Monocular vision for variable spray control system

The monocular vision-based system can obtain the leaf wall area characterizing the canopy parameter for online detection and real-time variable spraying, aiming to improve the accuracy of orchard spraying equipment and the utilization efficiency of pesticide. This study established a spraying system, in which canopy parameters were collected by monocular vision, and the spray volume decision coefficient was constructed by the leaf wall area and the L* value in International Commission on Illumination Lab color space to control the duty cycle of each solenoid valve to achieve variable spraying. Four spray flow models were designed to determine the spray volume decision coefficient. The coefficients of determination of the spray volumes with the duty cycle range of 15% to 65% were all over 94 and the error of the leaf wall area values obtained using the improved super green algorithm (calculated as ExG = 2.1G-1.1R-1.1B) was only 0.5%. The test showed that there is a negative relationship between canopy denseness and L*, and the value of L* is smaller in the dense area compared with the sparse area; the actual flow generated by the system is similar to the theoretical flow when the duty cycle is 65%. The field validation tests showed that the variable spraying system could refine the droplet size and increase the droplet density to a certain extent with the same coverage rate, which had advantages over the continuous spraying. In terms of droplet deposition, DV0.1 and DV0.9 were reduced by 2 μm and 18 μm, respectively, and the increase of droplet density to 75 droplets/cm2. At the same time, the improvement of droplet distribution uniformity and droplet penetration by 16% and 3%, respectively. Compared with continuous spraying, variable spraying can achieve 55.64% savings. The study demonstrates the feasibility of monocular vision in guiding spraying operations and provides a reference for the use of monocular vision in plant protection operations.