3D Scene Feature Extraction Using Ray and Height-Map Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for feature extraction in three-dimensional scenes, such as visual image features and depth-map features, face challenges with high data dimensions and redundancy, leading to increased learning costs and inefficiencies in machine learning tasks.
Innovation Solution
A method involving shooting cone-shaped viewing-frustum rays from a target object to extract object attribute information, converting it into ray-feature vectors, and combining these with height-value matrices from different granularities to reduce dimensions and integrate into a concise, semantically rich three-dimensional scene feature.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If visual image feature extraction is used to describe three-dimensional scene, then the scene can be represented in two-dimensional form, but the data dimension becomes high and learning costs increase
Solution Approach 1:
The patent extracts only the essential three-dimensional spatial information (depth, height, vertical distance) from the complete visual scene data, rather than using all visual image features. This selective extraction of key spatial parameters reduces the data dimension while maintaining the ability to represent the three-dimensional scene effectively.
2Loss of information
If depth-map feature extraction is used to include three-dimensional information, then spatial depth information is captured, but the data dimension remains high with duplication and redundancy
Solution Approach 1:
The patent extracts specific essential three-dimensional parameters (depth, height, vertical distance) from the depth-map data, separating the crucial spatial information from redundant pixel-level depth data. This selective extraction maintains three-dimensional information completeness while significantly reducing data dimension and eliminating duplication.
Solution Approach 2:
The patent segments the three-dimensional scene representation into distinct semantic components (depth information, height information, vertical distance information) rather than using complete depth-map pixel data. This segmentation allows each component to be processed independently with reduced dimensionality.
3Loss of information
If combined visual image feature and depth-map feature extraction is performed to enrich scene features, then more comprehensive scene information is obtained, but data dimension and learning costs remain high
Solution Approach 1:
The patent merges multiple three-dimensional spatial parameters (depth, height, vertical distance) into a unified compact feature representation. Instead of combining complete visual image features and depth-map features separately (which would increase dimension), the patent integrates the essential spatial components into a cohesive low-dimensional feature vector that comprehensively describes the three-dimensional scene.
Solution Approach 2:
The patent extracts only the essential spatial parameters needed for three-dimensional scene understanding, discarding redundant visual and depth information. This selective extraction achieves comprehensive scene feature representation with significantly reduced data dimension compared to combining complete visual and depth-map features.
Data Source
AI summary
This application disclose a feature extraction method performed by a computer device. The method includes: shooting a cone-shaped group of viewing-frustum rays from a target character object for a hit object in a three-dimensional scene picture; determining object attribute information of the hit object based on an interaction between the viewing-frustum rays and the hit object; performing vector conversion on the object attribute information of the hit object, to obtain a ray-feature vector; collecting, by using a location of the target character object as a collection center, a plurality of height-value matrices corresponding to different granularities; performing feature dimension reduction on the height-value matrix corresponding to each granularity, to obtain a height-map feature vector; and integrating the ray-feature vector and the height-map feature vector into a three-dimensional scene feature.