基于Faster-RCNN的苹果检测方法研究  

The Research on the Apple Detection Method Based on Faster-RCNN

作　　者：夏辉 王爱兵[1] 杜荣[1] 程嫣[1] XIA Hui;WANGAi-bing;DU Rong;CHENG Yan(Hebei Jiaotong Vocational and Technical College,Shijiazhuang,Hebei 050035,china)

机构地区：[1]河北交通职业技术学院,河北石家庄050035

出　　处：《邢台职业技术学院学报》2022年第1期101-104,共4页Journal of Xingtai Polytechnic College

基　　金：河北省“三三三人才工程”人才培养资助项目——“基于视觉SLAM的室内移动机器人自主导航关键技术研究”,项目编号:A202002009。

摘　　要：由于我国苹果市场分级不够严格,造成苹果附加值不高。文章提出一种利用Faster-RCNN深度学习网络检测苹果果径的方法,通过对苹果特征信息和位置回归参数的学习,识别并预测苹果在图像中的位置区域,进而根据像素点与实际尺寸的换算关系,计算得到果径大小。实验表明,该方法能够准确预测每个苹果的位置区域,具有很好的检测效果。Because the grading of apple market in China is not strict enough, the apple added value is not high.This paper proposes a method to detect apple fruit diameter by using ftL-RCNN deep learning network. By learning apple characteristic information and position regression parameters, apple position region in the image is identified and predicted, and then the fruit diameter size is calculated according to the conversion relationship between pixel point and actual size. The experimental results show that this method can accurately predict the location of each apple and has a good detection effect.

关 键 词：Faster-RCNN 位置预测 果径测量 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]