Radar Point-Cloud Filtering for Dynamic Object Ego-Motion Outliers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional radar-based ego-motion estimation methods face reliability issues due to incomplete removal of dynamic objects, which affects the recognition of static objects, especially when the ratio of dynamic objects in radar data is higher than static objects.
Innovation Solution
A dynamic object filtering method and apparatus that employs a pre-filter and post-filter in addition to an outlier filtering algorithm, using a velocity model to classify points into candidate static and dynamic objects, enhancing the accuracy of static object recognition and ego-motion estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional outlier filtering algorithms (RANSAC or Cauchy norm) are used to remove dynamic objects, then the processing is simple and fast, but the dynamic objects are not completely removed when their ratio is higher than static objects, reducing reliability
Solution Approach 1:
The filtering process is divided into three distinct stages: pre-filtering (using position values from track module), outlier filtering (using RANSAC or Cauchy norm), and post-filtering (using velocity model). Each stage targets specific aspects of dynamic object removal, with pre-filtering handling obvious dynamic objects based on position tracking, outlier filtering addressing moderate cases, and post-filtering catching remaining dynamic objects using velocity characteristics. This segmentation allows comprehensive dynamic object removal while maintaining reasonable computational complexity.
2Reliability
If multiple filtering stages (pre-filter, outlier filter, post-filter) are used to completely remove dynamic objects, then the reliability of static object recognition is improved, but the device complexity and processing time increase
Solution Approach 1:
The pre-filter stage performs preliminary removal of dynamic objects using position values tracked from previous times before the main outlier filtering process. By eliminating obviously dynamic objects early based on their tracked position trajectories, the system reduces the burden on subsequent filtering stages and improves overall efficiency while maintaining reliability.
Solution Approach 2:
The track module continuously tracks position values of objects across multiple time points and feeds this information back to the pre-filter and post-filter stages. This feedback mechanism allows the system to identify dynamic objects based on their motion patterns over time, enabling more accurate separation of dynamic and static objects while adapting to changing scene conditions.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The proposed method robustly filters dynamic objects, improving the reliability of radar-based ego-motion estimation by accurately distinguishing between static and dynamic objects, even in environments with a higher ratio of dynamic objects, thereby enhancing the overall accuracy of ego-motion estimation.
Implementation Method 1
because radar sensors may measure a Doppler velocity, research for using radar sensors in ego-motion estimation is being done
Data Source
AI summary
A method of filtering dynamic objects in radar-based ego-motion estimation includes converting measurement value at current time, measured by radar sensor, into point cloud, classifying the point cloud into points of a first object predicted as static object and points of a second object predicted as dynamic object, based on position value of dynamic object tracked at previous time, classifying the points of the first object into the points of the static object predicted as normal value and the points of the dynamic object predicted as outlier, based on outlier filtering algorithm, classifying the points of the second object into points of a candidate static object and points of a candidate dynamic object, based on velocity model of the static object, and tracking a position value of the dynamic object at current time, based on the points of the dynamic object and the points of the candidate dynamic object.