A defect detection method for granular products based on deep learning

Abstract

The invention discloses a granular product defect detection method based on deep learning. The present invention has crucial practical significance: first, because the training set data of the method of the present invention is comprehensive and balanced, the target that cannot be identified by the traditional algorithm can also be accurately identified by this algorithm; Separating the area from the background can better eliminate the influence of illumination changes and improve accuracy; third, the present invention detects granular product targets based on the C-HOG feature and the SVM classification algorithm. Compared with the traditional NCC template matching algorithm, this detection The method is faster and has higher precision; finally, the present invention uses an improved deep learning algorithm to classify granular product targets based on the characteristics of granular products, which can more accurately extract the texture information of the tablet itself and perform target classification more accurately. Compared with the traditional algorithm, the present invention comprehensively reduces the false detection rate and missed detection rate of detection, greatly reduces the production cost of the enterprise, and improves the production efficiency.

Description

technical field [0001] The present invention relates to the field of machine vision, in particular to the application of machine vision and deep learning technology in product inspection, especially medical product inspection, and more specifically, to a method for detecting defects in granular products based on deep learning. Background technique [0002] Because the quality of medical supplies is closely related to the life and health of drug users, drug production monitoring also has very strict requirements. In the process of pharmaceutical production, defects such as damage, missing, discoloration, powder leakage, dirt, and pressure bubbles will inevitably occur. These defective and defective tablets must be sorted out accurately. At present, in the actual production environment, most of the defective tablets are sorted by manual inspection. This method of visual inspection not only has a great rate of missed detection and false detection, but also is slow, which serio...

AI Technical Summary

Problems solved by technology

However, this detection method uses traditional image processing algorithms for detection, the system has a narrow range of use, and the detection rate of false detection and missed detection is high. It has obvious disadvantages in the process of use and is difficult to promote.

[0004] In addition, this method requires laser irradiation, which is costly, requires high irradiation accuracy, and has poor robustness.

Images