Object Volume Prediction From Container Changes Using 3D Depth Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for calculating food waste volume face challenges such as the need for precise angle and lighting in video and photographic measurements, and difficulties in estimating volumes of foods with variable densities using weight alone, while real-time volume prediction is hindered by inaccuracies in identifying reference times before and after volume changes.
Innovation Solution
A method utilizing multi-dimensional data from a container, including depth camera data and RGBD data, to identify changes in container content and predict volume by setting reference times based on data value thresholds, using neural networks for accurate volume estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If video analysis and photographic measurement are used to calculate food waste volume, then volume measurement can be performed, but accurate angle and lighting are required which reduces ease of operation and may reduce accuracy due to texture or color fluctuations
Solution Approach 1:
The patent replaces traditional video analysis and photographic measurement methods with a depth camera-based 3D reconstruction system. This substitution eliminates the need for complex lighting conditions and angle requirements, as the depth camera directly captures spatial information through time-of-flight measurement, thereby improving ease of operation while maintaining measurement accuracy
Solution Approach 2:
The patent transitions from 2D image-based volume estimation to 3D depth map reconstruction. By capturing depth information for each pixel and reconstructing three-dimensional container geometry, the system achieves accurate volume calculation without relying on texture or color features that cause fluctuations in traditional 2D methods
2Ease of operation
If metering device is used to estimate food volume, then weight measurement is simple, but it is difficult to estimate volume of foods with variable density
Solution Approach 1:
The patent replaces weight-based volume estimation with direct 3D spatial measurement using depth camera technology. This substitution allows accurate volume measurement of foods with variable density by directly measuring physical dimensions rather than inferring volume from weight, thereby maintaining operational simplicity while improving measurement precision for variable density materials
3Productivity
If deep camera is used to predict volume of food waste in real time, then real-time prediction is enabled, but it is difficult to accurately identify reference time before and after volume change
Solution Approach 1:
The patent implements a feedback mechanism that continuously monitors depth map changes over time and automatically identifies reference time points when significant volume changes occur. The system compares sequential depth maps, detects change magnitudes, and uses this feedback to precisely mark reference time points, thereby maintaining real-time prediction capability while improving reference time identification accuracy
Solution Approach 2:
The patent performs preliminary processing of depth map data by establishing baseline reference states before volume changes occur. By pre-processing and storing reference depth maps at identified reference time points, the system enables rapid real-time comparison and volume change detection without compromising the accuracy of reference time identification
Data Source
AI summary
Disclosed is a method for predicting a volume of an object, the method performed by one or more processors of a computing device according to an exemplary embodiment of the present disclosure.The method may include: obtaining a plurality of multi-dimensional data including a container capable of containing an object based on time-series information; obtaining first reference time multi-dimensional data based on whether the plurality of multi-dimensional data are changed; identifying whether a plurality of additional multi-dimensional data after the plurality of multi-dimensional data are changed; obtaining second reference time multi-dimensional data among the plurality of additional multi-dimensional data based on identified changed; and predicting the volume of the object based on the first reference time multi-dimensional data and the second reference time multi-dimensional data.