Image Processing-Based Weight Estimation for Steel Cylinders

We have been tackling a problem of weight estimation of steel cylinders using image processing with an error of less than ±10g. In our previous studies, the weight was estimated by capturing multiple images of an entire side of the steel cylinder, extracting the contours via image processing, and then constructing a 3D model based on the obtained data. However, the estimation performance was poor. To improve accuracy, a simple cylinder model was considered in addition to the previous method, and multiple regression analysis was applied on the weights calculated from both models. As a result, high estimation performance was achieved. However, detailed analyses and discussions of these results were insufficient. Therefore, they are investigated in this study.