An Edge-Detection Method for Capsule Defect on Embedded Platform

An edge intelligent detection method is proposed for identifying defects of medicinal capsules in this article, employing a lightweight convolutional neural network (CNN) model. The approach involves compressing the CNN model and optimizing the channel parameters such that the model becomes lightweight and suitable for edge-embedded devices. The lightweight model was then trained on a computer to optimize network parameters. Next, the optimized parameters were transplanted onto a field-programmable gate array (FPGA)-based edge detection device to detect the defects of medicinal capsules. The experimental results demonstrate that a lightweight network model can be successfully deployed on an FPGA-based edge detection apparatus, achieving an average identification accuracy of 95.50% for capsule defects. The proposed method provides an effective solution for intelligent edge detection in the medicinal-capsule production process.