Volumetric Analysis of an Solid Object By Supervised Learning based Segmentation of Explicit Homomorphed Images from Multiple Views
Radhamadhab Dalai · 2020
The calculation of absolute volume of a regular shaped solid object from image has been experimented and tried using image processing systems in this paper. Volume calculation method consists of image matching, feature extraction and comparison among multiple views of images. From segment based pixel counting approach the volume has been determined by pixel density relation. Finally by multiplying density the volume has been calculated from reconstructed segmented area. This approach has been experimented over several standard images of size 256 X 256 pixel size. The multiple views of a single image have been captured and they are compared to find similarity between various segmented regions. The SIFT based key features has been considered for feature comparisons. Pixel based segmentation method such as watershed segmentation has been tried and later grid based approach has been followed to find ROI (Region of Interest) first and using this segmented areas of the volume of object has been calculated. Using 3D based software the approximate height, width and length has been measured on which approximate volume of object has been calculated and then error margin has been minimized with the proposed segmented approach.