Thông tin chung
  English
  Đề tài NC khoa học
  Bài báo, báo cáo khoa học
  Hướng dẫn Sau đại học
  Sách và giáo trình
  Các học phần và môn giảng dạy
  Giải thưởng khoa học, Phát minh, sáng chế
  Khen thưởng
  Thông tin khác
  Tài liệu tham khảo
  Hiệu chỉnh
 
Số người truy cập: 107,323,372

 Implementation of YOLOv5 for Real-Time Maturity Detection and Identification of Pineapples
Tác giả hoặc Nhóm tác giả: Trịnh Trung Hải, Nguyễn Hà Huy Cường
Nơi đăng: International Information And Engineering Technology Association (IIETA); Số: ISSN: 0765-0019;Từ->đến trang: 1445-1455;Năm: 2023
Lĩnh vực: Công nghệ thông tin; Loại: Bài báo khoa học; Thể loại: Quốc tế
TÓM TẮT
Fresh fruit is particularly to spoilage during the harvest season. Fruit farmers and suppliers who do not know how to predict fruit freshness during sorting and packaging supply fruit that is not fit for consumption as the freshness of fruit gradually deteriorates over time. Use the tradictional methods to monitoring and evalution ripe fruit would be time-consuming, expensive, and labour intensive and for incorrect result. While machines could detect and sort fruit is quickly By using technology to detect the colour, texture, appearance, size, and shape of pineapples to determine the quality and value of the fruit. Agricultural and food industries benefit from accurate fruit evaluation. The use of organic products in many products, such as juice, jams, etc., have important role in the health of humans. Producing unsuitable fruit indirectly affects the economy and increase carbon dioxide emissions (CO2). This paper uses the Fast R-CNN and YOLO-v5 techniques to identify and predict pineapple quality. This paper presents a technical solution for detecting fresh of Pineapple fruit on Vietnamese farms. YOLO is trained on more than a hundred types of objects, and the total number of images used for preprocessing is 50,000 images. The best outcomes are chosen by the aggregation model, which does this by merging the findings of the pre-trained model with the transfer model. Our new model improves performance while also cutting down on training time by making use of the greatest possible amount of training data sets. The classifier has a sensitivity of accuracy of 94.5% based on a test set of 50.000 images. The experiment results show that trained YOLOv5s has achieved the accuracy of 98% for the ripe pineapples recognition, which was an improvement of 9,27% higher than that of Faster R-CNN, while 0,22% lower than that of YOLOv5x. It is 9.2ms to detect per single image, which is 67,88% that of Faster RCNN, which is only 34,06 that of YOLOv5x. The outcomes prove that the YOLOv5s target recognition model can meet the needs both of ripe pineapples recognition accuracy and speed. It is more suitable for agricultural embedded mobile devices deployment, which can provide some technical support about accurate operation to ripe pineapples harvesting machine
ABSTRACT
Fresh fruit is particularly to spoilage during the harvest season. Fruit farmers and suppliers who do not know how to predict fruit freshness during sorting and packaging supply fruit that is not fit for consumption as the freshness of fruit gradually deteriorates over time. Use the tradictional methods to monitoring and evalution ripe fruit would be time-consuming, expensive, and labour intensive and for incorrect result. While machines could detect and sort fruit is quickly By using technology to detect the colour, texture, appearance, size, and shape of pineapples to determine the quality and value of the fruit. Agricultural and food industries benefit from accurate fruit evaluation. The use of organic products in many products, such as juice, jams, etc., have important role in the health of humans. Producing unsuitable fruit indirectly affects the economy and increase carbon dioxide emissions (CO2). This paper uses the Fast R-CNN and YOLO-v5 techniques to identify and predict pineapple quality. This paper presents a technical solution for detecting fresh of Pineapple fruit on Vietnamese farms. YOLO is trained on more than a hundred types of objects, and the total number of images used for preprocessing is 50,000 images. The best outcomes are chosen by the aggregation model, which does this by merging the findings of the pre-trained model with the transfer model. Our new model improves performance while also cutting down on training time by making use of the greatest possible amount of training data sets. The classifier has a sensitivity of accuracy of 94.5% based on a test set of 50.000 images. The experiment results show that trained YOLOv5s has achieved the accuracy of 98% for the ripe pineapples recognition, which was an improvement of 9,27% higher than that of Faster R-CNN, while 0,22% lower than that of YOLOv5x. It is 9.2ms to detect per single image, which is 67,88% that of Faster RCNN, which is only 34,06 that of YOLOv5x. The outcomes prove that the YOLOv5s target recognition model can meet the needs both of ripe pineapples recognition accuracy and speed. It is more suitable for agricultural embedded mobile devices deployment, which can provide some technical support about accurate operation to ripe pineapples harvesting machine
[ iieta_the maturity of pineapples base yolo-v5.pdf ]
© Đại học Đà Nẵng
 
 
Địa chỉ: 41 Lê Duẩn Thành phố Đà Nẵng
Điện thoại: (84) 0236 3822 041 ; Email: dhdn@ac.udn.vn