Monocular Camera Distance Estimation via Feature Point Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Monocular camera-based distance estimation methods suffer from high error and reduced accuracy due to noise, calculation errors, and image acquisition timing issues, leading to slow convergence and reduced distance accuracy.
Innovation Solution
A distance estimation apparatus and method that sets feature points on images from a moving object's imaging device, tracks these points over time, determines the movement amount of both feature points and the object, and estimates distance based on these movements to reduce error influence and improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a monocular camera is used to detect surrounding information, then costs and processing complexity are reduced, but the distance to the feature point cannot be directly calculated
Solution Approach 1:
The patent changes the parameters used for distance estimation from direct geometric calculation (requiring stereo vision) to temporal parameter analysis. By utilizing the temporal sequence of feature point movements across multiple frames and correlating with vehicle velocity, the system achieves distance estimation with a monocular camera, resolving the contradiction between simplicity and measurement capability.
2Measurement precision
If monocular stereo method is used with particle filter or Kalman filter, then distance can be calculated from time-series information, but calculation convergence is slow and distance accuracy is reduced
Solution Approach 1:
The patent extracts only the essential parameters needed for distance calculation: feature point movement amount and vehicle velocity. By eliminating unnecessary calculation components of traditional filters and focusing solely on the correlation between these two parameters, the system achieves rapid convergence while maintaining accuracy, resolving the contradiction between accuracy and computation time.
Solution Approach 2:
The patent replaces complex iterative filtering mechanisms (particle filter, Kalman filter) with a direct mathematical calculation approach. By substituting the mechanical iterative filtering process with a straightforward correlation calculation between feature point movement and vehicle velocity, the system eliminates computation overhead while preserving measurement accuracy.
3Ease of operation
If feature point position contains error due to noise or calculation processing error, then distance estimation can still be performed, but error in distance becomes large and calculation may not converge
Solution Approach 1:
The patent incorporates feedback through the use of multiple sequential frames and iterative refinement of feature point tracking. By continuously comparing feature point movements across multiple frames and adjusting calculations based on accumulated data, the system compensates for individual frame errors and converges to accurate distance measurements even in noisy conditions.
Solution Approach 2:
The patent merges information from multiple sources: feature point movement data from sequential images and vehicle velocity data from the moving object. By combining these independent measurement streams, the system creates a more robust distance estimation that compensates for errors in individual measurements, resolving the contradiction between noise robustness and accuracy.
Data Source
AI summary
A distance to an object is estimated with a monocular camera that estimates a distance from a moving object to feature points on an image from an imaging device mounted on the moving object. The distance estimator sets one or more feature points on the image acquired from the imaging device at a first timing and detects the feature point on the image acquired from the imaging device at a second timing. The distance estimator also determines the movement amount of the feature point on the image between the first timing and the second timing and determines the movement amount of the moving object between the first and second timings. The distance estimator then estimates the distance from the moving object to the feature point based on the movement amount on the image and the movement amount of the moving object between the first and second timings.