Object Tracking via Stereo Disparity and Batch Association
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current in-store analytics systems face challenges in accurately detecting, tracking, and counting objects of interest in videos, particularly in retail environments, due to limitations in handling depth images and grayscale or color images, which affect the reliability and robustness of shopper traffic analysis.
Innovation Solution
The system employs a computing device to obtain and process multiple frames of images, using stereo disparity information to detect objects of interest, create object tracks, and associate them over consecutive batches, leveraging image rectification, object detection modules, and avatar dictionaries to enhance tracking accuracy and robustness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If stereo disparity information is used to detect objects in depth images, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent introduces an intermediary processing layer that converts stereo image pairs into depth maps through disparity computation. This intermediary representation simplifies subsequent object detection by providing explicit depth information, thereby improving measurement precision while managing system complexity through modular processing stages
Solution Approach 2:
The system segments the complex task of object detection into distinct modules: stereo image acquisition, rectification, disparity computation, depth map generation, and object detection. This segmentation allows each module to be optimized independently, improving overall measurement precision without proportionally increasing device complexity
2Reliability
If multiple frames are accumulated for object tracking, then reliability is improved, but loss of time increases
Solution Approach 1:
The system accumulates a limited number of frames (excessive but not excessive) for tracking decisions, balancing reliability improvement against time loss. By processing a small batch of frames rather than waiting for extensive accumulation, the system achieves sufficient tracking accuracy while minimizing processing delay
Solution Approach 2:
The system performs preliminary object detection on individual frames before accumulation, pre-processing data to facilitate faster batch processing. This preliminary action reduces the computational burden during the accumulation phase, thereby improving reliability without proportionally increasing time loss
3Reliability
If object tracks are associated over consecutive batches, then reliability is improved, but device complexity increases
Solution Approach 1:
The system maintains continuous tracking by associating object tracks across consecutive batches of frames. This continuity ensures reliable tracking consistency while using efficient association algorithms that prevent excessive complexity accumulation over time
Solution Approach 2:
The system uses simplified copy-based tracking where object states are copied and updated across batches rather than performing complex re-detection. This copying approach maintains tracking reliability while significantly reducing algorithmic complexity compared to full re-detection methods
Data Source
AI summary
Various embodiments are disclosed for detecting, tracking and counting objects of interest in video. In an embodiment, a method of detecting and tracking objects of interest comprises: obtaining, by a computing device, multiple frames of images from an image capturing device; detecting, by the computing device, objects of interest in each frame; accumulating, by the computing device, multiple frames of object detections; creating, by the computing device, object tracks based on a batch of object detections over multiple frames; and associating, by the computing device, the object tracks over consecutive batches.