Eccentricity Map Encoding for Video Motion Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Convolutional Neural Networks (CNNs) face difficulties in detecting speed and direction of motion between frames in video data, and traditional computer vision techniques require human expertise and are not robust for real-world applications.
Innovation Solution
The generation of an eccentricity map, which aggregates apparent motion from video frames, is used as input for CNNs, allowing for efficient and effective motion detection with minimal user-defined parameters, and enabling real-time processing and handling of concept drift and evolution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If CNNs are used for per-frame object identification, then object detection accuracy is improved, but motion detection capability deteriorates
Solution Approach 1:
The system segments motion detection from object identification by introducing a separate optical flow analysis module that processes temporal information independently, while CNNs handle spatial object recognition. This segmentation allows each component to specialize: CNNs for accurate object detection and optical flow for motion detection, resolving the contradiction between the two capabilities.
Solution Approach 2:
The patent introduces optical flow fields as an intermediary representation that bridges the gap between static frame analysis and motion detection. The optical flow module computes motion vectors between frames and feeds this temporal information to the classification system, enabling motion detection without compromising CNN object identification accuracy.
2Measurement precision
If traditional computer vision techniques are used for motion detection, then motion detection accuracy is improved, but system complexity and expertise requirements increase
Solution Approach 1:
The patent replaces complex traditional computer vision feature engineering with a streamlined optical flow computation pipeline. Instead of hand-crafting features requiring expert knowledge, the system uses automated optical flow algorithms that compute motion vectors directly from pixel intensities, reducing system complexity while maintaining motion detection accuracy through integration with CNN-based classification.
3Adaptability or versatility
If traditional computer vision techniques are used for motion analysis, then motion detection capability is improved, but robustness to real-world variations deteriorates
Solution Approach 1:
The system dynamically adjusts optical flow computation parameters and CNN classification thresholds based on scene characteristics and motion patterns. By adapting parameters such as flow field resolution, temporal window size, and classification confidence thresholds, the system maintains robustness across diverse real-world conditions while preserving motion detection capability.
Solution Approach 2:
The patent implements feedback mechanisms where classification results and detection confidence levels are used to adjust optical flow computation intensity and parameter selection. When motion is detected, the system increases optical flow analysis resolution; when scenes are static, it reduces computation, thereby maintaining robustness while adapting to varying real-world conditions.
Data Source
AI summary
The present invention extends to methods, systems, and computer program products for classifying time series image data. Aspects of the invention include encoding motion information from video frames in an eccentricity map. An eccentricity map is essentially a static image that aggregates apparent motion of objects, surfaces, and edges, from a plurality of video frames. In general, eccentricity reflects how different a data point is from the past readings of the same set of variables. Neural networks can be trained to detect and classify actions in videos from eccentricity maps. Eccentricity maps can be provided to a neural network as input. Output from the neural network can indicate if detected motion in a video is or is not classified as an action, such as, for example, a hand gesture.