Salient Event Detection in Video Analytics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current visual analytics frameworks face inefficiencies due to uneven task distribution between edge and cloud computing, leading to high costs, resource wastage, and difficulty in adapting to changing business logic, especially in video stream processing where most data is uninteresting and requires continuous processing by expensive sensors.
Innovation Solution
A system and method that focuses on salient event detection within defined areas of interest, using a combination of low-cost cameras and on-demand activation of additional sensors for supplemental data, reducing unnecessary processing and transmission of only relevant video streams for further analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If continuous processing of video streams by expensive sensors is implemented, then detection reliability is improved, but computational cost and energy consumption increase
Solution Approach 1:
The system performs preliminary actions by pre-defining areas of interest and pre-processing video streams to identify salient events before full analysis. This allows the system to prepare detection frameworks in advance and activate expensive processing only when salient events are detected, reducing continuous computational cost while maintaining detection reliability for important events.
Solution Approach 2:
The system implements periodic action by continuously monitoring video streams for salient events using low-cost sensors, then periodically activating expensive sensors and processing only when needed. This periodic activation pattern maintains detection reliability for salient events while significantly reducing average computational cost compared to continuous processing.
2Loss of time
If more visual analytics tasks are processed on the edge side, then latency is reduced and sensitive data remains on site, but device complexity and cost increase
Solution Approach 1:
The system segments visual analytics tasks into two categories: simple tasks processed locally on edge devices and complex tasks processed on the cloud. This segmentation allows low-cost edge devices to handle latency-sensitive simple tasks while maintaining data sensitivity, avoiding the need for expensive complex edge devices for all tasks.
Solution Approach 2:
The system introduces an intermediary cloud processing layer that receives video streams from edge devices, identifies salient events, and returns processed results. This intermediary allows edge devices to remain simple and cost-effective while still achieving low latency for local processing, as the cloud handles complex analysis only when needed.
3Productivity
If visual data is organized following a data model, then search efficiency is improved, but adaptability to changing business logic decreases
Solution Approach 1:
The system implements dynamics by using dynamic filtering mechanisms that adapt to changing business logic requirements. Instead of a static data model, the system dynamically adjusts which video streams are processed and which features are extracted based on current business needs, maintaining search efficiency while enabling flexible adaptation.
Solution Approach 2:
The system applies parameter changes by modifying detection parameters and filtering criteria based on changing business logic rather than restructuring the entire data model. This allows the system to maintain efficient search performance while adapting to new business requirements through parameter adjustments rather than schema changes.
4Adaptability or versatility
If simple flat video data is used, then adaptability to changing business logic is improved, but search efficiency decreases
Solution Approach 1:
The system performs preliminary action by pre-processing video data to identify and extract salient events and features before storage. This preliminary organization maintains adaptability to changing business logic while improving search efficiency, as the system can quickly query pre-identified salient events rather than searching through all flat video data.
Solution Approach 2:
The system extracts salient events and important features from flat video data and separates them into a structured representation. This extraction maintains the adaptability of flat data structures while creating an efficient search index of salient events, resolving the contradiction between simplicity and search efficiency.
Data Source
AI summary
The disclosure includes a system and method for providing visual analysis focalized on a salient event. A video processing application receives a data stream from a capture device, determines an area of interest over an imaging area of the capture device, detects a salient event from the data stream, determines whether a location of the detected salient event is within the area of interest, and in response to the location of the salient event being within the area of interest, identifies a portion of the data stream, based on the salient event, on which to perform an action.