Tile Image Scanning for Eye Tracking Head Position
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current eye tracking systems face limitations in frame rate due to full spatial resolution image acquisition, leading to increased latency and reduced data rate, especially during the initial ROI position search, which complicates subsequent data analysis and fails to meet high sampling rate requirements.
Innovation Solution
The method involves an initial head or eye position search mode where only a part of the image sensor is read out to detect features, with iterative searching and tiling of read-out parts to determine the region of interest (ROI) without requiring full frame readouts, allowing for faster ROI determination and maintaining a constant frame rate.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full spatial resolution image acquisition is used to search for head position, then measurement precision is improved, but productivity deteriorates due to lower frame rate
Solution Approach 1:
The image sensor area is divided into multiple smaller regions called tiles. Instead of reading out the entire sensor at full resolution, the system reads out only selected tiles that are likely to contain the head or eyes. This segmentation allows the system to maintain high measurement precision by searching multiple tiles while achieving higher productivity through reduced data bandwidth requirements.
Solution Approach 2:
The system performs partial image acquisition by reading out only a subset of the full sensor area (specific tiles rather than the complete frame). This partial action is sufficient to locate the head position while significantly reducing the data volume and increasing frame rate, thus resolving the contradiction between precision and productivity.
2Reliability
If full frame readout is performed to ensure complete coverage, then reliability is improved, but loss of time increases due to longer readout duration
Solution Approach 1:
The system performs preliminary scanning by reading out selected tiles in a systematic sequence to locate the head position before initiating full eye tracking. This preliminary action on a reduced dataset (selected tiles rather than full frame) significantly reduces the time to find the initial ROI while maintaining reliability through systematic tile scanning that ensures coverage of the entire sensor area.
Solution Approach 2:
The sensor area is segmented into multiple tiles that are read out in sequence or parallel. This segmentation allows the system to cover the entire sensor area (ensuring reliability) while reducing the time required for each individual readout operation, thus decreasing the total time to find the initial ROI.
3Productivity
If ROI based image acquisition is used, then productivity is improved through higher frame rate, but measurement precision deteriorates due to limited field of view
Solution Approach 1:
The system segments the sensor area into multiple tiles and performs ROI-based acquisition on each tile separately. By scanning through multiple tiles systematically, the system maintains a wide effective field of view for head position detection while still benefiting from the high frame rates enabled by ROI-based acquisition on each individual tile.
Solution Approach 2:
The system performs multiple partial readouts of different tile regions rather than a single full-frame readout. This approach achieves complete coverage (excessive action in terms of coverage) while maintaining high sampling rates through the efficiency of ROI-based acquisition on smaller regions.
Data Source
AI summary
An eye tracking method comprising: capturing image data by an image sensor; determining a region of interest as a subarea or disconnected subareas of said sensor which is to be read out from said sensor to perform an eye tracking based on the read out image data; wherein said determining said region of interest comprises: a) initially reading out only a part of the area of said sensor; b) searching the image data of said initially read out part for one or more features representing the eye position and/or the head position of a subject to be tracked; c) if said search for one or more features has been successful, determining the region of interest based on the location of the successfully searched one or more features, and d) if said search for one or more features has not been successful, reading out a further part of said sensor to perform a search for one or more features representing the eye position and/or the head position based on said further part.