Motion Vector Validation for Object Detection Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing systems face challenges in accurately detecting objects in motion due to incorrect global motion vector calculations caused by factors like large foreground movement, blur, parallax, illuminance changes, and sensor drift, which can result in unstable and variable output from motion sensors.
Innovation Solution
An image processing system and method that computes a first sensor offset based on both image-based and motion sensor-based motion vectors, using criteria to validate the first motion vector and correct it if invalid, thereby improving the accuracy of object extraction by utilizing the second motion vector when necessary.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If motion sensors are used to improve object detection accuracy, then measurement precision is improved, but device complexity increases due to additional sensors and processing requirements
Solution Approach 1:
The patent combines image-based motion vector computation with motion sensor-based motion vector computation into a unified global motion vector estimation system. The processor integrates data from both the image sensor and motion sensors (gyro sensor, accelerometer) to compute a more accurate global motion vector that compensates for the weaknesses of individual methods.
Solution Approach 2:
The patent introduces a high-pass filter as an intermediary component to process motion sensor output and remove drift components. The filter acts as a mediator between the raw motion sensor data and the final global motion vector computation, eliminating the need for additional complex hardware while maintaining accuracy.
2Measurement precision
If high pass filters are applied to motion sensor output to eliminate drift, then measurement precision is improved, but device complexity increases due to additional filtering hardware
Solution Approach 1:
The patent replaces physical filtering hardware with a computational high-pass filter implemented in software/processor. Instead of using additional analog filtering circuits or mechanical components, the system uses digital signal processing to remove drift from motion sensor output, significantly reducing hardware complexity while maintaining drift elimination effectiveness.
3Productivity
If global motion vector is calculated based on image frame comparison, then object detection is performed, but measurement precision deteriorates due to factors like blur, parallax, and illuminance change
Solution Approach 1:
The patent implements a feedback mechanism where the processor continuously compares image-based motion vectors with sensor-based motion vectors, computes their difference, and uses this feedback to correct the global motion vector. This closed-loop approach compensates for errors introduced by blur, parallax, and illuminance changes in the image-based method.
Solution Approach 2:
The patent creates a composite motion vector by combining information from multiple sources (image frame comparison and motion sensor data) into a unified global motion vector. This composite approach leverages the strengths of both methods while mitigating their individual weaknesses, similar to how composite materials combine properties of different materials to achieve superior performance.
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 is configured to compute a first sensor offset for a current frame based on a first motion vector and a second motion vector. A validation of the first motion vector is determined based on the second motion vector and one or more criteria. An object in motion from the current frame is extracted based on the first sensor offset of the current frame and the determined validation of the first motion vector.