Object Tracking Device Using KCF and Bird's-Eye View Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional object tracking systems face challenges in maintaining accurate tracking of objects near a vehicle, especially when the vehicle is in motion, due to changes in object appearance and movement, which can lead to inaccurate tracking.
Innovation Solution
An object tracking device that includes an image acquirer, a recognizer, an area setter, and an object tracker, which uses Kernelized Correlation Filter (KCF) to track objects by estimating future positions and sizes based on past changes and behavior information, and adjusts the image area size to avoid contact, projecting images into a bird's-eye view for enhanced tracking accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If object tracking is performed on images captured by a mobile imager, then the system can track objects in real-time, but tracking accuracy deteriorates due to changes in object appearance and movement
Solution Approach 1:
The area setter predicts the future image area where the object will be located by estimating object movement based on past changes and vehicle behavior information, before the actual tracking occurs in the next image frame. This preliminary prediction allows the tracking system to prepare the search area in advance, improving tracking accuracy despite the mobile imager's changing perspective
Solution Approach 2:
The system dynamically adjusts the image area position and size based on real-time analysis of object movement patterns and vehicle behavior. The area setter continuously updates the predicted image area according to the object's past movements and the vehicle's current state, allowing the tracker to adapt to changing appearance and position while maintaining high accuracy
2Reliability
If the image area size is increased to avoid contact with objects, then safety is improved, but tracking precision deteriorates due to reduced focus on the object
Solution Approach 1:
The area setter dynamically adjusts the image area size based on the vehicle's distance to the object and the object's movement characteristics. When the vehicle is close to the object or moving rapidly, the area is enlarged to ensure contact avoidance. When the vehicle is distant or moving slowly, the area is reduced to maintain tracking precision. This dynamic adjustment resolves the contradiction between safety and precision
Data Source
AI summary
An object tracking device according to embodiments includes an image acquirer configured to acquire image data including a plurality of image frames captured in time series by an imager mounted on a mobile object, a recognizer configured to recognize an object from image data acquired by the image acquirer, an area setter configured to set an image area including an object recognized by the recognizer, and an object tracker configured to track the object on the basis of an amount of time-series change in an image area set by the area setter, in which the area setter sets a position and a size of an image area for tracking the object in the future image frame on the basis of the amount of time-series change in an image area including the object in the past image frame and behavior information of the mobile object.