Food volume estimation method based on double-view three-dimensional reconstruction

Abstract

The invention provides a food volume estimation method based on double-view three-dimensional reconstruction. The method mainly comprises the steps of image 3D reconstruction, external calibration, dense reconstruction, and volume estimation. Particularly, the method comprises the steps of firstly performing image 3D reconstruction; performing external calibration through saliency point matching, relative gesture extraction and scale extraction; performing dense reconstruction through image correction, three-dimensional matching and point cloud generation; and finally eliminating background by means of a segmented graph and extracting a food surface, and calculating the volume of the food through integrating food surface height above a dish. According to the method of the invention, through acquiring the food image, the volume of the food is estimated through image reconstruction, thereby realizing convenient calculation process, high speed and accurate result. An accurate automatic meal evaluation tool is supplied for common groups and persons with specific nutrition requirements, and a novel full-automatic meal evaluation method for the people is realized.

Description

technical field [0001] The present invention relates to the field of image reconstruction, in particular to a dual-view Figure three A food volume estimation method for dimensional reconstruction. Background technique [0002] With the improvement of living standards, the prevalence of diet-related chronic diseases, such as obesity and diabetes, is gradually increasing worldwide, and the mortality rate is also on the rise. In addition to family genetic factors, the etiology of obesity and other diseases is mainly closely related to people's daily diet. By monitoring and analyzing the user's diet, it can help users obtain healthier eating habits. Traditionally, people need to roughly calculate the amount of food intake by recording their own dietary content and searching for relevant information, which is neither convenient nor accurate. The food volume estimation method proposed by the present invention can help users evaluate food intake through image reconstruction. [...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the problems of inconvenient use and low accuracy, the purpose of the present invention is to provide a dual-vision based Figure three The food volume estimation method of three-dimensional reconstruction, firstly perform image 3D reconstruction, then perform external calibration through salient point matching, relative pose extraction and scale extraction, then perform dense reconstruction through image correction, stereo matching and point cloud generation, and finally use the segmentation map Remove the background to extract the food surface and calculate the food volume by integrating the height of the food surface above the pan

Images