People Counting System Using Fluctuation Analysis to Eliminate False Positives
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Audience measurement systems face inaccuracies in counting people due to undetected faces, false positives, and double counting when using multiple image sensors, leading to unreliable media exposure data.
Innovation Solution
The system analyzes images over a period, uses grouping algorithms to eliminate transient detections, conditions frames to eliminate static false positives, and employs multiple image sensors with overlap detection to avoid double counting, while optimizing frame capture and processing to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If multiple image sensors are used to capture images of the environment, then the coverage area and detection capability are improved, but the risk of double counting individuals in overlap regions increases
Solution Approach 1:
The system divides the monitoring environment into distinct fields of view for each image sensor, identifying overlap regions where multiple sensors capture the same area. By segmenting the analysis into non-overlap and overlap regions, the system processes images from different sensors independently in their respective valid regions, then combines results without double-counting individuals in overlap zones.
2Productivity
If face detection algorithms are applied to all detected faces in images, then the counting speed is improved, but the accuracy decreases due to false positives from static objects and transient detections
Solution Approach 1:
The system performs preliminary actions by analyzing temporal patterns before final counting. It detects transient faces that appear and disappear quickly across frames and eliminates them before final tallying. Additionally, it pre-identifies overlap regions and applies specific filtering logic to detections in those areas, removing potential false positives before they contaminate the final count.
Solution Approach 2:
The system uses feedback by analyzing multiple frames sequentially, where detections in previous frames inform the processing of current frames. It tracks face detections across time, identifying patterns that indicate transient detections or false positives, and uses this temporal feedback to filter inaccurate detections before generating the final count.
3Speed
If the system processes every detected face in every frame, then the responsiveness is improved, but computational resources are wasted on false positives and transient detections
Solution Approach 1:
The system extracts and processes only the relevant portion of image data by focusing face detection and analysis specifically on identified overlap regions and areas with temporal patterns indicating genuine presence. It separates and eliminates transient detections and false positives from the processing pipeline, extracting only meaningful detections for final counting, thereby reducing unnecessary computational resource consumption.
Data Source
AI summary
Methods and apparatus to count people in images are disclosed. An example method includes generating a first fluctuation factor for a first frame by averaging fluctuation values of a random set of pixels of the first frame; generating a second fluctuation factor for a person indication area of the first frame by averaging fluctuation values of pixels of the person indication area; and marking the person indication area as a false positive when the second fluctuation factor is less than or equal to the first fluctuation factor.