3D Object Classification via Stereo Depth and Probability Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for object detection in scenes with varying illumination, such as outdoor scenarios, lack robustness and fail to effectively differentiate between multiple types of objects like pedestrians and vehicles while maintaining privacy, especially in uncalibrated environments.
Innovation Solution
A computer-implemented method using stereo/3D processing with two cameras to determine the absolute location and type of objects by combining probability levels based on size, movement, and image matching, allowing for robust object classification and tracking in mixed scenes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If video analysis is used to determine object information, then productivity and information extraction are improved, but privacy is compromised
Solution Approach 1:
The patent transitions from 2D video frame analysis to 3D point cloud processing by utilizing depth information from stereo cameras. This dimensional change enables more accurate object classification while maintaining privacy through aggregation at higher levels of abstraction, resolving the contradiction between detection capability and privacy protection
Solution Approach 2:
The patent introduces aggregated object information (counts, average dimensions, velocity) as an intermediary representation between raw video data and final analysis results. This intermediary layer enables productivity improvement through structured data while protecting privacy by eliminating identifiable individual characteristics
2Object-affected harmful factors
If low resolution video frames are used, then privacy is protected, but measurement precision and object classification accuracy deteriorate
Solution Approach 1:
The patent adds the depth dimension to transform 2D low-resolution video frames into 3D point clouds with accurate spatial information. This enables precise measurement of object dimensions and accurate classification while maintaining privacy through aggregated statistical representations rather than detailed individual data
Solution Approach 2:
The patent changes the parameter representation from pixel-based 2D coordinates to 3D spatial coordinates (x, y, z) with associated velocity and dimension information. This parameter transformation enables accurate object characterization while allowing privacy protection through aggregation and statistical summarization
3Device complexity
If single camera 2D detection is used, then device complexity is reduced, but reliability and robustness in varying illumination deteriorate
Solution Approach 1:
The patent combines data from multiple stereo camera pairs to create a unified 3D point cloud representation. This merging of multiple data sources improves detection reliability and robustness against varying illumination conditions while maintaining manageable system complexity through integrated processing
4Productivity
If detailed object information is extracted, then productivity and classification accuracy are improved, but privacy loss increases
Solution Approach 1:
The patent uses aggregated object information (counts, average dimensions, velocity distributions) as an intermediary between detailed 3D object data and final analysis results. This enables high productivity through structured quantitative data while protecting privacy by eliminating identifiable individual characteristics through aggregation
Solution Approach 2:
The patent extracts only the essential aggregated features (counts, average dimensions, velocity) needed for analysis while leaving out identifiable individual details. This selective extraction maintains productivity by providing sufficient structured data while protecting privacy by removing personally identifiable information
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention relates to a method for categorizing a moving object within a scene In particular, the present invention relates to a method specifically taking into account the three dimensional data within a scene for determining the location of and type of objects present within the scene. The method involves determining a first probability level for the type of object based on its size and shape, determining a second probability level for the type of object based on its relative speed, and defining the type of the object based on a combination of the first and the second probability level. The invention also relates to a corresponding system and a computer program product.