Target Tracking Using Full-Frame Detection and Deep Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing target tracking methods based on correlation filters suffer from tracking drift and loss of targets due to uncontrollable target movement and large changes in target region range, particularly when using local sliding windows.
Innovation Solution
Perform target detection in the overall video frame using a target detection model to identify candidate regions, extract deep features using a feature extraction model, and determine the tracking target based on feature similarity with previous frames, expanding the tracking range and utilizing deep features for accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If target tracking is performed using local sliding windows based on previous frame target region, then tracking speed is improved, but tracking reliability deteriorates due to target loss and drift
Solution Approach 1:
The patent performs target detection in the current video frame before tracking to obtain candidate regions, rather than relying solely on previous frame information. This preliminary detection action ensures that even if the target moves quickly outside the predicted range, candidate regions are already identified, preventing target loss while maintaining efficient tracking.
Solution Approach 2:
The patent introduces deep feature extraction as an intermediary between candidate region detection and target identification. By comparing deep features of candidate regions with the target template, the system achieves more reliable target identification that is insensitive to appearance changes, thereby improving tracking reliability without sacrificing speed.
2Device complexity
If target tracking range is limited to previous target region, then device complexity is reduced, but measurement precision deteriorates due to tracking drift
Solution Approach 1:
The system performs preliminary target detection in the current frame to identify candidate regions before the tracking decision is made. This ensures that even if the target has moved significantly from its previous position, the detection system has already located potential new positions, preventing tracking drift while maintaining computational efficiency.
Solution Approach 2:
The patent replaces traditional morphological feature comparison with deep feature extraction and comparison. Deep features capture semantic information that is more robust to appearance changes, providing higher measurement precision for target identification without requiring complex mechanical adjustments to the tracking system.
3Productivity
If morphological features are used for target identification, then processing speed is improved, but manufacturing precision deteriorates due to tracking inaccuracy
Solution Approach 1:
The patent introduces deep feature extraction as an intermediary layer between image data and target identification. Although deep feature extraction requires more computation than simple morphological operations, it provides much more discriminative power for accurate target identification, achieving high tracking accuracy while maintaining acceptable processing speeds through efficient implementation.
Solution Approach 2:
The patent changes the feature representation parameters from simple morphological features to deep features extracted by neural networks. This parameter change enables the system to capture complex patterns and semantic information, significantly improving tracking accuracy while the efficient deep learning architecture maintains acceptable processing speeds.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Embodiments of this application disclose a target tracking method and apparatus, a device, and a medium. The method includes: obtaining, by a server, a first video stream, and performing the following data processing for each video frame in the first video stream: performing, by the server, detection according to a target detection model and a current video frame, to obtain candidate regions in the current video frame; performing, by the server, extraction according to a feature extraction model and the candidate regions, to obtain a deep feature corresponding to each candidate region; calculating, by the server, a feature similarity corresponding to the each candidate region according to the deep feature corresponding to the each candidate region and a deep feature of a target detected in a previous video frame; and finally determining, by the server based on the feature similarity corresponding to the each candidate region, the target detected in the current video frame. Target detection is performed in a range of the current video frame by using the target detection model, and target tracking is performed based on the deep feature, so that occurrence of cases such as a tracking drift or losing a tracking target can be effectively prevented, to ensure the accuracy of target tracking.