基于改进YOLOv7的复杂环境下香梨和果梗的采摘识别

doi:10.13733/j.jcam.issn.2095-5553.2025.06.019

摘要/Abstract

摘要：

为解决遮挡、光照、密集程度等复杂生长环境造成香梨采摘识别成功率不高及夹取采摘使果实破损的问题，提出一种基于YOLOv7的改进模型与一种剪抓相辅的采摘方式；制作基于真实复杂采摘环境下的香梨与果梗的数据集；增加BiFormer注意力机制提高模型对小目标的检测精度及效率；增加损失函数WIoU提升目标检测锚框的质量。经试验测试，改进的模型对样本数据集的平均精度均值mAP达到89.9%，比原模型提升1.9%，少特征的果梗平均精度AP值提高3.0%。与SSD、Faster R—CNN、Deformable DETR、YOLOv7模型对比，mAP值与检测速度均有较大的提升。结果表明，提升后的模型能较好完成对香梨与果梗的检测，模型识别速度快，能够满足实时检测需求。

关键词: 香梨与果梗, 注意力机制, 损失函数, 复杂环境, 图像识别

Abstract:

To address the challenges of low success rate in picking and recognizing fragrant pears, caused by a complex growth environment with factors such as occlusion, illumination, and density, as well as fruit damage during harvesting, an improved model based on YOLOv7 and a cutting‑and‑grasping picking method were proposed. A dataset was created for fragrant pears and fruit stems, reflecting real‑world complex picking conditions. The BiFormer attention mechanism was incorporated to enhance the detection accuracy and efficiency of the model, particularly for small targets. Additionally, the WIoU loss function was introduced to improve the quality of anchor frame, thereby enhancing target detection precision. The experimental results showed that the mAP of the improved model for the sample dataset reached 89.9%, which was a 1.9% increase over the original model. The AP value for detecting fruit stems, which have fewer features, increased by 3.0%. When compared to SSD, Faster R—CNN, Deformable DETR, and YOLOv7, both the mAP and detection speed showed significant improvements. These findings indicate that the enhanced model can more effectively detect pears and fruit stems while maintaining fast recognition speeds, making it suitable for real‑time detection.

Key words: fragrant pear and fruit stem, attention mechanism, loss function, complex environment, image recognition

中图分类号:

邵明明, 张立萍, 郑威强, 吕文涛, 雷佳乐. 基于改进YOLOv7的复杂环境下香梨和果梗的采摘识别[J]. 中国农机化学报, 2025, 46(6): 128-135.

Shao Mingming, Zhang Liping, Zheng Weiqiang, Lü Wentao, Lei Jiale. Picking recognition of fragrant pear and fruit stem based on improved YOLOv7[J]. Journal of Chinese Agricultural Mechanization, 2025, 46(6): 128-135.

参考文献

[ 1 ] 王瑾瑾, 王彦发, 马东芬, 等. 巴州品牌农业发展模式分析——以库尔勒香梨为例[J].农村经济与科技, 2023, 34(4): 6-9.

[ 2 ] Sébastien P, Tanguy L, Sandrine R, et al. Effect of carbohydrates and night temperature on night respiration in rice [J]. Journal of Experimental Botany, 2015, 66(13): 3931-3944.

[ 3 ] 陈夕朦, 崔菁, 赵秀文, 等. 技术伦理视野下人工智能的未来展趋势[J]. 中国软科学, 2022(S1): 76-82.

Chen Ximeng, Cui Jing, Zhao Xiuwen, et al. Future development of “artificial intelligence”: A study from the perspective of technical ethics [J]. China Soft Science, 2022(S1): 76-82.

[ 4 ] Hameed K, Chai D, Rassau A. Score‑based mask edge improvement of Mask—RCNN for segmentation of fruit and vegetables [J]. Expert Systems with Applications, 2022, 190: 116205.

[ 5 ] Yu Y, Zhang K, Yang L, et al. Fruit detection for strawberry harvesting robot in non‑structural environment based on Mask—RCNN [J]. Computers and Electronics in Agriculture, 2019, 163: 104846.

[ 6 ] Zhang J, Karkee M, Zhang Q, et al. Multi‑class object detection using Faster R—CNN and estimation of shaking locations for automated shake‑and‑catch apple harvesting [J]. Computers and Electronics in Agriculture, 2020, 173: 105384.

[ 7 ] Tian Y, Yang G, Wang Z, et al. Apple detection during different growth stages in orchards using the improved YOLO—V3 model [J]. Computers and Electronics in Agriculture, 2019, 157: 417-426.

[ 8 ] 王新彦, 易政洋. 基于改进YOLOv5的割草机器人工作环境障碍物检测方法研究[J]. 中国农机化学报, 2023, 44(3): 171-176.

Wang Xinyan, Yi Zhengyang. Research on obstacle detection method of mowing robot working environment based on improved YOLOv5[J]. Journal of Chinese Agricultural Mechanization, 2023, 44(3): 171-176.

[ 9 ] Liu T, Kang H, Chen C. ORB—Livox: A real‑time dynamic system for fruit detection and localization [J]. Computers and Electronics in Agriculture, 2023, 209: 107834.

[10] Kalfas I, Ketelaere B D, Bunkens K, et al. Towards automatic insect monitoring on witloof chicory fields using sticky plate image analysis [J]. Ecological Informatics, 2023, 75: 102037.

[11] 宋怀波, 马宝玲, 尚钰莹, 等. 基于YOLOv7—ECA模型的苹果幼果检测[J]. 农业机械学报, 2023, 54(6): 233-242.

Song Huaibo, Ma Baoling, Shang Yuying, et al. Young apple fruit detection based on YOLOv7—ECA model [J]. Transactions of the Chinese Society for Agricultural Machinery, 2023, 54(6): 233-242.

[12] 张三林, 张立萍, 郑威强, 等. 基于YOLOv5的核桃品种识别与定位[J]. 中国农机化学报, 2022, 43(7): 167-172.

Zhang Sanlin, Zhang Liping, Zheng Weiqiang, et al. Identification and location of walnut varieties based on YOLOv5[J]. Journal of Chinese Agricultural Mechanization, 2022, 43(7): 167-172.

[13] Wang C Y, Bochkovskiy A, Liao H Y M. YOLOv7: Train‑able bag‑of‑freebies sets new state‑of‑the‑art for real‑time object detectors [J]. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023: 7764-7475.

[14] Zhu L, Wang X, Ke Z, et al. BiFormer: Vision transformer with Bi—level routing attention [J]. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023: 10323-10333.

[15] Tong Z, Chen Y, Xu Z, et al. Wise—IoU: Bounding box regression loss with dynamic focusing mechanism [J]. arXiv preprint arXiv: 2301.10051, 2003.

[16] Jocher G, Stoken A, Chaurasia A, et al. Ultralytics/YOLOv5: v6.0—YOLOv5n “Nano” models, roboflow integration, tensorFlow export, openCV DNN support [J]. Zenodo, 2021.

[17] Li C, Li L, Jiang H, et al. YOLOv6: A single‑stage object detection framework for industrial applications [J]. arXiv preprint arXiv: 2209.02976, 2022.

[18] Yang L, Zhang R Y, Li L, et al. Simam: A simple, parameter‑free attention module for convolutional neural networks [C]. International Conference on Machine Learning, 2021: 11863-11874.

[19] Kirillov A, Mintun E, Ravi N, et al. Segment anything [J]. arXiv preprint arXiv: 2304.02643, 2023.

[20] Woo S, Park J, Lee J Y, et al. CBAM: Convolutional block attention module [C]. Proceedings of the European Conference on Ccomputer Vision (ECCV), 2018: 3-19.

[1]	王江晴, 马春, 莫海芳, 帖军, 田娟娟, . 基于融合双注意力机制的野生菌图像识别方法[J]. 中国农机化学报, 2025, 46(6): 70-76.
[2]	邹小波, 高文健, 石吉勇, 史永强, 申婷婷. 基于改进DenseNet与迁移学习的食品图像分类技术[J]. 中国农机化学报, 2025, 46(6): 77-84.
[3]	崔岩, 庄卫东, 秦韬, 王楠. 基于改进YOLOv5的玉米植株检测与识别研究[J]. 中国农机化学报, 2025, 46(6): 136-141.
[4]	王启航, 顾寄南, 蒋兴宇, 范天浩, 潘知瑶. 基于改进YOLOv7—tiny的茶叶嫩芽识别方法[J]. 中国农机化学报, 2025, 46(6): 150-155.
[5]	裴文杰, 刘拥民, 胡魁, 石婷婷, . 面向复杂背景的CFFA—ResNet苹果叶片病害识别模型构建[J]. 中国农机化学报, 2025, 46(6): 169-175.
[6]	甄又陈, 王佳宇, 王宁, 刘升平, 林克剑, 李艳艳. 基于改进ResNet50的草原蝗虫种类智能识别APP系统#br#[J]. 中国农机化学报, 2025, 46(5): 68-78.
[7]	刘坤, , 吉宏亚, 黄程菲, 王晓, 朱一帆. 基于改进YOLOv5s的番茄成熟度识别技术研究[J]. 中国农机化学报, 2025, 46(5): 79-85.
[8]	王靖, 姜文刚, 程耀, 钱伟. 基于Lab颜色空间改进U—Net的稻田杂草分割方法[J]. 中国农机化学报, 2025, 46(5): 148-154.
[9]	陈浩宇, 苗玉彬. 基于GASF变换和深度学习的柑橘内部品质分析[J]. 中国农机化学报, 2025, 46(4): 133-138.
[10]	赵玉清, 焦雨杰, 李宏, 王天允, 李嘉舜, 张悦, . 改进ShuffleNet V1算法的咖啡豆分级方法研究[J]. 中国农机化学报, 2025, 46(4): 194-203.
[11]	王丹阳, 梁伟红, 李玉萍, 黄贵修, . 面向不平衡数据的木薯叶部病害图像识别方法[J]. 中国农机化学报, 2025, 46(3): 101-107.
[12]	吕宗旺, 王甜甜, 孙福艳, 祝玉华, . 基于改进YOLOv8m的小麦仓储粮虫检测方法[J]. 中国农机化学报, 2025, 46(3): 108-114.
[13]	买买提·沙吾提, 阿尔庆·西力克. 基于改进YOLOv8的复杂果园环境下杏果的目标检测[J]. 中国农机化学报, 2025, 46(3): 246-252.
[14]	郭敬涛, 吕凤, 章慧婷, 杨彪, 刘大洋. 基于迁移学习与轻量化YOLOv5s的草莓目标检测方法[J]. 中国农机化学报, 2025, 46(3): 253-260.
[15]	颜士军, 朱红梅, 王雅童, 张亮. 基于字词融合和注意力机制的兽药文本命名实体识别#br#[J]. 中国农机化学报, 2025, 46(3): 336-342.