基于机器人视觉系统的物体检测技术探究  

Research on Object Detection Technology Based on Robot Vision System

作　　者：贾佳[1] JIA Jia(School of Mechanical and Automotive Engineering,Kaifeng University,Kaifeng 475004,Henan)

机构地区：[1]开封大学机械与汽车工程学院,河南开封475004

出　　处：《开封大学学报》2022年第4期88-93,共6页Journal of Kaifeng University

基　　金：开封市科技攻关项目《基于Basra的果蔬采摘机构关键技术研究》(2202014);河南省科技攻关项目《4SD—800联合大蒜收获机的研究与试验分析》(202102110272)。

摘　　要：机器人视觉系统的物体检测,是指依托人工智能深度学习技术、卷积神经网络模型,对某一图像内的目标特征点、边缘稳定的极值点等作出检测,预测图像像素点在立体空间中的具体位置、物体尺度特征。而原有的AlexNet分类检测模型、RCNN位置框定模型、Fast RCNN卷积训练模型,对空间物体的视觉图像检测来说,存在着结构复杂、检测时间长等缺点。基于此,在卷积主干网络、区域网络背景下,可利用Faster RCNN算法模型、特征点描述检测子,对框定区域的物体图像信息、公开数据集进行筛选、模型训练等检测,以增强固定视觉网络下物体检测的实时性,提高检测质量。Object detection in vision system of robot is based on the deep learning technology of artificial intelligence and convolutional neural network model,it can detect target feature points in an image and extremum point with stable edge,it can also predict specific location and object scale characteristics of image pixels in three-dimensional space.The structure of the original detection model is complex,such as AlexNet classification detection model,RCNN position framing model and FastRCNN convolution training model,it needs long time when applied to visual image detection of space objects.Faster RCNN algorithm model and feature points can describe detector in the context of convolution backbone network and regional network,it is also able to detect image information and public database,which improves the real-time and quality of object detection under fixed vision network.

关 键 词：机器人 视觉系统 物体检测技术 改进Faster RCNN算法 

分 类 号：TP242.62[自动化与计算机技术—检测技术与自动化装置]