A
computer vision-based method for calculating the volume of stacked goods within a regular cube space is disclosed. This method trains a neural network for semantic segmentation using the projected images of the stacked goods on both sides of the cube to obtain a model for calculating the
projected area of the stacked goods. This model is then used to calculate the
projected area of the collected image data to determine the volume of the stacked goods within the regular cube space. This method avoids the problems of
human error and subjective judgment inherent in traditional manual observation methods, and also avoids the drawbacks of instrument measurement methods, such as high equipment costs and maintenance expenses, and susceptibility to environmental interference. It enables accurate and efficient real-time calculation of the volume of stacked goods within a regular cube space.