Deep neural network and clustering algorithm combined large infusion liquid medicine foreign matter target detection method

Abstract

The invention discloses a deep neural network and clustering algorithm combined large infusion liquid medicine foreign matter target detection method. The method comprises the following steps: firstly, carrying out simple preprocessing on an image, and then detecting a foreign matter target in large infusion liquid medicine by combining a deep neural network for detecting and positioning a small target and a clustering algorithm and adopting a processing method of combining a single-frame image and a multi-frame image. Firstly,the FasterR-CNN deep neural network is used for detecting a single-frame image, and then hierarchical clustering and a K-means++ clustering algorithm are used to cluster the position coordinate points of all suspected foreign matter targets in the eight frames of detected images; therefore, the motion trail of each foreign matter target and the motion trail of the noise point can be recorded, the motion trail of each foreign matter target and the motion trail ofthe noise point can be conveniently distinguished, and the detection accuracy is greatly improved. The visual detection method is accurate and stable and meets the online detection requirement of a general production line.

Description

technical field [0001] The invention belongs to the field of medical image detection, and in particular relates to a large infusion liquid foreign object detection method combined with a deep neural network and a clustering algorithm. Background technique [0002] The production and consumption of large infusions in my country has already ranked first in the world, and the packaging of large infusions mainly includes glass bottles, plastic bottles, and plastic soft bags. At present, large infusion products in glass bottles still occupy a huge market share in my country. In the production process of pharmaceutical companies, due to the limitation of production technology, many visible foreign objects with a diameter larger than 50 microns such as fibers, glass chips, white spots and hairs are easily mixed in the production of traditional Chinese medicine liquid. Generally speaking, fibers come from the air. Most of the dust and glass debris come from the collision between bo...

AI Technical Summary

Problems solved by technology

Manual light inspection relies on "artificial darkroom observation", which relies on manual inspection, which is labor-intensive, slow, and has many unfavorable factors such as missed detection and false detection.

Imported equipment is expensive, difficult and costly to maintain

Images