An Algorithm for Cigarette Capsules Defect Detection Based on Lightweight Faster RCNN

Abstract

Defect detection before packaging for cigarette capsules is of great significance to ensure the quality of cigarette production. This paper proposes a new type of bead defect detection algorithm based on lightweight Faster RCNN, which can detect four typical defects of bubbles, dents, scratches, and small tails in cigarette capsules. In order to meet the requirements of industrial real-time detection, the MobileNet V1 network is used to replace the Vgg16 network in the traditional Faster RCNN to achieve feature extraction. The detection speed was further improved by optimizing the number of defect candidate frames. Adam algorithm was selected to replace the traditional Momentum algorithm to realize network parameter learning. The detection results show that the average detection accuracy of the algorithm in this paper for the four defects can reach 97.34%, and the detection speed can reach 50.66 frames per second. Both the accuracy and detection speed can meet the requirements of real-time detection.