People Counting via Overlapping Sensor Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional audience measurement systems face inaccuracies in counting people due to partial visibility, lighting conditions, and field of view limitations, leading to undetected faces and false positives, which affect the accuracy of media exposure data.
Innovation Solution
The system analyzes images over a period of time, using multiple image sensors with overlapping fields of view to eliminate redundant detections and condition frames for more accurate face detection, and employs fluctuation analysis to differentiate between human and static objects, thereby improving the accuracy of people tallies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional audience measurement systems use single image sensors with limited field of view, then device complexity is reduced, but measurement precision deteriorates due to undetected faces and false positives
Solution Approach 1:
The patent combines multiple image sensors with overlapping fields of view into a unified measurement system. The sensors work together to capture a broader area, and their detections are integrated through a people counter that processes data from all sensors, eliminating blind spots and improving counting accuracy.
Solution Approach 2:
The patent divides the measurement task into specialized components: multiple image sensors capture different regions, a face detector identifies faces in each region, and a people counter aggregates results. This segmentation allows each component to be optimized for its specific function while contributing to overall accuracy.
2Measurement precision
If the system uses multiple image sensors with overlapping fields of view, then measurement precision improves by reducing blind spots, but device complexity increases
Solution Approach 1:
The patent uses multiple copies of image sensors with identical or similar specifications, each capturing a portion of the environment. Rather than using one complex sensor, multiple simpler sensors are deployed, and their outputs are processed to achieve the desired measurement precision.
3Measurement precision
If the system processes images from multiple sensors over time, then measurement precision improves by eliminating false positives, but loss of time increases due to additional processing
Solution Approach 1:
The patent performs preliminary processing of image data by capturing multiple frames over a time interval and pre-processing them to identify and eliminate false positives before final tallying. This preliminary action reduces the computational burden on the final counting operation.
Solution Approach 2:
The system continuously captures and processes image data over a sustained time interval, maintaining a running tally that is refined as more data becomes available. This continuous processing allows the system to converge on accurate counts while managing computational load through incremental updates.
Data Source
AI summary
Methods and apparatus to count people in images are disclosed. An example method includes analyzing frame pairs of a plurality of frame pairs captured over a period of time to identify a redundant person indication detected in an overlap region, the overlap region corresponding to an intersection of a first field of view and a second field of view; eliminating the identified redundant person indication to form a conditioned set of person indications for the period of time; grouping similarly located ones of the person indications of the conditioned set to form groups; analyzing the groups to identify redundant groups detected in the overlap region; and eliminating the redundant groups from a people tally generated based on the groups.