Feature Point Speed Filtering for Accurate Vehicle Motion Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for calculating the movement amount of a moving object, such as robots and automobiles, face accuracy issues when three-dimensional objects are present in the camera image, leading to tracking failures and erroneous calculations due to height and movement influences.
Innovation Solution
A system that captures images of the road surface using an imaging device and processes them to extract feature points, track these points between frames, filter based on speed, and calculate the movement amount of the moving object using only feature points within a predetermined speed range, thereby enhancing accuracy and reducing processing load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the average value of movement amounts of a plurality of feature points is taken to calculate the movement amount of the moving object, then the processing can be performed with relatively simple calculation, but the accuracy cannot be ensured when three-dimensional objects are present in the camera image
Solution Approach 1:
The patent changes the selection criterion for feature points from using all extracted feature points to selecting only those with movement amounts within a predetermined range. This parameter-based filtering (speed threshold) eliminates the influence of three-dimensional objects while maintaining simple average calculation processing
Solution Approach 2:
The patent extracts and removes feature points that do not meet the predetermined speed range criterion from the set of all feature points. By taking out the problematic feature points (those influenced by three-dimensional objects), the calculation accuracy is improved without complex processing
2Reliability
If feature points from three-dimensional objects are included in the calculation, then more feature points are available for processing, but tracking failure and erroneous calculation occur due to height and movement influence
Solution Approach 1:
The patent introduces a speed range parameter as a filtering criterion to distinguish between feature points on the road surface and those on three-dimensional objects. By changing from using all feature points to using only those within the predetermined speed range, reliable tracking is achieved while reducing the number of problematic feature points
Solution Approach 2:
The patent applies different treatment to different feature points based on their movement characteristics. Feature points with movement amounts within the predetermined range are selected for calculation, while others are excluded, creating a locally optimized set of high-quality feature points for reliable tracking
Data Source
Figure 1
Figure 2
Figure 3~3(B)
AI summary
An object is to provide a moving object capable of being calculated a movement amount with a low processing load and high accuracy even when a three-dimensional object which is stationary or moves appears in a camera image. The invention is directed to a moving object which includes an imaging device which captures an image of a road surface and an image processing unit which calculates a movement amount of the moving object based on an image captured by the imaging device, wherein the image processing unit extracts a plurality of first feature points from a first image captured at a first timing, extracts a plurality of second feature points from a second image captured at a second timing after the first timing, performs tracking on each of the plurality of first feature points to each of the plurality of second feature points, calculates a movement amount and a speed of each of the plurality of second feature points, and calculates a movement amount of the moving object based on a feature point having a speed within a predetermined range among the plurality of second feature points.