Object Volume Prediction Using Before-and-After 3D Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for calculating the volume of objects, such as food waste, face challenges in accuracy due to angle and lighting requirements, texture fluctuations, and difficulty in estimating volume based on weight with variable density, especially when using deep cameras to predict volume changes.
Innovation Solution
A method utilizing multi-dimensional data from a container, including steps to identify object areas in images and correct empty spaces in point cloud data, followed by integrating and filtering data to predict the volume accurately.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If video analysis or photographic measurement is used to calculate volume, then the measurement process is simple, but the accuracy is reduced due to angle and lighting requirements and texture fluctuations
Solution Approach 1:
The patent replaces traditional mechanical/optical measurement systems (video analysis, photographic measurement) with a depth camera-based 3D sensing system. This substitution eliminates the need for controlled lighting conditions and specific viewing angles, as the depth camera directly captures spatial information through time-of-flight or structured light methods, thereby maintaining measurement simplicity while significantly improving volume calculation accuracy.
Solution Approach 2:
The patent transitions from 2D image analysis to 3D depth-based measurement. By capturing depth information in addition to visual data, the system obtains true spatial coordinates of object surfaces, enabling accurate volume calculation without being affected by lighting conditions or texture variations that plague 2D imaging methods.
2Ease of manufacture
If metering device is used to estimate volume, then the measurement process is simple, but it is difficult to estimate volume of food with variable density only with weight
Solution Approach 1:
The patent replaces weight-based measurement systems with optical-depth sensing systems. Instead of measuring mass and attempting to infer volume through density assumptions, the depth camera directly captures the three-dimensional shape and dimensions of the object, providing accurate volume measurement independent of the object's density or composition.
3Ease of manufacture
If deep camera is used to predict volume of food waste, then the measurement process is simple, but it is difficult to calculate accurate volume even with data before and after change
Solution Approach 1:
The patent replaces standard 2D deep cameras with depth-capable sensing systems that capture true 3D spatial information. This substitution enables accurate volume calculation by directly measuring the three-dimensional boundaries of objects, eliminating the need for complex 2D-to-3D reconstruction algorithms that are sensitive to lighting, angle, and texture variations.
Solution Approach 2:
The patent transitions from 2D image processing to 3D depth mapping. By capturing depth information for each pixel, the system creates accurate three-dimensional point clouds that directly represent object geometry, enabling precise volume calculation through simple geometric integration rather than complex image analysis.
4Loss of information
If reference time before and after volume change is identified to secure data, then more data is available for analysis, but it is still difficult to calculate accurate volume through the secured data
Solution Approach 1:
The patent replaces complex temporal data analysis methods with direct 3D spatial measurement. Instead of attempting to calculate volume changes by analyzing 2D images at different time points and inferring three-dimensional changes, the depth camera directly captures the three-dimensional state at each time point, allowing accurate volume calculation through direct geometric measurement rather than indirect inference.
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 first-time image and a second-time image including a container capable of containing an object; identifying a second object area for the second-time image and identifying a first object area for the first-time image; obtaining first multi-dimensional data based on the first object area included in the first-time image and obtaining second multi-dimensional data based on the second object area included in the second-time image; and predicting a volume of the object based on the first multi-dimensional data and the second multi-dimensional data.