVM-based disease identification model. The experimental results on the test set show that the identification model could recognize 93% of hydroponic lettuce diseases, achieving an excellent identification effect.