Research on real-time detection of large-granularity green pellets based on YOLOV3 algorithm

• Published in: Metalurgija Vol. 63; no. 3-4; pp. 329 - 332
• Main Author: Z. Yang
• Format: Journal Article
• Language: English
• Published: Croatian Metallurgical Society 2024
• Subjects: deep learning; detection; granularity; green pellets; YOLOv3
• ISSN: 0543-5846; 1334-2576

In order to realize the real-time detection of abnormal green pellet particle size. First, image data of large-granularity green balls at different disk pelletizing machine material disk speeds and different camera angles are collected on site; then LabelImg software is used to label the image data of large-granularity green balls; and finally based on the YOLOv3 algorithm under the Pytorch deep learning framework train and detect large-grained ball image data. The experimental results show that: under the condition of high rotation speed of the material disk of the disc pelletizing machine, the detection accuracy can reach more than 90,58 % for the image data of a single large-grained green ball, and the comprehensive detection rate can reach more than 85 %.