徐印赟,江 明,李云飞,吴云飞,卢桂馥.基于改进 YOLO 及 NMS 的水果目标检测[J].电子测量与仪器学报,2022,36(4):114-123 |
基于改进 YOLO 及 NMS 的水果目标检测 |
Fruit target detection based on improved YOLO and NMS |
|
DOI: |
中文关键词: 水果目标检测 YOLO 网络 SPP 模块 NMS 信息熵 |
英文关键词:fruit target detection YOLO network SPP block NMS information entropy |
基金项目:国家自然科学基金(61976005)项目资助 |
|
Author | Institution |
Xu Yinyun | 1. Key Laboratory of Advanced Perception and Intelligent Control of High-End Equipment, Ministry of Education, Anhui Polytechnic
University,2. School of Electrical Engineering, Anhui Polytechnic University |
Jiang Ming | 1. Key Laboratory of Advanced Perception and Intelligent Control of High-End Equipment, Ministry of Education, Anhui Polytechnic
University,2. School of Electrical Engineering, Anhui Polytechnic University |
Li Yunfei | 1. Key Laboratory of Advanced Perception and Intelligent Control of High-End Equipment, Ministry of Education, Anhui Polytechnic
University,2. School of Electrical Engineering, Anhui Polytechnic University |
Wu Yunfei | 1. Key Laboratory of Advanced Perception and Intelligent Control of High-End Equipment, Ministry of Education, Anhui Polytechnic
University,2. School of Electrical Engineering, Anhui Polytechnic University |
Lu Guifu | 1. Key Laboratory of Advanced Perception and Intelligent Control of High-End Equipment, Ministry of Education, Anhui Polytechnic
University,3. School of Computer and Information, Anhui Polytechnic University |
|
摘要点击次数: 1063 |
全文下载次数: 1635 |
中文摘要: |
为使水果采摘机器人在复杂情况下如树叶遮挡、果实目标尺度变化大等情况能准确地检测出水果,提出一种 YOLO
(you only look once)改进模型与 NMS(non-maximum suppression)改进算法的目标检测方法。 首先,对传统 YOLO 深度卷积神经
网络架构进行改进,设计一种更细化的 SPP5(spatial pyramid pooling)特征融合网络模块,强化特征图多重感受野信息的融合,
并基于此模块提出一种 YOLOv4-SPP2-5 模型,在标准 YOLOv4 网络中跨层添加并改进 SPP 层,重新分布池化核大小,增强感受
野范围,从而降低目标误检率;其次,提出一种 Greedy-Confluence 的 NMS 改进算法,通过对高度接近的检测框直接抑制和对重
叠检测框综合考虑距离交并比 DIOU( distance-intersection over union)和加权接近度 WP(weighted proximity)的方法,均衡 NMS
的计算消耗并减少检测框的错误抑制,从而提高遮挡、重叠物体的检测精度;最后,分别对改进方法进行性能测试,验证方法的
可行性,随后制作水果检测数据集并进行格式转换和标签标注,然后采用数据增强技术对训练数据进行扩充,并使用
K-means++聚类方法获取先验锚定框,在计算机上进行了水果检测实验。 结果表明,基于改进 YOLO 网络及改进 NMS 的水果检
测方法在准确率方面有显著的提高,平均精度均值(mean average precision,MAP) 在 YOLOv4 上达到了 96. 65%,较原网络提升
1. 70%,并且实时性也得到了保证,在测试设备上达到了 39. 26 帧/ s。 |
英文摘要: |
To enable fruit picking robots that accurately detect target in complex conditions such as leaf covering and variances of fruit
sizes, etc. , improved YOLO( you only look once) model and NMS ( non-maximum suppression) algorithm are proposed. First, the
traditional YOLO deep convolutional neural network architecture is upgraded. A more fine-grained SPP5(spatial pyramid pooling)feature
fusion network module is generated to enhance the integration of multiple sensory field information in feature maps, based on which a
YOLOv4-SPP2-5 model is proposed. The SPP layer is added and improved in the standard YOLOv4 network across layers and
redistributed the pooling kernel size and to enlarge perceptual field range, thus decreasing the false detection rate. Moreover, an
improved Greedy-Confluence NMS algorithm is proposed. Through direct suppression of high-proximity detection boxes and
comprehensive consideration of Distance-Intersection over Union (DIOU) and weighted proximity (WP) for overlapping detection boxes,
the computational consumption of NMS was balanced and the error suppression of detection boxes was reduced, so as to improve the
detection accuracy of occlusion and overlapping objects. Finally, performance tests are conducted to verify the feasibility of the method,
followed by format converting and annotation labelling with fruit training datasets. The training datasets are expanded via data
augmentation techniques and the K-means ++ clustering approach is utilized to obtain a priori anchor frames, and the fruit detection experiments are carried out on a computer. The results demonstrate that the improved YOLO network and NMS algorithm-based approach
significantly increase the accuracy rate of fruit detection. The mean average precision (MAP) reaches 96. 65% at YOLOv4, which is
1. 70% higher than the previous network. Real-time performance is also guaranteed, hitting 39. 26 frames per second on the test device. |
查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|