3D Face Recognition With Overlapping 2D Camera Views
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Recognizing 3D objects, particularly human faces, from non-cooperative subjects is challenging due to occlusion and varying viewpoints, making it difficult to mathematically map different views of a 3D object like a human face, unlike flat objects such as license plates.
Innovation Solution
An array of 2D cameras with overlapping fields of view captures multiple images, normalizes them to a canonical view, extracts feature data, and compares it with stored data to recognize 3D objects using a feature extraction and recognition processor.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple 2D cameras are used to capture images of a 3D object from different viewpoints, then recognition accuracy can be improved, but the complexity of processing and mapping these different views increases significantly
Solution Approach 1:
The system segments the 3D object recognition problem into multiple 2D image captures from different camera viewpoints. Each camera captures a portion of the object, and these segmented views are then processed individually before being integrated to form a complete recognition result, thereby improving accuracy while managing processing complexity through division of the problem space.
Solution Approach 2:
The system creates multiple 2D image copies of the 3D object from different viewpoints using multiple cameras. These copied views serve as alternative representations that can be processed independently and then combined, allowing the system to overcome occlusion and viewpoint limitations without requiring complex real-time 3D reconstruction.
2Reliability
If images are captured from different viewpoints to overcome occlusion, then complete object recognition is improved, but mathematical mapping between views becomes impossible for 3D objects
Solution Approach 1:
The system transitions from attempting to map between different 3D viewpoints (which is mathematically complex) to capturing multiple 2D projections from different angular dimensions. By working in the 2D image space rather than trying to perform 3D coordinate transformations, the system simplifies the mapping problem while still capturing complete object information through the combination of multiple angular views.
Solution Approach 2:
The system uses a universal feature extraction and comparison process that can handle multiple different 2D viewpoints simultaneously. Rather than requiring viewpoint-specific mapping algorithms, the same feature extraction and template matching procedures are applied to each camera's images, making the system universally applicable to any viewpoint configuration.
3Measurement precision
If 3D cameras are used to capture 3D images, then recognition of non-cooperative subjects is improved, but the cost and complexity of the imaging system increases
Solution Approach 1:
The system replaces expensive, complex 3D imaging systems with multiple inexpensive 2D cameras. While each individual 2D camera image is transient and limited in information content, the combination of multiple such images provides sufficient 3D recognition capability at a fraction of the cost and complexity of dedicated 3D cameras.
Solution Approach 2:
The system substitutes complex mechanical 3D imaging systems with a simpler optical arrangement using multiple 2D cameras. By using standard 2D image capture devices positioned at different locations rather than complex 3D scanning or structured light systems, the solution achieves 3D recognition capability through a more simple and cost-effective mechanical configuration.
Data Source
AI summary
3D objects in the form of human faces are recognised at or adjacent an entrance to a building. Cameras are spaced horizontally from one another in an array to provide multiple viewpoints of a common scene with overlapping fields of view. For each of the cameras, a sequence of images of a face is generated and each of those images is normalised to a canonical view. One or more best normalised image for each camera is selected and feature data is extracted from the selected image. A recognition processor compares the extracted feature data of each selected image with stored, corresponding feature data of known 3D objects. A recognition result is output when the extracted feature data of at least one of the selected images corresponds to stored, corresponding feature data of at least one of the known 3D objects.