Depth-Based Head Detection for Biometric Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Skeletal tracking systems often fail to track human subjects effectively, particularly when they are seated, covered, or occluded, leading to incomplete biometric identification and reduced processing resource utilization.
Innovation Solution
Implementing depth-based head detection to limit the search area for facial recognition in light intensity images, allowing for reduced facial recognition processing and enabling biometric identification even when skeletal tracking is unreliable, by using a pipeline that includes a head detector and face recognizer seeded with depth-based head-detection data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If facial recognition is performed on the entire light intensity video, then identification accuracy is improved, but processing resource utilization increases and frame rate decreases
Solution Approach 1:
The patent divides the light intensity video processing into two stages: first, depth-based head detection identifies candidate head regions; second, facial recognition is applied only to these segmented regions. This segmentation reduces the processing area from the entire video frame to specific head regions, thereby maintaining identification accuracy while reducing computational load and increasing frame rate.
Solution Approach 2:
The patent performs preliminary head detection using depth video data before conducting facial recognition on light intensity video. This preliminary action identifies and isolates head regions of interest, allowing the subsequent facial recognition process to focus only on relevant areas rather than processing the entire video frame, thus improving processing efficiency.
2Measurement precision
If facial recognition is performed on the entire light intensity video, then identification accuracy is improved, but processing resource utilization increases
Solution Approach 1:
The patent segments the processing task by first detecting head regions in depth video and then applying facial recognition only to these segmented regions in the light intensity video. This reduces the total number of pixels requiring intensive processing, thereby lowering energy consumption and processing resource utilization while maintaining identification accuracy.
Solution Approach 2:
The patent introduces depth-based head detection as an intermediary step between video capture and facial recognition. This intermediary process filters and identifies relevant regions, allowing the computationally intensive facial recognition algorithm to operate only on necessary data, thus reducing overall processing resource utilization.
3Adaptability or versatility
If skeletal tracking is used for identification, then gesture-based interactions are enabled, but tracking reliability decreases when subjects are seated or occluded
Solution Approach 1:
The patent introduces depth-based head detection as an intermediary mechanism that bridges the gap when skeletal tracking fails. By detecting heads in depth video independently of full-body skeletal tracking, the system maintains identification capability in scenarios where subjects are seated, covered, or occluded, thereby improving reliability without sacrificing gesture interaction versatility.
Data Source
AI summary
A candidate human head is found in depth video using a head detector. A head region of light intensity video is spatially resolved with a three-dimensional location of the candidate human head in the depth video. Facial recognition is performed on the head region of the light intensity video using a face recognizer.