Monocular Camera Distance Estimation via Object Area Division
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing distance estimation apparatuses face challenges in accurately calculating the distance from a host vehicle to an obstacle, especially when the road is inclined, and struggle to detect the contact area of an object with the road surface, leading to potential inaccuracies in distance measurement.
Innovation Solution
The proposed solution involves an information processing apparatus equipped with a monocular camera and a series of processing units that utilize area division technologies to distinguish individual objects within an image, even when they are overlapped or adjacent, by dividing the object area based on variations in the Y-axis direction, allowing for accurate distance estimation from the camera to each object.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If distance estimation is performed using a monocular camera and distance tables, then the system structure remains simple, but measurement precision deteriorates when roads are inclined
Solution Approach 1:
The image is divided into multiple regions of interest (ROIs) based on detected line segments. Each ROI corresponds to a specific area where distance estimation should be performed. This segmentation allows the system to focus computational resources on relevant areas and apply appropriate distance calculation methods for each region, improving overall measurement precision while maintaining system simplicity.
Solution Approach 2:
Line segment detection serves as an intermediary step between image capture and distance estimation. The detected line segments (representing road boundaries, curbs, or other linear features) act as mediators to identify ROIs and provide geometric constraints for accurate distance calculation. This intermediary processing enables the system to compensate for road inclination effects without requiring complex additional hardware.
2Loss of information
If multiple objects are present in the image, then the information processing becomes more complex, but the ability to distinguish individual objects improves
Solution Approach 1:
The image is divided into multiple discrete regions of interest, with each ROI associated with a specific object or feature. This segmentation is achieved by detecting line segments and using them to define boundaries between different objects. By processing each ROI separately, the system can accurately distinguish between multiple objects while keeping the processing complexity manageable through localized analysis.
Solution Approach 2:
Different processing methods and parameters are applied to different regions of interest based on their specific characteristics. Each ROI is analyzed with locally optimized parameters, allowing the system to handle diverse objects (vehicles, pedestrians, obstacles) with appropriate specialized processing. This local quality approach improves object distinction capability without requiring a single complex processing pipeline for all objects.
3Measurement precision
If the contact area of an object with the road surface is used for distance calculation, then distance estimation can be performed, but accuracy deteriorates when the object's contact point is difficult to detect
Solution Approach 1:
Line segment detection serves as an intermediary that indirectly identifies object positions without requiring direct contact point detection. The line segments (representing road features, curbs, or boundaries) provide reference frames from which object distances can be calculated. This mediator approach bypasses the difficulty of detecting contact points by using easily detectable linear features as reference.
Solution Approach 2:
Instead of directly detecting the contact point between object and road surface, the system detects line segments that represent or copy the geometric relationships in the scene. These line segments serve as proxies that encode the necessary spatial information for distance calculation, making the measurement process more robust when direct contact detection is difficult.
Data Source
Figure 1~2
Figure 3A~3G
Figure 4
AI summary
To provide an information processing apparatus and a program, which are capable of accurately estimating the distance from a monocular camera to an object. The information processing apparatus acquires an image captured by one monocular camera, determines an object area in the image, divides the object area based on variation in a predetermined direction at an end portion in the predetermined direction of the determined object area, and estimates a distance from the monocular camera to an object corresponding to at least one partial object area resulting from the division of the object area.