| 基于MSRCR-YOLOv4-tiny的田间玉米杂草检测模型 |
| |
| 引用本文： | 刘莫尘,高甜甜,马宗旭,宋占华,李法德,闫银发.基于MSRCR-YOLOv4-tiny的田间玉米杂草检测模型[J].农业机械学报,2022,53(2):246-255，335. |
| |
| 作者姓名： | 刘莫尘 高甜甜 马宗旭 宋占华 李法德 闫银发 |
| |
| 作者单位： | 山东农业大学 |
| |
| 基金项目： | 国家自然科学基金青年科学基金项目（32001419）、财政部和农业农村部：国家现代农业产业技术体系项目（CARS-18-ZJ0402）和山东省现代农业产业技术体系建设项目（SDAIT-18-06） |
| |
| 摘 要： | 为实现田间环境下对玉米苗和杂草的高精度实时检测，本文提出一种融合带色彩恢复的多尺度视网膜(Multi-scale retinex with color restoration, MSRCR)增强算法的改进YOLOv4-tiny模型。首先，针对田间环境的图像特点采用MSRCR算法进行图像特征增强预处理，提高图像的对比度和细节质量；然后使用Mosaic在线数据增强方式，丰富目标检测背景，提高训练效率和小目标的检测精度；最后对YOLOv4-tiny模型使用K-means++聚类算法进行先验框聚类分析和通道剪枝处理。改进和简化后的模型总参数量降低了45.3%,模型占用内存减少了45.8%,平均精度均值(Mean average precision, mAP)提高了2.5个百分点，在Jetson Nano嵌入式平台上平均检测帧耗时减少了22.4%。本文提出的Prune-YOLOv4-tiny模型与Faster RCNN、YOLOv3-tiny、YOLOv4 3种常用的目标检测模型进行比较，结果表明：Prune-YOLOv4-tiny的mAP为96.6%,分别比Faster RCNN和YOLOv3...
|
| 关 键 词： | 杂草识别 YOLOv4-tiny 带色彩恢复的多尺度视网膜增强算法 模型剪枝 嵌入式设备 |
| 收稿时间： | 2021/11/4 0:00:00 |
Target Detection Model of Corn Weeds in Field Environment Based on MSRCR Algorithm and YOLOv4-tiny |
| |
| LIU Mochen,GAO Tiantian,MA Zongxu,SONG Zhanhu,LI Fade,YAN Yinfa.Target Detection Model of Corn Weeds in Field Environment Based on MSRCR Algorithm and YOLOv4-tiny[J].Transactions of the Chinese Society of Agricultural Machinery,2022,53(2):246-255，335. |
| |
| Authors: | LIU Mochen GAO Tiantian MA Zongxu SONG Zhanhu LI Fade YAN Yinfa |
| |
| Institution: | Shandong Agricultural University |
| |
| Abstract: | To solve the problem of low accuracy and poor real-time performance of weed recognition in corn field,a detection method of weed based on multi-scale retinex with color restoration(MSRCR)and improved YOLOv4tiny algorithm was proposed.Firstly,according to the image characteristics of weed in corn field environment,the MSRCR algorithm was used for image feature enhancement preprocessing to improve the image contrast and detail quality.Then,Mosaic online data augmentation method was used to enrich the object detection background,improve the training efficiency and the detection accuracy of small objects.Finally,The K-means++was used for a priori anchor boxes clustering analysis and channel pruning for the YOLOv4tiny model.The total parameters of the improved and simplified model were reduced by 45.3%,the model size was reduced by 45.8%,the mean average precision(mAP)was increased by 2.5 percentage points,and the average detection frame time on the Jetson Nano embedded platform was reduced by 22.4%.The proposed PruneYOLOv4tiny model was compared with Faster RCNN,YOLOv3tiny,and YOLOv4,the experimental results showed that the mAP of the PruneYOLOv4tiny model was 96.6%,which was 22.1 percentage points and 3.6 percentage points higher than that of the Faster RCNN and YOLOv3tiny,and 1.2 percentage points lower than that of the YOLOv4 model;the model size of the PruneYOLOv4tiny was 12.2 MB,which was 3.4%of the Faster RCNN,36.9%of the YOLOv3tiny,and 5%of the YOLOv4;the average detection frame time on the Jetson Nano embedded platform was 131 ms,which was 32.1%of the YOLOv3tiny and 7.6%of YOLOv4.The optimization method proposed was superior to other commonly used object detection algorithms in model size,detection time and detection accuracy,which could provide a feasible real-time weed recognition method for the field precision weeding system with limited hardware resources. |
| |
| Keywords: | weed recognition YOLOv4tiny model multi-scale retinex with color restoration(MSRCR)enhancement algorithm model pruning embedded device |
| 本文献已被 维普 等数据库收录！ |
| 点击此处可从《农业机械学报》浏览原始摘要信息 |
| 点击此处可从《农业机械学报》下载免费的PDF全文 |
|
