Motion Vector Validation Using Sensor Data for Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing systems for detecting objects in motion often fail due to incorrect calculation of global motion vectors caused by factors like large foreground movement, blur, parallax, illuminance changes, and roller shutter effects, leading to inaccurate object detection.
Innovation Solution
An image processing system and method that computes a first motion vector based on pixel differences between current and previous frames, and a second motion vector using sensor information, with validation criteria to determine the validity of the first motion vector, allowing for accurate extraction of objects in motion by switching to sensor-based vectors when necessary.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If global motion vector is calculated based on pixel difference between current and previous frames, then object detection can be performed, but detection accuracy deteriorates due to large foreground movement, blur, parallax, illuminance change, and roller shutter effect
Solution Approach 1:
The patent divides the global motion vector calculation into two segments: image-based motion vector from pixel differences and sensor-based motion vector from motion sensor data. Each segment handles different aspects of motion detection, with the sensor-based approach compensating for the weaknesses of the image-based approach under challenging conditions.
Solution Approach 2:
The motion sensor acts as an intermediary that provides alternative motion information when image-based calculation fails. The sensor-based motion vector serves as a mediator to correct or replace inaccurate image-based vectors, particularly during large foreground movement, blur, or illuminance changes.
2Manufacturing precision
If image-based global motion vector is used for object extraction, then extraction can be performed, but extraction accuracy deteriorates when motion vector calculation is incorrect
Solution Approach 1:
The patent implements a feedback mechanism where the system validates the image-based motion vector by comparing it with sensor-based motion vector. When the image-based vector fails validation (indicating incorrect calculation), the system switches to using the sensor-based vector, providing corrective feedback to maintain extraction accuracy.
Solution Approach 2:
The system dynamically switches between image-based and sensor-based motion vectors based on validation results. This dynamic adaptation allows the system to maintain high extraction accuracy under varying conditions by selecting the most reliable motion vector source for each situation.
3Measurement precision
If sensor-based motion vector is used to replace invalid image-based vector, then detection accuracy improves, but device complexity increases due to additional sensor and validation logic
Solution Approach 1:
The motion sensor serves multiple functions: it provides alternative motion vector calculation, validates image-based vectors, and compensates for various error conditions (large foreground movement, blur, parallax, illuminance change, roller shutter effect). This multi-functionality justifies the additional hardware by consolidating multiple correction capabilities into a single sensor.
Data Source
AI summary
Various aspects of a system and a method are provided for detection of objects in motion are disclosed herein. In accordance with an embodiment, the system includes an electronic device, which computes a first motion vector based on a difference of pixel values of one or more pixels in a current frame with respect to a previous frame. The current frame and the previous frame comprises at least an object in motion. The system further computes a second motion vector by use of motion-based information obtained from a sensor in the electronic device. The system determines validation of the first motion vector based on the second motion vector and one or more criteria and extracts the object in motion from the current frame based on the determined validation of the first motion vector.