Design and Experimental Verification of the YOLOV5 Model Implanted with a Transformer Module for Target-Oriented Spraying in Cabbage Farming

Due to large line spacing and planting distances, the adoption of continuous uniform pesticide spraying in vegetable farming can lead waste, thus increasing cost environmental pollution. In this paper, by applying deep learning online identification methods, control technology for target-oriented is studied with cabbages as research object. To overcome motion blur low average precision under strong light conditions during operation sprayers, an innovative YOLOV5 model implanted a transformer module utilized achieve accurate cabbage fields complex environments. Based on concept, new spray system built NVIDIA Jetson Xavier NX. Indoor test results show that 96.14% image processing time 51.07 ms. When occurs, target 90.31%. Then, field experiment, when intensity within range 3.76–12.34 wlx, advance opening distance less than 3.51 cm, delay closing 2.05 error diameter 1.45 cm. The experimental indicate changes have no significant impact effect. 98.65%, savings rate reaches 54.04%. general, designed study achieves expected provide technical support spraying.