Vehicle-Mounted Passenger Counting with Occlusion-Resistant Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing systems for counting passengers in public transportation systems face challenges with occlusion and mis-re-identification, particularly when using mobile sensors, leading to inaccuracies and decreased reliability.
Innovation Solution
The method employs cameras mounted on transportation vehicles to capture images continuously around predefined locations, using a window of interest and filtering conditions to track and count passengers, avoiding occlusion and mis-re-identification by applying boundary boxes and unique identifiers, and optionally utilizing machine learning for optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If mobile sensors are used for passenger counting, then the system can capture images from multiple angles and positions, but occlusion effects and mis-tracking of persons occur leading to reduced counting accuracy
Solution Approach 1:
The patent segments the image processing task by introducing a window of interest that divides the image into relevant and irrelevant regions. Only persons within the window of interest are tracked and counted, which segments the processing scope to avoid occlusion effects from persons outside the critical area while maintaining comprehensive capture capability.
Solution Approach 2:
The patent applies local quality by assigning different processing priorities to different regions of the image. The window of interest receives enhanced processing attention with strict tracking criteria, while other regions are processed with lower priority or filtered out, optimizing resource allocation and improving counting accuracy in the critical zone.
2Productivity
If similarity analysis is used to count persons by comparing actual photos with historical photos, then the system can estimate passenger numbers, but the method is difficult to scale and leads to errors due to varying outfits and postures
Solution Approach 1:
The patent performs preliminary action by pre-defining windows of interest and tracking criteria before the actual counting process. This preliminary setup establishes strict boundary conditions and identification rules that guide the subsequent real-time processing, enabling both efficiency and accuracy without relying on computationally intensive similarity analysis of historical data.
Solution Approach 2:
The patent changes the processing parameters by shifting from global image similarity comparison to localized region-based tracking with specific geometric and temporal parameters. This parameter transformation enables scalable real-time processing while maintaining high accuracy through objective mathematical criteria rather than subjective similarity measures.
3Loss of information
If historical distributions and historical parameters are used for comparison, then the system can provide baseline data, but the choice of time period significantly affects results and decreases reliability
Solution Approach 1:
The patent extracts only the essential counting information from the image data within the window of interest, discarding unnecessary historical comparison data. This extraction approach eliminates the reliability issues associated with historical period selection while maintaining the ability to provide accurate baseline data through direct real-time measurement.
Solution Approach 2:
The system performs self-service by using its own real-time measurements to establish baselines, eliminating dependence on external historical data. The counting system serves itself by continuously accumulating accurate measurement data that automatically forms reliable baselines without requiring manual selection of historical periods or external reference data.
4Device complexity
If fixed camera systems at static infrastructure are used, then the system structure is simple, but the systems suffer from crowded scenes and occlusion leading to poor counting performance
Solution Approach 1:
The patent introduces dynamics by implementing a moving window of interest that adapts its position and size based on the transportation vehicle's movement and the scene layout. This dynamic approach allows the system to maintain simple fixed camera infrastructure while achieving superior counting precision through adaptive region-of-interest tracking that responds to changing scene conditions.
Data Source
AI summary
Method for counting the number of persons being at a predefined location before entering a transportation vehicle 1, wherein the method includes the steps of receiving images taken by one or more cameras 2a, 2b mounted at the transportation vehicle 1 and performing for each image processing steps of, wherein the method further comprises track boundary boxes BB of each detected person in all received images and count the number of persons.