中圖分類(lèi)號(hào)： TP393 文獻(xiàn)標(biāo)識(shí)碼： A DOI： 10.19358/j.issn.2096-5133.2021.12.008 引用格式： 錢(qián)雪，李軍，唐球，等. 基于YOLOV5的藥品表面缺陷實(shí)時(shí)檢測(cè)方法[J].信息技術(shù)與網(wǎng)絡(luò)安全，2021，40(12)：45-50.
Real-time detection method of pill surface defects based on YOLOV5
Qian Xue1，Li Jun1，Tang Qiu2，Qian Xiaoyu1
(1.School of Information Management，Beijing Information Science and Technology University，Beijing 100192，China； 2.National Computer System Engineering Research Institute of China，Beijing 100083，China)
Abstract： In the actual production process of drugs, there are always surface defects such as foreign matter, particle shortage and drug body damage. These defects may affect the use effect of products, or cause huge accidents in the use process, resulting in loss of life and property. Aiming at the application problems of deep learning model in the surface defect detection of practical industrial products with few defect samples and low detection accuracy of small defects, YOLOV5, one of the current mainstream target detection algorithms, is applied to the drug detection scene, and a kind of one stage real-time defect detection system(RDD_YOLOV5) with high accuracy, less required standard samples and fast detection speed is proposed. The primary features of the original image are used for data enhancement, combined with attention mechanism and multi-scale feature fusion, the ability of the backbone network to extract cross-channel semantic information is increased, and the high-level semantic information and low-level fine-grained information are fully integrated to improve the recognition effect of the detection efficiency of this model in small defect detection. Under the condition of limited samples, the method achieves high accuracy, and the detection efficiency of this method reaches 96.6% mAP, 32 FPS.
Key words : defect detection；deep learning；object detection；YOLOV5
