Background Model Update via Similarity-Based Sample Replacement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing background subtraction methods face challenges in handling dynamic and multi-modal backgrounds, leading to increased false detections and high processing and memory requirements due to the need for large numbers of background samples.
Innovation Solution
A sample-based background model that updates by replacing background samples similar to the current pixel data, reducing the number of samples needed while maintaining long memory of modalities, thereby reducing processing and memory requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If many background samples are stored per pixel to be robust to dynamic and multi-modal backgrounds, then the model can handle dynamic backgrounds effectively, but processing and memory requirements become excessively high
Solution Approach 1:
The patent changes the parameter of background sample management by introducing a similarity-based selection criterion. Instead of storing all past background samples, the system selects samples based on their similarity to current pixel values, replacing only those most similar to current data. This parameter change enables effective handling of dynamic backgrounds with significantly reduced memory and processing requirements.
Solution Approach 2:
The patent extracts only the necessary background samples for storage and processing. By using similarity-based selection, the system extracts and stores only those background samples that are most representative of current conditions, discarding redundant samples. This extraction approach maintains robustness while reducing the number of samples needed, thereby lowering memory and processing requirements.
2Loss of information
If a large number of background samples are stored to maintain long memory of modalities, then the model can represent background variations accurately, but the processing time and memory usage increase significantly
Solution Approach 1:
The patent changes the parameter of sample selection by using similarity-based criteria. This allows the system to maintain long-term memory of modalities by selectively retaining only the most relevant background samples, reducing processing time while preserving essential information about background variations over time.
Solution Approach 2:
The patent extracts only the essential background samples that contribute to long-term modality memory. By extracting samples based on their representativeness and similarity to current conditions, the system maintains accurate long-term memory with reduced processing requirements compared to storing all historical samples.
3Device complexity
If background samples are updated using first-in-first-out principle, then the collection can be maintained efficiently, but the model cannot effectively handle sporadic background movements requiring long memory
Solution Approach 1:
The patent changes the updating parameter from fixed FIFO replacement to similarity-based selective replacement. This allows the system to handle sporadic background movements effectively by retaining background samples that are similar to current pixel values, even when they are not the oldest samples. The similarity criterion enables the model to maintain long-term memory of modalities while keeping the updating mechanism computationally efficient.
Data Source
AI summary
There is provided a method and an apparatus for updating a background model used for background subtraction of an image. An image is received and each pixel in the image is classified as being foreground or background by performing background subtraction using a background model which comprises a collection of background samples for each pixel in the image. For each pixel in the image being classified as background, the background model is updated by selecting a background sample in the collection of background samples associated with the pixel which fulfills a similarity condition, and replacing the background sample which fulfills the similarity condition with image data of the pixel.