Content-Adaptive Blob Filtering for Video Analytics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video analytics systems face inefficiencies in blob detection and tracking due to noisy blobs generated by background elements, leading to reduced detection and tracking rates, especially when simple size-based filtering methods fail to distinguish between noise and actual objects.
Innovation Solution
The implementation of content-adaptive blob filtering techniques, which include a pre-filtering stage filtering blobs by size, a post-filtering stage filtering by size, height, and width, and an optional blob merging process to refine blob identification, based on adaptive thresholds and multipliers adjusted according to the number of blobs in a frame.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If simple size-based filtering is used to remove blobs, then processing speed is improved, but detection precision deteriorates because small relevant objects are incorrectly filtered out
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting the size threshold based on scene characteristics. Instead of using a fixed size threshold, the system adapts the threshold parameter according to the number of blobs detected and other scene-specific factors, allowing small but relevant objects to be retained while filtering out noise
Solution Approach 2:
The patent implements dynamics by transitioning from static size-based filtering to dynamic content-adaptive filtering. The filtering criteria change adaptively based on the current scene content, with the size threshold being adjusted in real-time according to the number of blobs and other contextual information
2Measurement precision
If no blob filtering is applied, then detection precision is improved by retaining all potential objects, but processing complexity increases due to handling noisy blobs
Solution Approach 1:
The patent applies local quality by implementing different filtering strategies for different types of blobs based on their characteristics. Rather than applying a uniform filtering approach, the system evaluates each blob's context and applies appropriate filtering, preserving relevant small objects while removing noise
Solution Approach 2:
The patent implements preliminary action by performing content-adaptive analysis before final blob filtering. The system预先 determines the appropriate size threshold and filtering parameters based on scene characteristics, preparing the filtering criteria in advance to balance precision and complexity
3Ease of operation
If fixed size threshold filtering is used, then ease of operation is improved, but adaptability deteriorates because the same threshold cannot handle varying scene conditions
Solution Approach 1:
The patent implements dynamics by making the size threshold adaptive rather than fixed. The threshold dynamically adjusts based on the number of blobs detected and other scene-specific parameters, allowing the same system to effectively handle varying scene conditions without manual reconfiguration
4Reliability
If aggressive blob filtering is applied to remove noise, then tracking reliability is improved, but loss of information occurs when small relevant objects are filtered out
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting the filtering threshold to prevent information loss. Instead of using a fixed aggressive threshold, the system adapts the threshold parameter based on scene characteristics, ensuring that small but relevant objects are not incorrectly filtered out while still maintaining noise removal
Data Source
AI summary
Techniques and systems are provided for processing video data. For example, techniques and systems are provided for performing content-adaptive blob filtering. A number of blobs generated for a video frame is determined. A size of a first blob from the blobs is determined, the first blob including pixels of at least a portion of a first foreground object in the video frame. The first blob is filtered from the plurality of blobs when the size of the first blob is less than a size threshold. The size threshold is determined based on the number of the plurality of blobs generated for the video frame.