Virtual Stream Centering for Fisheye Motion Monitoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Surveillance systems using fisheye cameras face high processing power requirements due to the need to extract multiple rectilinear video streams from the fisheye feed for AI monitoring, which is costly and inefficient.
Innovation Solution
A method and system that dynamically generates virtual rectilinear streams from fisheye video streams, allowing for pan and tilt adjustments to center the motion of interest, reducing the need for continuous monitoring of all streams and lowering processing demands by using low-cost CPUs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If multiple rectilinear video streams are extracted from fisheye feed to cover surveillance area, then area coverage is improved, but processing power requirement increases
Solution Approach 1:
The patent divides the fisheye video stream into multiple virtual rectilinear streams, each covering a specific area of interest. This segmentation allows the system to process only relevant portions of the surveillance area independently, reducing the overall processing burden while maintaining comprehensive coverage.
Solution Approach 2:
The patent implements dynamic pan-tilt adjustments for virtual streams based on detected motion. When motion is detected in a particular area, the corresponding virtual stream is adjusted to center on that motion, dynamically reallocating processing focus to areas requiring monitoring while reducing processing of static areas.
2Measurement precision
If GPU-based AI systems are used for real-time video monitoring, then detection accuracy is improved, but system cost increases
Solution Approach 1:
The patent segments the video processing into multiple virtual streams that can be handled by lower-cost CPU resources. By dividing the processing load across multiple independent virtual streams rather than requiring a single high-power GPU system, the solution achieves comparable detection accuracy at reduced cost.
Solution Approach 2:
The patent replaces GPU-based processing with CPU-based processing for video monitoring tasks. By implementing motion detection and tracking algorithms on standard CPU hardware rather than requiring dedicated GPU acceleration, the system achieves acceptable detection accuracy at significantly lower system cost.
3Area of stationary object
If fisheye cameras are used to cover wider area, then area coverage is improved, but video stream processing complexity increases
Solution Approach 1:
The patent converts the complex fisheye video stream into multiple simpler virtual rectilinear streams, each representing a specific viewing angle or area. This segmentation transforms the complex single-stream processing into multiple simpler parallel processing tasks, reducing overall processing complexity.
Solution Approach 2:
The patent introduces virtual streams as an intermediary layer between the fisheye camera and the AI monitoring system. These virtual streams act as mediators that translate the complex fisheye format into multiple standard rectilinear formats, simplifying the interface with downstream processing systems.
4Reliability
If continuous monitoring of all video streams is performed, then detection reliability is improved, but processing power consumption increases
Solution Approach 1:
The patent implements dynamic adjustment of virtual stream processing based on detected motion. When motion is detected, processing resources are dynamically allocated to focus on areas with motion activity. When no motion is present, processing power consumption is reduced by minimizing monitoring of static areas, maintaining detection reliability only where needed.
Solution Approach 2:
The patent extracts and focuses processing resources on areas containing motion of interest, rather than continuously processing all video streams uniformly. By taking out only the relevant portions requiring monitoring and applying processing power selectively, the system maintains detection reliability while reducing overall power consumption.
Data Source
AI summary
Present disclosure provides methods and apparatuses for monitoring motion in a video stream. The method comprises: identifying a motion of interest from motion present in a virtual stream, the virtual stream covering an area of the video stream at which motion is present, the motion of interest being a motion that is targeted for monitoring, wherein the area of the video stream being covered is controllable by a pan tilt value of the virtual stream; and adjusting the pan tilt value of the virtual stream to position the motion of interest at a centre of the virtual stream.