Real-Time 3D Texturizing Using ORB-SLAM Feature Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating textures in real-time are limited by a predefined set of conditional textures, failing to provide an actual representation of object surfaces in output images, and are computationally intensive, making them unsuitable for use on mobile devices.
Innovation Solution
The implementation of a system that uses the ORB-SLAM algorithm with preliminary image processing to identify promising areas for feature detection, reducing computational intensity and enabling the texturization of 3D models without a reference textures database, allowing for the use of feature-based algorithms like ORB-SLAM on mobile devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a predefined set of conditional textures is used for texturizing 3D models, then the texturizing process can be simplified and executed in real-time, but the output image cannot display the actual texture of the object surface
Solution Approach 1:
The patent performs preliminary image processing to identify promising areas for feature detection before executing the ORB-SLAM algorithm. By pre-processing the image to locate key regions, the system reduces the computational burden during real-time texturizing, enabling both actual texture capture and real-time performance
Solution Approach 2:
The patent extracts only the necessary feature points from the image using preliminary processing, rather than analyzing the entire image. This extraction of key information allows the system to capture actual object textures while reducing computational intensity for mobile device deployment
2Measurement precision
If the ORB-SLAM algorithm is used for feature-based image processing, then the texturizing precision and actual texture representation improve, but the computational intensity increases making it unsuitable for mobile devices
Solution Approach 1:
The system performs preliminary image processing to identify promising areas for feature detection before running ORB-SLAM. This pre-localization of key regions reduces the search space and computational load, enabling mobile devices to execute the algorithm with acceptable energy consumption while maintaining high feature detection accuracy
Solution Approach 2:
The patent applies feature detection selectively in identified promising areas rather than uniformly across the entire image. This localized approach concentrates computational resources on regions most likely to contain useful features, improving detection accuracy while reducing overall energy consumption on mobile devices
3Ease of manufacture
If a database of reference textures is stored for texturizing objects, then the texturizing process can be simplified, but the device complexity and memory requirements increase
Solution Approach 1:
Instead of relying on pre-stored reference texture databases, the system captures and processes actual object textures directly from images in real-time. The object itself serves as the texture source, eliminating the need for external reference databases and reducing device complexity while maintaining texturizing simplicity through automated feature-based mapping
Data Source
AI summary
The technical result obtained herein is providing the ability to display an output image having the actual texture of the photo or video image, simplifying the implementation by eliminating the need to store a database of reference textures of objects, enabling texturizing of a 3D model areas invisible on the 2D object. A feature-based ORB-SLAM algorithm can be used for image processing using sparse point cloud. An advantageous solution is provided that allows to improve the efficiency of the texturizing process and therefore allows to use the ORB-SLAM algorithm on mobile devices by reducing computational intensity of the overall process. In order to achieve these advantages, a system of preliminary image processing is implemented. Said preliminary image processing allows to identify the most promising areas for detecting object features on the images coming from the camera.