Facial Recognition Image Cropping for Low-Bandwidth Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Facial recognition systems struggle to identify individuals accurately when operating under low bandwidth conditions, such as those provided by Land Mobile Radio (LMR) walkie-talkies used by law enforcement officers, due to the need for higher resolution images that consume excessive battery power and time.
Innovation Solution
Send an initial lower resolution image for preliminary matching, identify distinguishing features, and request specific portions of the higher resolution image for detailed analysis, reducing bandwidth usage by focusing on relevant features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full-resolution facial images are transmitted for recognition, then recognition accuracy is improved, but network bandwidth consumption increases
Solution Approach 1:
The facial image is segmented into multiple resolution zones: a high-resolution region containing the facial features of interest and lower-resolution peripheral regions. This segmentation allows the system to transmit only the necessary high-detail information for accurate recognition while reducing overall data transmission requirements, thereby resolving the contradiction between recognition accuracy and bandwidth consumption.
Solution Approach 2:
Different regions of the facial image are assigned different quality levels based on their importance for recognition. The central facial features receive high resolution while peripheral areas use lower resolution. This local quality differentiation maintains recognition accuracy in critical areas while reducing total bandwidth consumption across the entire image.
2Productivity
If image resolution is reduced to save bandwidth, then network efficiency is improved, but recognition accuracy deteriorates
Solution Approach 1:
The image is divided into segments with different resolution requirements. By segmenting the image and applying appropriate resolution levels to each segment, the system achieves overall network efficiency through reduced total data transmission while preserving recognition accuracy in the critical high-resolution segments that contain essential facial features.
Solution Approach 2:
The resolution parameter is changed selectively across different regions of the image rather than uniformly. High-resolution parameters are applied only where necessary for recognition accuracy, while lower-resolution parameters are used in less critical areas, thereby improving network efficiency without sacrificing the recognition accuracy needed in key regions.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Techniques for saving bandwidth in performing facial recognition are provided. An image including a face may be received, over a wireless link, at a first resolution. A facial recognition system may identify a subset of people who may be associated with the face, wherein the facial recognition system cannot definitively associate the face with an individual person in the subset of people, based on the image including the face at the first resolution. A feature of the subset of people that may be used to identify a person within the subset of people may be determined. A request for the portion of the image containing the feature at a second resolution may be sent over the wireless link. The second resolution may be higher than the first.