Video Person Detection With Multi-Resolution Alert Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video surveillance systems struggle with accurately identifying and categorizing meaningful segments of video streams and efficiently conveying this information to users, often resulting in excessive irrelevant alerts due to sensitivity issues and the need for manual review.
Innovation Solution
Implement methods for categorizing motion events using motion categories and confidence levels, sending alerts based on elapsed time and category consistency, and analyzing video frames to detect and confirm the presence of persons.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If motion detection sensitivity is set high to detect all potential events, then detection coverage is improved, but the quantity of irrelevant alerts increases
Solution Approach 1:
The patent segments the detection process into multiple stages: initial motion detection, person verification, and category classification. By dividing the alert generation process into discrete steps with filtering at each stage, the system maintains high sensitivity for detection while reducing the quantity of irrelevant alerts that reach the user.
Solution Approach 2:
The system performs preliminary verification actions before generating alerts. It pre-verifies detected motion events by analyzing video frames to confirm person presence and categorizing the event type. This preliminary action filters out false positives before they become alerts, resolving the contradiction between comprehensive detection and alert reduction.
2Quantity of substance
If motion detection sensitivity is set low to reduce false alerts, then the quantity of alerts is reduced, but detection coverage deteriorates
Solution Approach 1:
The system performs preliminary verification of all detected motion events through person detection algorithms and event categorization. This preliminary action ensures that even with low sensitivity settings, all relevant events are captured and verified before alert generation, maintaining detection coverage while reducing false alerts.
Solution Approach 2:
The system implements feedback mechanisms where detection results are continuously refined. By analyzing the characteristics of detected events and adjusting verification thresholds based on historical data, the system maintains optimal detection coverage regardless of the initial sensitivity setting, while still reducing irrelevant alerts through intelligent filtering.
3Reliability
If all motion events are alerted to ensure complete coverage, then detection coverage is improved, but user time to review events increases
Solution Approach 1:
The patent extracts and removes irrelevant information from the alert stream by implementing event categorization and person verification. Only verified person-related events are extracted for alert generation, while other motion events are filtered out. This reduces the volume of events requiring user review while maintaining complete coverage of relevant incidents.
Solution Approach 2:
The system applies different quality levels of verification to different types of events. High-priority events receive thorough verification and categorization, while low-priority events receive minimal processing. This local quality approach ensures complete coverage of important events while minimizing user review time by reducing noise from less critical events.
4Measurement precision
If manual review of all video segments is required to ensure accurate event identification, then measurement precision is improved, but productivity deteriorates
Solution Approach 1:
The system performs self-verification of detected events through automated person detection algorithms and event categorization. Instead of requiring manual review of all segments, the system autonomously verifies events and filters out false positives, maintaining high identification accuracy while dramatically improving review efficiency by eliminating the need for manual inspection of every detected event.
Solution Approach 2:
The system performs preliminary verification and categorization of all detected events before presenting them for review. This preliminary action pre-filters and pre-classifies events, so that when users do review events, they are already verified and organized. This maintains measurement precision while improving productivity by reducing the time users spend on basic verification tasks.
Data Source
AI summary
The various embodiments described herein include methods, devices, and systems for providing event alerts. In one aspect, a method includes obtaining a video feed. A frame of the video feed is analyzed at a first resolution to determine whether the frame includes a potential instance of a person. In accordance with the determination that the image includes the potential instance, a region is denoted around the potential instance. The region is analyzed at a second resolution, greater than the first resolution. In accordance with a determination that the region includes the instance of the person. a determination that the frame includes the person is made. An indication of the determination is stored for use in subsequent alert notification processing.