Highlights A cartesian robotic spraying system was developed for precision apple blossom thinning. Flower clusters were detected and localized with deep learning model for target spraying. A communication algorithm was developed for positioning the spray end-effector to the target flowers. The cartesian robotic system greatly reduced chemical usage while maintaining thinning effectiveness in the final green fruit set. Abstract. Crop thinning, including blossom thinning, is one of the critical management strategies that determines the annual profitability of apple orchards. Challenges still remain for applying appropriate amounts of chemical thinner; if thinning is inadequate and too many fruits remain on the tree, fruit size will be small, fruit quality will be poor, and flower bud initiation for the following year’s crop may be either reduced or eliminated. Over-thinning also carries economic perils since yield and crop value in the year of application will be reduced. In addition, chemical thinning with excessive spray volume may cause leaf damage and fruit russeting. Thus, a robotic apple blossom thinning system was proposed, aiming to reduce the usage of chemical thinner while maintaining good thinning performance. The robotic system consisted of three major components: (1) a machine vision system that can identify and localize the apple flower clusters in tree canopies, (2) a cartesian robotic system with the guidance of a machine vision system to reach target flower clusters, and (3) a flat-shaped spraying nozzle connected with a solenoid valve as a spraying end-effector to deposit chemical thinner to the targeted flower clusters. A set of field tests was conducted to evaluate the performance of the robotic thinning system by comparing it to conventional air-blast and boom-type sprayers. In the test, the flower cluster detection reached a precision of 93.82%. The integrated robotic system used 2.3 L of chemical thinner to finish the chemical thinning for 18 apple trees, followed by the boom sprayer and air blast sprayer with 4.2 and 6.8 L usage, respectively. The robotic system also obtained an average fruit set of 2.4 per cluster after thinning, which was comparable to that of the air blast sprayer. The results showed that the robotic thinning system saved 66.7% and 45.5% of chemicals compared to the air-blast sprayer and boom-typed sprayers, respectively, while achieving a similar fruit set per cluster. The outcomes of the study provided guidance for developing a full-scale robotic chemical thinning system for modern apple orchards. Keywords: Apple orchard, Blossom thinning, Cartesian robot, Chemical thinning, Machine vision.
Read full abstract