基于一种改进的Faster-RCNN的公路中央隔离带绿化识别  

Greening Recognition of Central Isolation Zone of Highway Based on an Improved Faster-RCNN

作　　者：卜华雨 杨国平 Bu Huayu;Yang Guoping(School of Mechanical and Automotive Engineering,Shanghai University of Engineering Science,Shanghai 201620,China)

机构地区：[1]上海工程技术大学机械与汽车工程学院,上海市201620

出　　处：《农业装备与车辆工程》2022年第9期114-117,共4页Agricultural Equipment & Vehicle Engineering

摘　　要：研究了一种结合不感兴趣区域的Faster-RCNN改进算法对公路隔离带中绿化的识别。通过将次最深特征层同样进行3×3的卷积并进行L2正则化得到新的特征层,将浅层次特征层与深层次特征层的特征信息进行融合,使得该算法在增强识别精度和准度的同时不增加特征选取框候取层的厚度。同时设置了不感兴趣区域对冗余检测信息进行筛除,以提高检测准确率。经试验验证,改进算法可有效完成检测任务,从而为无人驾驶浇灌车的控制提供支持。This paper studies an improved Faster-RCNN algorithm that combines regions of no interest to identify greenery in highway isolation belts.This paper uses the same 3×3 convolution of the second deepest feature layer and L2 regularization to obtain a new feature layer,and fuses the feature information of the shallow feature layer with the deep feature information.This allows the algorithm to enhance the recognition accuracy and accuracy while not increasing the thickness of the candidate layer of the feature selection frame.At the same time,this paper sets the regions of no interest to filter out the redundant detection information to improve the detection accuracy.It has been verified that the improved algorithm can effectively complete the detection task,thereby providing support for the control of the driverless irrigation vehicle.

关 键 词：浇灌车 智能驾驶 目标识别 Faster-RCNN 改进的RPN网络 

分 类 号：U463.6[机械工程—车辆工程] TP391.4[交通运输工程—载运工具运用工程]