关键词: 核桃, 树干, 摇振采收, 非接触式检测, 共振频率, YOLOv5s

Abstract:  In order to improve the efficiency of walnut shaking harvesting equipment in orchard environment, a non-contact method for detecting the resonance frequency of walnut trunk was proposed. Firstly, in order to efficiently and accurately obtain the motion information of feature points on walnut trunks at each excitation frequency, the goal of lightweight integration in the orchard environment was achieved by improving the YOLOv5 network model, and the feasibility of the improved strategy was verified by ablation experiments and comparison experiments between different models, and the experiments showed that the mAP value of the improved YOLOv5s model was increased by 2.09%, the number of model parameters was reduced by 72.4%, and the detection speed of a single image could reach 9.2ms. Then, the high-frame camera was mounted on the walnut vibration equipment to carry out the field walnut trunk resonance frequency detection test on two sample trees, and finally the trunk vibration data measured by the displacement sensor were quantitatively analyzed. The results of field experiment show that the non-contact detection method proposed in this paper is more time-saving and labor-saving, and the mean errors of the two sample trees in the range of 10-25 Hz excitation frequency are 0.434 mm and 0.245 mm, respectively.

Key words: walnuts, trunk, shake harvesting, non-contact detection, resonant frequency, YOLOv5s