Video Surveillance Parallel Processing via GOP Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Modern video surveillance systems face challenges in processing video analytics in real-time due to resource-intensive tasks, particularly with deep learning requirements, which can lead to prolonged processing times for tasks like face recognition and video analysis, impacting incident response and forensic investigations.
Innovation Solution
Implementing a video surveillance system that utilizes parallel processing across multiple computer-implemented digital devices, allowing for simultaneous processing of video streams and metadata using GPUs and CPUs, enabling faster analytics results by distributing processing tasks across multiple cores and devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If sequential processing is used for video analytics, then processing complexity is manageable, but processing time increases linearly with complexity causing loss of time
Solution Approach 1:
The patent segments video processing into separate encoding and decoding stages, with encoded video data stored in groups of pictures (GOPs) that can be independently processed. This segmentation allows parallel processing of multiple GOPs simultaneously, reducing overall processing time while maintaining manageable complexity for each individual processing stage.
Solution Approach 2:
The patent introduces parallel processing across multiple processors or processing cores, transitioning from single-thread sequential processing to multi-thread parallel processing. This dimensional change in processing architecture enables simultaneous execution of multiple video analytics tasks, dramatically reducing processing time without exponentially increasing individual task complexity.
2Measurement precision
If deep learning analytics are performed on video data, then analysis accuracy is improved, but resource consumption increases significantly
Solution Approach 1:
The patent performs preliminary encoding of video data into compressed GOPs with embedded metadata before analytics processing. This preliminary action reduces the data volume requiring deep learning analysis, allowing accurate analytics to be performed on compressed representations rather than full-resolution video, thereby reducing computational resources while maintaining analysis accuracy.
Solution Approach 2:
The patent extracts and processes only relevant metadata and key features from video data using deep learning, rather than analyzing entire video frames. This extraction approach focuses computational resources on critical information elements, improving analysis accuracy for specific tasks while significantly reducing overall resource consumption.
3Speed
If face recognition and video analytics are performed in real-time, then incident response is improved, but processing time requirements create system bottlenecks
Solution Approach 1:
The patent segments video streams into independent GOPs that can be processed in parallel by multiple workers. Each GOP contains self-contained video data with metadata, allowing independent real-time analytics processing without waiting for sequential completion of previous frames, thereby maintaining high incident response speed while maximizing processing throughput.
Solution Approach 2:
The patent combines encoding and analytics processing into an integrated pipeline where video encoding and analytics operations are merged and executed concurrently. This merging eliminates sequential bottlenecks by allowing both functions to operate simultaneously on different GOPs, improving both real-time response capability and overall system throughput.
Data Source
AI summary
A video surveillance system comprising a video management system, one or more computer-implemented digital devices and a plurality of application programs, the system being configured to run the programs on one of or across several of the said one or more computer-implemented digital devices, and configured to perform parallel processing of one or more groups of pictures from one or more video streams with the plurality of application programs in one or across several of the said one or more computer-implemented digital devices.