In the practical application of intelligent weed control devices in agricultural fields, the key issue is how to deploy the detection model to mobile devices and achieve rapid and accurate identification of weeds with different shapes and features in complex backgrounds. In response to the deployment requirements of agricultural mobile weeding equipment, this study focused on weed data during the corn seedling stage and made improvements to the YOLOv8 (You Only Look Once) detection algorithm, proposing a lightweight weed object detection algorithm. To reduce the model size and improve the operational efficiency of the model on mobile devices, the study replaced the backbone network with an improved MobileViT (Mobile Vision Transformer) lightweight model backbone. Simultaneously, the study designed a neck network that incorporated lightweight convolution modules such as GSConv (Group-Steerable Convolution) and VoVGSCSP (Voting-based Vision Guided Superpixel Co-Segmentation with Pooling) to enhance the model's feature extraction capability and the accuracy of object detection, thereby improving the accuracy and speed of model detection. By combining the Distance-IoU (DIoU) algorithm with Soft-NMS (soft non-maximum suppression), the convergence speed of the model was improved, and higher average accuracy was achieved even with multiple overlapping targets. Experimentally verifying by the self-constructed cornfield weed dataset, the number of parameters of the lightweight weed recognition model is only 1 083 025, which is 64% less compared to YOLOv8n. The precision and recall of the model reach 93% and 92.6%, the mAP50 and mAP50-95 metrics are 97.4% and 88.1%, respectively, and the detection speed of a single image reaches 92.5 frames/second, which supplies the conditions for deploying it to agricultural edge devices. The experimental results indicate that the algorithm achieves significant improvements in accuracy, speed, and model size, providing technical support for the practical application of the intelligent weeding device in agricultural fields.
LIZhaoxu, JIANGHonghai, YANGXiao, et al. Bean seedling-weed detection method based on lightweight deep learning model[J]. Agricultural Equipment and Vehicle Engineering, 2022, 60(9): 98-102. (in Chinese)
YUANTao, HUDong, MAChao, et al. An algorithm for weed target detection in rice fields based on YOLOv4[J]. Shanghai Journal of Agriculture, 2023, 39(6): 109-117. (in Chinese)
RAKHMATULINI, KAMILARISA, ANDREASENC. Deep neural networks to detect weeds from crops in agricultural environments in real-time: A review[J]. Remote Sensing, 2021, 13(21): 4486.
[7]
THANHLEV N, TRUONGG, ALAMEHK. Detecting weeds from crops under complex field environments based on Faster RCNN[C]//2020 IEEE Eighth International Conference on Communications and Electronics (ICCE). IEEE, 2021: 350-355.
[8]
SHEKHAWATN, VERMAS, AGARWALM, et al. Weed and crop detection in rice field using r-cnn and its hybrid models[C]//International Conference on Intelligent Human Computer Interaction. Cham: Springer Nature Switzerland, 2024: 37-49.
[9]
MUY, FENGR, NIR, et al. A faster R-CNN-based model for the identification of weed seedling[J]. Agronomy, 2024, 12(11): 2867.
[10]
王宇. 基于改进yolov5s模型的小麦田间杂草检测研究[D].合肥: 安徽农业大学, 2022.
[11]
ZHANGH, WANGZ, GUOY, et al. Weed detection in peanut fields based on machine vision[J]. Agriculture, 2022, 12(10): 1541.
JIWenli, LIUZhou, XINGHaihua. Lightweight method for identifying farmland weeds based on YOLO v5[J]. Transactions of the Chinese Society for Agricultural Machinery, 2024, 55(1): 212-222. (in Chinese)
[14]
HUAY, XUH, LIUJ, et al. A peanut and weed detection model used in fields based on BEM-YOLOv7-tiny[J]. Mathematical Biosciences and Engineering: MBE, 2023, 20(11): 19341-19359.
[15]
HUW, WANES O, ZHUJ, et al. Review of deep learning-based weed identification in crop fields[J]. International Journal of Agricultural and Biological Engineering, 2023, 16(4): 1-10.
[16]
ROMERO-GONZÁLEZJ.A comprehensive review of YOLO architectures in computer vision: from YOLOv1 to YOLOv8 and YOLO-NAS[J].Machine Learning & Knowledge Extractio, 2023, 5(4): 1680-1716.
[17]
MEHTAS, RASTEGARIM. Mobilevit: light-weight, general-purpose, and mobile-friendly vision transformer[DB/OL]. (2022-03-04)[2024-04-22].
[18]
SANDLERM, HOWARDA, ZHUM, et al. Mobilenetv2: Inverted residuals and linear bottlenecks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 4510-4520.
YANGJing, JINYanxia, LIUYabian, et al. Research on fabric material recognition method based on improved Transformer[J]. Journal of North University of China (Natural Science Edition), 2023, 44 (2): 138-145. (in Chinese)
[21]
LIH, LIJ, WEIH, et al. Slim-neck by GSConv: A lightweight-design for real-time detector architectures[DB/OL].(2022-06-06)[2024-04-22].
[22]
BODLAN, SINGHB, CHELLAPPAR, et al. Soft-NMS-improving object detection with one line of code[C]//Proceedings of the IEEE International Conference on Computer Vision, 2017: 5562-5570.
[23]
SHEPLEYA J, FALZONG, KWANP, et al. Confluence: A robust non-IoU alternative to non-maxima suppression in object detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023, 45(10): 11561-11574.
[24]
YANGG, WANGJ, NIEZ, et al. A lightweight YOLOv8 tomato detection algorithm combining feature enhancement and attention[J]. Agronomy, 2023, 13(7): 1824.
[25]
ZHENGZ, WANGP, LIUW, et al. Distance-IoU loss: Faster and better learning for bounding box regression[J]. Proceedings of the AAAI conference on artificial intelligence, 2020, 34(7): 12993-13000.
MENGQingkuan, ZHANGMan, YANGXiaoxia, et al. Recognition of maize seedling and weed based on light weight convolution and feature fusion[J]. Transactions of the Chinese Society for Agricultural Machinery, 2020, 51(12): 238-245. (in Chinese)
LIUZiyang, XUHuiying, ZhUXinzhong, et al. Bi-YOLO:An improved lightweight object detection algorithm based on YOLOv8n[J].Computer Engineering and Science,2024,46(8):1444-1454. (in Chinese)
YANGFeng, YAOXiaotong. A lightweight model for wheat leaf pest and disease detection based on improved YOLOv8[J]. Smart Agriculture, 2024, 6(1): 147-157. (in Chinese)
[33]
RAI N, MAHECHAM V, CHRISTENSENA, et al. Multi-format open-source weed image dataset for real-time weed identification in precision agriculture[J]. Data in Brief, 2023, 51: 109691.