Object Tracking Using Depth Information for Robot Positioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object tracking methods, such as those based on Kernelized correlation Filters (KCF), have high computational complexity and are not suitable for robots due to limited resources, leading to reduced tracking accuracy over extended periods.
Innovation Solution
A method utilizing depth information to track objects, where a robot captures images, calculates feature templates, and adjusts its position based on depth change ratios to maintain a relative position with the target object, reducing computational complexity and improving tracking accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If KCF-based tracking methods are used, then tracking functionality is achieved, but computational complexity is high and tracking accuracy reduces over time
Solution Approach 1:
The patent extracts and utilizes depth information from the image data to augment traditional 2D tracking. By incorporating the Z-dimension (depth) into the tracking process, the system creates a more robust tracking representation that improves accuracy without proportionally increasing computational complexity, as depth information can be obtained from standard RGB-D sensors.
Solution Approach 2:
The patent transitions from traditional 2D image tracking to 3D tracking by incorporating depth information. This dimensional expansion allows the system to track objects more accurately by considering their position in three-dimensional space, while the depth data from RGB-D sensors provides this additional dimension without requiring complex computational overhead.
2Duration of action of moving object
If traditional tracking methods are used, then object tracking is possible, but resource consumption is high for extended period tracking
Solution Approach 1:
The system uses the depth information inherently captured by RGB-D sensors alongside the image data. By utilizing information that is already being collected by the sensor (the depth map), the system avoids additional computational overhead for depth acquisition, making the tracking more resource-efficient for extended operation.
3Measurement precision
If 2D matching is performed without depth information, then computational simplicity is maintained, but tracking accuracy reduces as object moves
Solution Approach 1:
The patent merges 2D image matching with depth information processing. By combining the visual appearance information from 2D images with spatial depth information, the system creates a more comprehensive tracking representation that maintains algorithmic efficiency while significantly improving tracking accuracy during extended object movement.
Data Source
AI summary
A technology for tracking a target object using depth information is disclosed. An object image contain a target object can be obtained from a reference image. A target distance from a robot to the target object can also be derived or accessed. When the robot captures a new image, 2D matching can be performed to find the horizontal and vertical movement of the target object. Distance change or depth change of the target object can also be calculated from the new image. A depth change ratio can be calculated and be utilized to scale a bounding box. The robot can then move to a new location according to the horizontal, vertical and depth changes. A new object image can be extracted from the new image according to the scaled bounding box. The object tracking can be continued using a similar process when the next image is captured.