An Improved YOLOv5 Algorithm of Target Recognition
Lei Hu
Abstract
In order to improve the accuracy and efficiency of target recognition of mature apples, and solve the problem of apple picking robot quickly avoiding irregular obstacles to effectively pick mature apples in complex environments. Based on the Yolo v5 network spatial coordinate system and the feature information network, an adaptive scaling mechanism and position focal loss function are used, an improved YOLOv5 algorithm is proposed. The experimental results show that the accuracy of the improved algorithm is improved by 8.1% and the speed of pattern recognition is improved by 3.9 frames per second. It provides a solution for apple-picking robot to efficiently recognition mature apples and positioning it in complex environments.
Topics & Concepts
Computer scienceArtificial intelligenceComputer visionRobotPosition (finance)Feature (linguistics)AlgorithmFunction (biology)Mobile robotPattern recognition (psychology)FinancePhilosophyLinguisticsBiologyEconomicsEvolutionary biologySmart Agriculture and AIIndustrial Vision Systems and Defect Detection