ABSTRACT
Aiming at the problems of poor environmental anti-interference, poor mobility of equipment, slow detection rate, false detection and missed detection in the traditional image recognition method of sprouted potato, this paper proposes a sprouted potato based on a lightweight image deep learning model (MobileNet-YOLOv4). The identification method can improve the accuracy and speed of sprouting potato identification, and at the same time improve the mobility of identification equipment. First, the collected sprouting potato picture sample data is enhanced by CutMix and Mosaic data to improve the generalization of the sample data, and the LabelImg tool is used for image labeling and data set production, and then based on the YOLOv4 model, use the more lightweight MobileNeV3 network structure replaces the main network structure (Backbone) in the original YOLOv4 model, thereby reducing the overall parameter amount of the model, improving the detection speed and the generalization of the model used in the device. Second, optimize the regression box loss function of the model, using the EIoU loss function with higher positioning accuracy is used to improve the accuracy of sprouting potato identification. Finally, the experimental results show that the improved EIoU+MobileNetv3-YOLOv4 model reduces the number of parameters by about 78% compared to the original YOLOv4 model, and the detection speed is improved 28%, the identification speed of sprouted potatoes is faster, the false detection and missed detection rate is lower, and the identification accuracy rate reaches 97.18%, thus providing better technical support for potato automation, high quality and rapid storage.
- Yamin Ji, , Non-destructive classification of defective potatoes based on hyperspectral imaging and support vector machine, Infrared Physics & Technology, 99(2019)71-79, https://doi.org/10.1016/j. infrared. 2019. 04. 007.Google ScholarCross Ref
- Wang Feiyun, Xu Huiqing, Yang Yanchen, Li Jia, Lu Chengg, Zhao Bo, Yang Bingnan. Image recognition of sprouted potato based on YOLOv3 model [C]. Potato Industry and Green Development (2021), 2021: 386-387. DOI: 10.26914/c. cnkihy. 2021. 013230.Google Scholar
- Gao Shang. Research on detection method of sprouted potato based on SVM and weighted Euclidean distance [D]. Jilin Agricultural University, 2020. DOI: 10. 27163/d. cnki. gjlnu. 2020. 000117.Google Scholar
- Huang Tao. Research on the detection method of potato internal and external defects based on semi-transmission hyperspectral imaging technology [D]. Huazhong Agricultural University, 2015.Google Scholar
- Li Jinwei, Liao Guiping, Jin Jing, Yu Xiaojuan. Detection method of potato surface defects based on grayscale interception segmentation and ten-color model [J]. Chinese Journal of Agricultural Engineering, 2010, 26(10): 236-242.Google Scholar
- Wang Feiyun, Xu Huiqing, Yang Yanchen. Image recognition of sprouted potato based on YOLOv3 model [C]. Potato Industry and Green Development (2021), 2021: 386-387. DOI: 10.26914/c. cnkihy. 2021. 013230.Google Scholar
- FARHADI A, REDMON J. Yolov3: an incremental improvement [DB/OL]. https://arXiv.org/pdf/1804.02767.pdf. 2018. 04.Google Scholar
- BOCHKOVSKIY A, WANG C Y, LIAO H Y, Yolov4: optimal speed and accuracy of object detection [DB/OL]. https://arXiv.org/pdf/2004.10934. 2020. 04Google Scholar
- Wan Lang, Ling Yutao, Zheng Xicong, Li Xiayu. Vehicle Recognition Based on MobileNetV3-YOLOv4 [J]. Software Guide, 2021, 20(12): 173-178.Google Scholar
- HOWARD A, SANDLER M, CHU G, Searching for mobilenet-netv3 [C], Proceedings of the IEEE/CVF International Conference on Computer Vision, 2019: 1314-1324.Google Scholar
- WANG C Y, LIAO H Y M, WU Y H, CSPNet: a new back-bone that can enhance learning capability of CNN [C]. Proceedings of the IEEE/CVF Coference on Computer Vision and Pattern Recog-nition Workshops, 2020: 390-391.Google Scholar
- HE K, ZHANG X, REN S, Spatial pyramid pooling in deep convolutional networks for visual recognition [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2015, 37(9): 1904-1916.Google Scholar
- LIU S, QI L, QIN H, Path aggregation network for instance segmentation [C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Analysis and Machine Intelligence, 2015, 37(9): 1904-1916.Google Scholar
Recommendations
A Lightweight Target Detection Algorithm Based on Improved MobileNetv3-YOLOv3
Knowledge Science, Engineering and ManagementAbstractTo solve the problems of complex model structure, large number of parameters, and high resource consumption that make it difficult to meet the real-time requirements of embedded target detection tasks, this paper proposed a lightweight target ...
Detection of powdery mildew on strawberry leaves based on DAC-YOLOv4 model
Highlights- Real time detection of strawberry leaves powdery mildew using deep neural network.
AbstractStrawberry powdery mildew (PM) is the main disease affecting the yield and quality of strawberries in recent years, which always appears on the back side of leaves in the early stage. Traditional methods of disease detection are labor-...
Fabric Defect Target Detection Algorithm Based on YOLOv4 Improvement
Web Information Systems and ApplicationsAbstractFabric defect detection is a key part of product quality assessment in the textile industry. It is important to achieve fast, accurate and efficient detection of fabric defects to improve productivity in the textile industry. For the problems of ...
Comments