基于改进SSD网络的番茄采摘机器人目标识别方法

打开文本图片集

中图分类号：TP242.3 文献标志码：A 文章编号：1671-5276（2025）05-0208-04

Target Recognition Method of Tomato Picking Robot Based on Improved SSD Networking

REN Xiaoming

（Qinhuangdao Agricultural Comprehensive Administrative Law Enforcement Detachment，Qinhuangdao O660o1，China）

Abstract：To improvethelow eficiencyandlimitedaccuracyof tomatopicking targetrecognition，atomato picking robot targetrecognition method based on the improved single lensmultiple box detector network is proposed.Feeling field module is increased，loss function is eliminated，the maximum suppression optimization target detectionalgorithmis softened，and binocular visual target recognitionsystemis constructedforsimulation analysis.The results showthat the loss value of the improved algorithmis reduced by1.1，the area undertherecall curveis 0.945、0.931，with average mean accuracy as 0.938.The absolute error of the proposed system for locating a single tomato is less than 16mm ，whilethe relative error less than 1.60% ，the absolute error of locating multiple tomatoes is lower than 15mm ，and the relative errorbelow 1.33% ：

Keywords：tomato picking；SSD algorithm；target recognition； binocular vision；triangulatior

0 引言

随着农业现代化的发展，智能采摘机器人被广泛应用于番茄种植，而目标识别技术是提高采摘精度和效率的关键[1]。（剩余4244字）