Image Tracking via Dynamic Masking Map for Partial Occlusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object tracking systems face accuracy issues when target objects are partially masked by other objects, particularly in crowded scenes or lane changes, leading to reduced tracking precision.
Innovation Solution
An image processing device that generates a masking map to quantify the degree of masking for each block of the target object, allowing for weighted template matching to improve tracking accuracy by adjusting weights based on masking levels and updating the masking map dynamically.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If template matching is performed using conventional methods, then tracking can be implemented, but tracking accuracy is significantly reduced when the target object is masked by other objects
Solution Approach 1:
The template is divided into multiple blocks, and masking detection is performed for each block independently. This segmentation allows the system to identify which specific regions are masked while maintaining information from unmasked regions, thereby preserving tracking accuracy even when partial masking occurs.
Solution Approach 2:
The invention applies different processing strategies to different regions of the template based on their masking status. Unmasked blocks are used for reliable matching, while masked blocks are handled differently (e.g., excluded or given lower weight). This local differentiation maintains overall tracking accuracy by leveraging reliable regional information.
2Reliability
If multiple templates are used to prevent accuracy reduction, then tracking robustness improves, but the method fails when the target vehicle is widely masked
Solution Approach 1:
By segmenting the template into blocks and evaluating masking status for each block, the system can determine whether sufficient unmasked regions remain for accurate tracking. This approach maintains reliability by adapting to different masking scenarios rather than relying on a fixed multiple-template strategy.
Solution Approach 2:
The invention dynamically changes the masking detection parameters (such as threshold values or block sizes) based on the current masking situation. When wide masking is detected, the system adjusts its parameters to maximize the use of available unmasked regions, thereby maintaining tracking accuracy in previously failing scenarios.
3Measurement precision
If verification of tracking correctness is performed after individual vehicle tracking, then final result accuracy can be determined, but accuracy cannot be maintained in cases of wide masking or complicated vehicle behavior
Solution Approach 1:
The invention performs preliminary masking detection and evaluation before final template matching. By identifying masked blocks in advance and adjusting the matching strategy accordingly, the system prepares for potential accuracy issues rather than attempting verification after tracking fails. This preliminary action enables the system to handle wide masking and complicated behaviors more effectively.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An image processing apparatus for tracking a target object to be tracked included in a plurality of consecutively captured images comprises an image acquisition unit configured to acquire an image; a template acquisition unit configured to acquire a template corresponding to a target object to be tracked included in the image; a masking map generation unit configured to generate a masking map serving as a map indicative of a degree of masking the target object to be tracked, which is included in the image, by another object; and a tracking unit configured to detect the target object to be tracked included in the image by using the template, and acquire over a plurality of images an amount of movement of the target object to be tracked and a direction of the movement of the target object to be tracked, based on both of the template and the masking map.