3D Object Classification via 2D Projection for Map Clutter Removal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Service providers face challenges in accurately classifying objects at a geo-location and providing an uncluttered presentation in map applications due to the complexity of 3D surface models, which can be time-consuming and inefficient.
Innovation Solution
A method that determines regions of interest by processing textured three-dimensional representations to generate two-dimensional image representations and classify objects based on two-dimensional image and three-dimensional geometry information, using techniques like photogrammetry and machine learning algorithms for efficient object classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If 3D surface models are used to represent all objects at a geo-location, then comprehensive coverage of objects is achieved, but processing time and computational complexity increase significantly
Solution Approach 1:
The patent segments the 3D surface model processing by dividing objects into different categories (buildings, natural features, vehicles, etc.) and applying different processing strategies to each segment. This allows selective classification and rendering optimization, reducing overall processing time while maintaining accurate identification of all object types.
Solution Approach 2:
The patent extracts and removes clutter objects (such as trees, rocks, and other non-essential elements) from the 3D presentation after classification. By separating essential navigation-related objects from unnecessary clutter, the system reduces rendering complexity and processing time for the final presentation while maintaining complete object detection capability.
2Loss of information
If all objects in a 3D surface model are rendered in the presentation, then complete information is provided, but the presentation becomes cluttered and difficult to interpret
Solution Approach 1:
The patent extracts and removes clutter objects (such as trees, rocks, and other non-essential elements) from the 3D presentation after classification. By separating essential navigation-related objects from unnecessary clutter, the system reduces rendering complexity and processing time for the final presentation while maintaining complete object detection capability.
Solution Approach 2:
The patent applies different rendering qualities and levels of detail to different object types based on their importance. Essential objects like buildings and navigation markers receive high-quality rendering, while clutter objects are either removed or rendered with lower detail, optimizing both information preservation and visual clarity.
3Measurement precision
If manual classification methods are used to identify and remove clutter objects, then accurate object identification can be achieved, but the process becomes time-consuming and inefficient
Solution Approach 1:
The patent replaces manual classification methods with automated machine learning algorithms and computer vision techniques. These systems use image processing, feature extraction, and classification models to automatically identify and categorize objects in 3D surface models, achieving both high accuracy and efficient processing speeds.
Solution Approach 2:
The patent changes the processing parameters by using pre-trained machine learning models with optimized classification thresholds and confidence levels. This allows the system to rapidly classify objects with high accuracy by adjusting parameters such as classification confidence thresholds, object size filters, and category-specific detection parameters.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables efficient and accurate classification of objects, allowing for cluttered presentations to be uncluttered, improving user experience by clearly presenting relevant objects in map applications.
Implementation Method 1
processing and/or facilitating a processing of the at least one textured three-dimensional representation to determine at least one two-dimensional footprint and three-dimensional geometry information
Data Source
AI summary
An approach is provided for classifying objects that are present at a geo-location and providing an uncluttered presentation of images of some of the objects in an application such as a map application. The approach includes determining one or more regions of interest associated with at least one geo-location, wherein the one or more regions of interest are at least one textured three-dimensional representation of one or more objects that may be present at the at least one geo-location. The approach also includes processing and/or facilitating a processing of the at least one textured three-dimensional representation to determine at least one two-dimensional footprint and three-dimensional geometry information for the one or more objects. The approach further includes causing, at least in part, a generation of at least one two-dimensional image representation of the one or more regions of interest by causing, at least in part, a projection of three-dimensional texture information of the at least one textured three-dimensional representation onto the at least one two-dimensional footprint. The approach also includes causing, at least in part, a classification of the one or more objects based, at least in part, on the at least one two-dimensional image representation and the three-dimensional geometry information.