UCIW-YOLO: Multi-category and high-precision obstacle detection model for agricultural machinery in unstructured farmland environments
文献类型： 外文期刊
作者： Gangwei Liu;Chengqian Jin;Youliang Ni;Tengxiang Yang;Zheng Liu
作者机构：
关键词： Coordinate attention;Obstacle detection;Universal inverted bottleneck;WIoU;YOLO
期刊名称： Expert Systems with Applications
ISSN： 0957-4174
年卷期： 2025 年 294 卷
页码：
收录情况： SCIE（2025版） ; ; EI（2025版）
摘要： Unstructured farmland environments contain numerous unpredictable obstacles that pose significant challenges to the operational safety of agricultural machinery. To address this issue, this paper proposes a high-accuracy and multi-class obstacle detection model named UCIW-YOLO. Based on the YOLOv5s framework, the Universal Inverted Bottleneck (UIB) module from the lightweight MobileNetV4 network and the Coordinate Attention (CA) mechanism are integrated to construct the C3UIB and C3UIBCA modules. These enhancements strengthen the model's feature extraction capabilities for obstacle detection while reducing parameter count and computational complexity. In addition, a novel bounding box regression loss function, termed Inner-WIoU, is proposed. It assigns adaptive gradient weights to anchors of varying quality through a dynamic non-monotonic focusing mechanism, and incorporates auxiliary bounding boxes into the IoU computation, thereby accelerating model convergence. Experimental results demonstrate that UCIW-YOLO achieves strong performance in farmland obstacle detection. It achieves a mAP@0.5:0.95 of 0.845 and an F1-score of 0.965, reflecting improvements of 2 % and 0.9 % over YOLOv5s, respectively. Moreover, it maintains real-time detection at 10.09 FPS on resource-constrained edge computing platforms. The UCIW-YOLO strikes a favorable balance between detection accuracy and deployment efficiency, thereby supporting research and application in autonomous obstacle avoidance for agricultural machinery.
分类号：
- 相关文献
作者其他论文 更多>>
-
Estimation of the orientation of potatoes and detection bud eye position using potato orientation detection you only look once with fast and accurate features for the movement strategy of intelligent cutting robots
作者：Jie Huang;Xiangyou Wang;Chengqian Jin;Fernando Auat Cheein;Xinyu Yang
关键词：Cutting robot;Deep learning;Lightweight;Potato;Rotated object detection;You-only-look-once-v8
-
ANALYSIS OF VIBRATION CHARACTERISTICS AND STRUCTURAL OPTIMIZATION OF THE CHASSIS FRAME OF CRAWLER-TYPE COMBINE HARVESTER [履带式联合收割机底盘机架振动特性分析与结构优化]
作者：Zihao Zhao;Chengqian Jin;Guangyue Zhang;Youliang Ni;Zhenjie Qian
关键词：chassis;static analysis;vibration characteristics;vibration velocity
-
Variation in the response of maize (Zea mays L.) root-shoot growth and grain yield to tillage practices under various soil compactions
作者：Zhuohan Gao;Xinbing Wang;Zhigang Wang;Zaisong Ding;Lu Liang;Wenchao Zhen;Zheng Liu;Congfeng Li;Ming Zhao;Baoyuan Zhou
关键词：Grain yield;Maize;Root growth;Shoot growth;Soil bulk density;Tillage
-
EFFECTS OF MOISTURE CONTENT ON MECHANICAL CRUSHING PERFORMANCE OF SOYBEAN SEEDS AND ITS APPLICATION IN MECHANIZED HARVESTING
作者：Panpan LI;Zheng LIU;Jin WANG;Lulu LV;Anqi JIANG;Han YAN;Chengqian JIN
关键词：Breakage Rate;Mechanical Properties;Mechanized Harvesting;Moisture Content;Soybean Seeds
-
Wheat Full-Width harvesting navigation line extraction method using improved Swin-Transformer
作者：Gong Cheng;Chengqian Jin;Man Chen;Zeyu Cai;Zheng Liu
关键词：Deep learning;Harvest;Image Segmentation;Intelligent harvesting;Visual Navigation
-
Basal-topdressing trade-offs in maize production: Lower current N losses versus elevated soil residual N and subsequent losses in the Northeast China
作者：Zheng Liu;Wenli Zhou;Tujin Wang;Hong Ren;Baoyuan Zhou;Zaisong Ding;Xinbing Wang;Congfeng Li
关键词：
-
Classification and Recognition of Soybean Quality Based on Hyperspectral Imaging and Random Forest Methods
作者：Man Chen;Zhichang Chang;Chengqian Jin;Gong Cheng;Shiguo Wang;Youliang Ni
关键词：classification and identification;feature extraction;hyperspectral imaging;random forest;soybean