Target Object Estimation Using Road Surface Region Exclusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing target object estimating apparatuses face increased computational load due to the need to calculate estimated values for almost all pixels in high-resolution stereo camera images, leading to inefficiencies and potential errors in estimating the position and speed of target objects.
Innovation Solution
The apparatus detects road surface regions in the image and excludes these areas from calculations, only estimating values for pixels outside the road surface region, thereby reducing the computational load and preventing errors in target object estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If estimated values are calculated for all pixels in high-resolution stereo camera images, then the position and speed of target objects can be estimated, but the computational load increases significantly
Solution Approach 1:
The image is segmented into road surface regions and non-road surface regions based on parallax information. Estimation calculations are then performed only on non-road surface regions where target objects are likely to exist, rather than processing all pixels in the high-resolution image. This segmentation approach maintains estimation accuracy for relevant areas while reducing overall computational load.
Solution Approach 2:
Different processing approaches are applied to different regions of the image. Road surface regions use parallax-based detection to identify the road area, while non-road surface regions undergo full estimation calculation for target objects. This local differentiation ensures accurate target detection where needed while avoiding unnecessary computations on the road surface.
2Measurement precision
If the resolution of the stereo camera is increased to improve image quality, then the detail of target objects can be captured, but the number of pixels increases causing higher computational load
Solution Approach 1:
The road surface region is extracted from the high-resolution image using parallax information and road surface detection algorithms. By removing the road surface area from the processing scope, the number of pixels requiring estimation calculations is significantly reduced, thereby lowering computational complexity while preserving full image quality for target object detection.
3Reliability
If estimation calculations are performed for all pixels to ensure no target object is missed, then estimation leakage is prevented, but the processing time increases
Solution Approach 1:
Road surface regions are detected and identified in advance using parallax information before the estimation calculation phase. This preliminary identification of areas where target objects cannot exist allows the system to skip estimation calculations on the road surface, reducing processing time while maintaining detection completeness for actual target objects.
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
This approach effectively reduces the computational load while ensuring accurate estimation of target object positions and speeds by focusing calculations on regions where target objects are likely to exist, thereby preventing estimation leakage.
Implementation Method 1
calculates a parallax of each of pixels in the taken image
Implementation Method 2
calculates a change amount (an optical flow) of a position of each of the pixels
Data Source
AI summary
A target object estimating apparatus comprises parallax information calculating means for calculating parallax information including a parallax between corresponding pixels of taken images taken at a same computation timing, road surface region detecting means for detecting a road surface region in the taken image based on the parallax information, road surface region determining means for determining whether or not each pixel in the taken image corresponds to the road surface region, transition information calculating means for calculating transition information including a change amount of a position of each pixel, using temporally sequential taken images, and estimated value calculating means for estimating a position and a speed of a target object by calculating, based on the parallax/transition information, estimated values of the position and the speed of each pixel. The estimated value calculating means does not calculate the estimated values for a pixel determined to correspond to the road surface region.