Optical Flow Target Tracking with Edge Detection and Error Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video surveillance systems in coal mines face challenges in accurately and stably tracking moving objects due to low illumination, noise, and uneven background conditions, leading to incomplete target extraction and incorrect tracking.
Innovation Solution
A method and system that utilize edge-detection and optical flow techniques, combined with a SUSAN corner detection algorithm and forward-backward error algorithm, to generate a complete moving target template for accurate tracking, while adapting to environmental changes through variable coefficient template updating.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the optical flow method is used to extract target information, then target extraction can be achieved, but the target contour is incomplete
Solution Approach 1:
The patent combines optical flow method with edge detection method to extract target information. The optical flow provides motion vector fields while edge detection provides contour information, and their fusion results in complete target contours that resolve the incompleteness issue of using optical flow alone
2Productivity
If the optical flow method is used for next frame estimation, then target tracking can be performed, but noise causes wrong matching leading to incorrect target extraction
Solution Approach 1:
The patent introduces a forward-backward error verification mechanism where the optical flow estimation is checked by comparing forward and backward tracking errors. This feedback mechanism identifies and eliminates false matching points caused by noise, thereby improving tracking reliability
Solution Approach 2:
The patent uses edge detection information as an intermediary to verify and correct optical flow estimation results. The edge information acts as a mediator to filter out false matches caused by noise while preserving true target movements
3Reliability
If traditional video surveillance systems are used with operators observing video recordings, then potential dangers can be detected, but tremendous human and material resources are required
Solution Approach 1:
The patent implements automated target extraction and tracking systems that perform safety monitoring without human operators. The system automatically detects moving targets, tracks their trajectories, and identifies potential dangers, making the system self-serving and eliminating the need for tremendous human resources
Data Source
AI summary
A method for tracking a moving target based on an optical flow method, including, providing video images, and implementing pre-processing of the images to generate pre-processed images; implementing edge-detection of the pre-processed images and using an optical flow method to extract target information from the pre-processed images, and on the basis of a combination of the edge-detection information and the extracted target information, generating a complete moving target; using an optical flow method to perform estimation analysis of the moving target and using a forward-backward error algorithm based on feature point trace to eliminate light-generated false matching points; and creating a template image and implementing template image matching to track the moving target. The method and system for tracking a moving target based on an optical flow method have the advantages of accurate and complete extraction and the ability to implement stable tracking over a long period of time.