基于Faster R-CNN的服务机器人物品识别研究  

Item recognition based on Faster R-CNN in service robot

作　　者：石杰 周亚丽[1] 张奇志[1] Shi Jie;Zhou Yali;Zhang Qizhi(School of Automation,Beijing Information Science & Technology University,Beijing 100192,China)

机构地区：[1]北京信息科技大学自动化学院

出　　处：《计算机应用研究》2019年第10期3152-3156,共5页Application Research of Computers

基　　金：国家自然科学基金资助项目(11672044,11172047);北京信息科技大学教改项目(2016JGYB09);2018北京信息科技大学研究生科技创新项目

摘　　要：传统的日用商品识别流程通常使用较为经典的图像识别和机器学习算法,如支持向量机(SVM)、随机森林或AdaBoost,然后利用目标图像的梯度、纹理或颜色的基本特征来对日用商品进行识别,可以在比较简单的背景中得到应用,但是在复杂的背景环境中很难有比较突出的表现,并且难以达到较高的准确率。目前在目标识别中表现比较优异的是卷积神经网络(CNN),并成为很多目标识别场景中的首选。考虑到服务机器人的硬件配置成本,将基于区域的卷积神经网络(R-CNN)的快速算法Faster R-CNN引入系统中,并以CPU计算的方式进行物品识别。利用CNN网络提取图像特征,在其后面接入一个区域提议层。实验结果表明,将深度学习的识别方法应用到服务机器人平台是可行的,识别效果准确,且在实验中得到较好的检测效果。Traditional commodity recognition processes generally use the more classic image recognition and machine learning algorithms such as support vector machines(SVM),random forest or AdaBoost,then use the basic characteristics of the gradient,texture or color of the target image to recognize commodities.It can be applied in a relatively simple background,but it is hard to have a more prominent performance in a complicated background environment,and it is difficult to achieve a high accuracy.At present,the convolution neural network(CNN),which is superior in target recognition,has become the first choice in many target recognition scenarios.Considering the hardware configuration cost of service robot,Faster R-CNN,a fast algorithm of region-based convolutional neural network(R-CNN),was introduced into the system and identified by CPU.The CNN network was used to extract image features and access to a regional proposal layer behind it.The experimental results show that it is feasible to apply the deep learning recognition method to the service robot platform.The recognition effect is accurate and the test results are good.

关 键 词：服务机器人 深度学习 FasterR-CNN 物品识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]