Human-Body Feature Tracking for Speed and Reliability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multiple object tracking (MOT) technologies, such as optical flow and deep learning algorithms, face challenges with slow tracking speed and poor tracking effectiveness, making them unsuitable for industrial applications.
Innovation Solution
An object tracking method utilizing deep sort technology, which includes object detection, feature extraction, state estimation, and trajectory processing, using a combination of human-face and human-body features to improve tracking speed and accuracy, with the ability to handle occlusions and adjust human-body region extraction based on image aspect ratios and expansion parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If deep learning algorithms are used for MOT, then tracking effectiveness is improved, but tracking speed becomes slow
Solution Approach 1:
The patent segments the tracking process into distinct modules: human-face region detection, human-body region determination, feature extraction, and trajectory matching. Each module processes specific information independently, allowing optimization of each component without compromising overall system performance. This segmentation enables the system to maintain high tracking effectiveness while improving speed by avoiding unnecessary computational operations in each module.
Solution Approach 2:
The patent applies partial action by extracting only the necessary features from the human-body region rather than processing the entire region. It uses a subset of historical object trajectories for matching purposes, and determines human-body regions based on expanded face regions rather than detecting all possible body parts. This selective approach reduces computational complexity while maintaining sufficient tracking accuracy.
2Speed
If optical flow and KCF algorithms are used, then tracking speed is fast, but tracking effectiveness becomes poor
Solution Approach 1:
The patent combines multiple tracking strategies into a composite system: it integrates human-face region detection with human-body region tracking, combines feature extraction with trajectory matching, and merges historical trajectory information with current frame data. This composite approach leverages the speed advantages of traditional algorithms while incorporating the effectiveness of deep learning through composite processing stages.
Solution Approach 2:
The patent introduces an intermediary feature extraction mechanism that bridges the gap between fast traditional algorithms and accurate deep learning methods. The feature extraction module acts as a mediator, processing visual information into compact representations that can be efficiently matched against historical trajectories, enabling both speed and accuracy.
3Productivity
If multiple execution threads are used for processing, then processing efficiency is improved, but system complexity increases
Solution Approach 1:
The patent divides the processing system into multiple independent execution threads, each responsible for specific tasks: one thread handles human-face region detection, another handles human-body region determination, a third performs feature extraction, and a fourth handles trajectory matching. This segmentation allows parallel processing to improve efficiency while keeping each thread's logic simple and manageable.
Solution Approach 2:
The patent designs a universal processing framework where each execution thread follows the same basic structure and can be independently optimized. The threads share common data structures and control mechanisms, allowing the system to scale by adding more threads without proportionally increasing complexity. Each thread performs multiple functions within its designated task scope.
Data Source
AI summary
The present disclosure provides an object tracking method and an object tracking device. The method includes: acquiring a human-face region of an image frame so as to determine a human-body region; extracting a human-body feature from the human-body region, and determining whether a plurality of historical object trajectories match the human-body feature; in response to that one of the plurality of historical object trajectories matches the human-body feature, updating an age of the human-body feature to a preset value; and in response to that none of the plurality of historical object trajectories matches the human-body feature, adding an object trajectory corresponding to the human-body feature to the plurality of historical object trajectories. Thus, a better tracking effect may be achieved.


