Video Activity Detection via Frame Segmentation and Binary Optimization

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current automated video surveillance systems are inefficient in accurately detecting and distinguishing human activities of concern, such as theft or compliance violations, due to high false alert rates and limited scalability, especially in environments with visual clutter and varying lighting conditions.

Innovation Solution

A method that segments video streams into discrete frame image primitives, combines them into visual events using hypotheses, and optimizes these events with binary variables based on constraints, associating them with non-video transaction data to issue alerts only when certain conditions are met, thereby reducing false alarms and improving detection accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If automated video analysis systems are implemented to detect human activities, then productivity is improved by reducing manual review time, but measurement precision deteriorates due to high false alert rates and missed event recognitions

Engineering Contradiction:

Improvevideo analysis efficiencyVSAvoidactivity detection accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The video stream is segmented into discrete frame image primitives, which are then combined into visual events. This segmentation allows the system to process video data in manageable units, improving both efficiency and detection accuracy by focusing on specific visual patterns rather than analyzing entire video streams manually.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

A binary variable acts as an intermediary between visual event detection and alert generation. This intermediary layer processes the detected events through optimization constraints before issuing alerts, reducing false positives by filtering out spurious detections while maintaining true event recognition.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Loss of time

If automated analysis algorithms are used to distinguish human activities, then loss of time is reduced compared to human review, but reliability deteriorates due to false alerts and missed event recognitions

Engineering Contradiction:

Improvereview timeVSAvoiddetection reliability

Core Design Contradiction:

Loss of timeVSReliability

Solution Approach 1:

The system incorporates feedback loops where detected visual events are optimized against constraints and compared against logged transactions. This feedback mechanism continuously refines detection accuracy by learning from past events and adjusting detection thresholds, thereby improving reliability while maintaining automated processing speed.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

Visual events are pre-processed and optimized against constraints before final alert generation. This preliminary action filters out potential false positives early in the processing pipeline, ensuring that only high-confidence events trigger alerts, thus improving reliability without sacrificing processing speed.

Inventive Principle:

Principle #10Preliminary action

3Measurement precision

If visual events are optimized using binary variables and constraints, then measurement precision is improved by reducing false alerts, but device complexity increases

Engineering Contradiction:

Improveevent detection accuracyVSAvoidsystem complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system changes parameters by introducing binary variables that represent the presence or absence of specific visual event characteristics. These binary parameters simplify the optimization process by converting continuous visual data into discrete decision variables, improving detection accuracy while managing computational complexity through efficient binary optimization algorithms.

Inventive Principle:

Principle #35Parameter changes

4Loss of information

If the system issues alerts based on binary variable optimization, then loss of information is reduced by filtering out false positives, but productivity deteriorates due to stricter processing constraints

Engineering Contradiction:

Improvefalse alert reductionVSAvoidprocessing throughput

Core Design Contradiction:

Loss of informationVSProductivity

Solution Approach 1:

The system applies partial optimization by focusing computational resources on optimizing only those visual events that meet initial detection thresholds. Not all detected events undergo full optimization processing, which maintains processing throughput while still filtering out false positives through selective application of the binary variable optimization framework.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS9471832B2Human activity determination from video

Publication Date: 2016.10.18 KYNDRYL INC

AI summary

Automated analysis of video data for determination of human behavior includes segmenting a video stream into a plurality of discrete individual frame image primitives which are combined into a visual event that may encompass an activity of concern as a function of a hypothesis. The visual event is optimized by setting a binary variable to true or false as a function of one or more constraints. The visual event is processed in view of associated non-video transaction data and the binary variable by associating the visual event with a logged transaction if associable, issuing an alert if the binary variable is true and the visual event is not associable with the logged transaction, and dropping the visual event if the binary variable is false and the visual event is not associable.