Video Region Identification Using Motion-Based Resolution Scaling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional client devices require manual user input to configure and adjust video capture settings, limiting their ability to automatically adapt to changing conditions and accurately identify moving objects within video frames.
Innovation Solution
A client device with an image capture device and a controller that applies machine learning models to differentiate between foreground and background in video frames, using a two-stage process to identify and prioritize regions of interest, allowing for automatic adjustment of video capture parameters and enhanced object tracking.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If manual user input is required to configure video capture settings, then user control over video data capture is maintained, but the system cannot automatically adapt to changing conditions or accurately identify moving objects
Solution Approach 1:
The system automatically captures video data and applies machine learning models to identify regions of interest and moving objects without requiring manual user input for configuration. The controller autonomously processes video frames, detects motion, and adjusts capture parameters based on detected objects, enabling the system to serve itself rather than requiring continuous user direction.
Solution Approach 2:
The system dynamically changes video capture parameters such as focal point and magnification based on detected moving objects and regions of interest. The controller modifies these parameters in real-time during video capture to optimize the view of detected objects, transitioning from static manual configuration to dynamic automatic adjustment.
2Measurement precision
If the system processes entire video frames to identify moving objects, then detection accuracy is improved, but computational resources and processing time increase
Solution Approach 1:
The system divides the video frame into multiple regions and processes only those regions that show motion or contain detected objects of interest. Instead of analyzing every pixel of every frame, the controller identifies and processes specific regions where moving objects are detected, reducing overall computational load while maintaining detection accuracy.
Solution Approach 2:
The system applies machine learning models to identify regions of interest and then processes only those partial regions in detail, rather than processing the entire video frame. This partial action approach reduces computational resources while maintaining sufficient accuracy for object identification and tracking.
3Measurement precision
If the image capture device continuously captures video data at high resolution, then object tracking accuracy is improved, but data processing load and storage requirements increase
Solution Approach 1:
The system dynamically adjusts the resolution and processing intensity based on detected motion and objects of interest. When no moving objects are detected, the system processes data at lower intensity; when objects are detected, it increases processing resolution for those specific regions, creating a dynamic adaptive system rather than continuous fixed-high resolution processing.
Solution Approach 2:
The system applies different processing qualities to different regions of the video data. High-resolution processing is applied locally to regions containing detected moving objects, while lower processing is applied to background regions. This local quality differentiation maintains object tracking accuracy for important regions while reducing overall processing load.
Data Source
AI summary
Multiple users communicate over a network via client devices that include one or more cameras and a display to enable video messaging. At least one of the client devices modifies regions of video data captured by the client device's camera to more prominently identify the people within the video data. To identify a person, the client device disambiguates between actual people and static objects that may appear like people. The client device uses pose models to identify bounding boxes and applies a motion model to determine if a bounding box may include a person based on an amount of movement within the bounding box. If a threshold amount of movement is detected in a bounding box, the client device obtains a higher resolution portion of the scene including the bounding box and classifies whether the bounding box contains a person based on movement within the higher resolution video.