A deep learning-enhanced vision system for precision target spraying in chinese cabbage cultivation
Year: 2026
Type: article
Abstract: Aiming at the issue that increasing the forward speed during target spraying of calcium and other agents onto the heart and inner leaf margins of Chinese cabbage plants leads to changes in the effective sprayed area covered per unit time and the pesticide spray volume, a target spray control system based on deep learning was designed. This system aims to investigate the relationship between the forward speed of target spraying and the spray volume and coverage in the scenario of applying calcium agents to the heart of Chinese cabbage plants. Firstly, the structure and working principle of the target spray control system were elaborated. Secondly, the general YOLOv8 model was improved, proposing a lightweight BWT-YOLO model that integrates long-range recognition, sprayer forward speed, and spray response frequency. Utilizing a Jetson Xavier NX controller equipped with the lightweight BWT-YOLO model, a pressure-stabilized execution unit and a target control unit were designed, and the system performance was tested through intermittent spray experiments. During the experiments, the forward speed of the spray platform was gradually increased, and the spray volume of the corresponding nozzle was obtained based on the response frequency of the solenoid valve. The response times of the three main parts of the target spray system were recorded: the average image processing time was 29.50 ms, the decision signal transmission time was 6.40 ms, and the spray process duration was 88.83 ms. Comprehensive analysis indicated that the total response time of the target spray lagged behind the electrical signal by approximately 124.73 ms. By compensating for the solenoid valve response lag time and conducting acquisition experiments, the difference between the actual spray volume after compensation and the required volume was derived. It was determined that under the condition of 7.2 km/h speed, the corresponding difference between actual and required volume was 0.01 L/min, which was the smallest difference and met the operational requirements for target spraying. This study provides a reference for the application of target application robots in spray systems and the selection of operational parameters.
Source: Precision Agriculture
Authors: Changxi Liu, Hang Shi, Hao Sun, Hui Zhang, Qingda Li, Jun Hu
Language: English
Topic: Smart Agriculture and AI
Subfield: Plant Science
Domain: Life Sciences
SDG: Zero hunger
Open Access status: green
OpenAlex ID: W7213275445