Dynamic Detection Zone for Fall Event Identification in Video Surveillance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video analytics systems face challenges in accurately detecting falls in hospital environments due to dynamic furniture arrangements and the complexity of distinguishing between individuals lying on beds and those who have fallen, as traditional zone-based approaches are not adaptable to changing layouts.
Innovation Solution
A method that first detects a person and classifies their posture, then defines a detection zone around them to identify overlapping objects, using machine learning algorithms to determine if the person is lying on furniture, thereby eliminating the need for predefined zones and enhancing accuracy in fall detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional zone-based approaches are used for fall detection, then the system can cover the entire camera field of view, but the system cannot adapt to changing furniture arrangements and produces false alarms
Solution Approach 1:
The patent transitions from static predefined zones to dynamic detection zones that are automatically generated based on real-time detection of furniture objects and person postures. The detection zone adapts to changing furniture arrangements by continuously updating its boundaries according to the detected furniture positions and the lying person's location, thereby maintaining both adaptability and reliability.
Solution Approach 2:
The system performs preliminary detection and classification of furniture objects and person postures before defining the detection zone. By first identifying the furniture items and classifying the person's posture as lying, sitting, or standing, the system prepares the necessary information in advance to accurately define the detection zone boundaries, ensuring reliable fall detection even when furniture arrangements change.
2Measurement precision
If the field of view is divided into multiple fixed zones with algorithms, then movement in specific zones can trigger alarms, but the system cannot distinguish between persons lying on beds and those who have fallen
Solution Approach 1:
The patent extracts the detection zone definition from fixed predefined zones and makes it dynamic based on detected objects. Instead of managing multiple fixed zones with complex boundary definitions, the system extracts only the necessary detection zone around the lying person based on detected furniture boundaries, significantly simplifying zone management while improving the ability to distinguish falls from normal lying positions.
Solution Approach 2:
The patent introduces furniture objects as intermediaries to define detection zone boundaries. Rather than using fixed geometric zones, the system uses detected furniture items (beds, chairs, sofas) as natural boundaries to define where a person should be lying. This intermediary approach provides intuitive and accurate reference frames for distinguishing falls from normal positions without complex zone management.
3Measurement precision
If object detection algorithms are applied to the entire video frame, then all objects can be detected, but the processing time increases and accuracy decreases for specific objects
Solution Approach 1:
The patent segments the video frame processing by first detecting persons and classifying their postures, then only searching for furniture objects within a defined detection zone around the lying person. This segmentation of the search space from the entire frame to a localized region significantly reduces processing time while improving detection accuracy by concentrating computational resources on the relevant area where furniture objects are likely to be located.
Data Source
AI summary
A computer implemented method for identifying an event by processing video surveillance data from a video camera having a field of view uses a trained machine learning algorithm to detect a person in the video surveillance data and determine the location of the person in the field of view. A trained machine learning algorithm is used to classify a posture of the detected person and a detection zone within the field of view around the location of the detected person is defined. A trained machine learning algorithm is used to search for an object of a predetermined type overlapping the detection zone. An event is identified based on the posture detection and the result of the object detection.