Static Occlusion Handling via Directional Pixel Replication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video tracking algorithms face challenges in maintaining object tracking when static occlusions occur, such as objects being occluded by semi-permanent objects like trees or signs, leading to loss of track due to mismatched appearance and reliance on computationally expensive search windows and morphological processing.
Innovation Solution
A method and system that replicate pixels from neighboring regions along the predicted motion direction of the tracked object to replace occlusion pixels, allowing continuous tracking by estimating the approaching direction of the object relative to the occlusion and using dynamic pixel replication algorithms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional tracking algorithms use feature matching and search/optimization algorithms to track objects frame-by-frame, then tracking can be performed with standard computational resources, but the track is lost when objects are occluded by static occlusions due to appearance mismatch
Solution Approach 1:
The system performs preliminary action by replicating pixels from previous frames before the occlusion event occurs. The pixel replication is prepared in advance based on the predicted motion trajectory, so when occlusion happens, the replicated pixels are already positioned to maintain the track continuity without requiring complex real-time search algorithms.
Solution Approach 2:
The invention applies copying by replicating pixels from the tracked object in previous frames to current and future frames. This creates virtual copies of the object's appearance that can be used for tracking even when the actual object is occluded, maintaining the track through the occlusion event without losing information.
2Productivity
If Kalman filter method is used to estimate future object locations based on past trajectory, then computational resources are reduced, but the tracked object is lost when actual occlusion time differs from filter prediction
Solution Approach 1:
Instead of relying solely on predictive filtering, the system copies actual pixel data from previous frames where the object was visible. This provides concrete visual evidence of the object's appearance and position, which is more reliable than predictions when occlusion timing deviates from expected patterns.
Solution Approach 2:
The pixel replication acts as an intermediary between the tracked object and the tracking algorithm. Rather than directly tracking the occluded object or relying purely on prediction, the replicated pixels serve as a bridge that maintains the visual connection through the occlusion event.
3Reliability
If other approaches analyze blob deformations and use search windows with morphological processing to recover lost tracks, then track recovery is attempted, but computational cost increases significantly
Solution Approach 1:
The system copies pixels directly from previous frames where the object was visible, avoiding the need for complex morphological processing and search window analysis. This direct copying approach maintains track recovery capability while significantly reducing computational complexity compared to analyzing blob deformations and using morphological operations.
Data Source
AI summary
This disclosure provides a static occlusion handling method and system for use with appearance-based video tracking algorithms where static occlusions are present. The method and system assumes that the objects to be tracked move in according to structured motion patterns within a scene, such as vehicles moving along a roadway. A primary concept is to replicate pixels associated with the tracked object from previous frames to current or future frames when the tracked object coincides with a static occlusion, where the predicted motion of the tracked object is a basis for replication of the pixels.