Range Image Object Detection Using Distance Continuity Merging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle-mounted object detection systems face challenges in processing load and object detection performance due to the need for excessive processing time and potential over-grouping of objects in range images, which degrades real-time detection capabilities.
Innovation Solution
An object detection apparatus that generates a range image and divides it into column regions, determines distance continuity between subgroups, and merges horizontally continuous subgroups to reduce processing load while maintaining detection accuracy, using a two-step grouping technique in vertical and horizontal directions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the object detection apparatus divides the grouping region using brightness information on the grayscale image, then object detection accuracy is improved, but processing load increases and real-time detection becomes difficult
Solution Approach 1:
The patent segments the grouping region detection process into two independent stages: first dividing the range image into column regions based on distance information, then detecting objects in each column region separately. This segmentation allows the system to maintain high detection accuracy by processing regions independently while reducing overall processing load through the structured division approach.
Solution Approach 2:
The patent extracts and utilizes only the distance information from the range image for the initial division into column regions, rather than processing full brightness information across the entire image. This extraction of essential distance data enables efficient region division without the computational burden of analyzing complete grayscale information, thereby improving real-time detection capability.
2Productivity
If the object detection apparatus detects objects in the grouping region of the range image to reduce processing load, then real-time detection is improved, but object detection performance degrades
Solution Approach 1:
The patent applies local quality by detecting objects in each column region with properties tailored to that specific region's characteristics. Each column region is processed independently with detection parameters optimized for its local distance range, ensuring high detection performance in each area while maintaining overall real-time capability through the distributed processing approach.
3Productivity
If a grouping region is made by grouping adjacent regions in the range image, then processing efficiency is improved, but different objects at substantially the same distance may be over-grouped
Solution Approach 1:
The patent segments the range image into multiple column regions based on distance information before object detection. This segmentation prevents over-grouping of different objects at the same distance by creating natural boundaries between column regions, while still maintaining processing efficiency through the structured organization of detection tasks across these segments.
Data Source
AI summary
In an object detection apparatus, a range image generator, based on distance information indicative of distances from a given measurement point to objects in real space, generates a range image indicative of a distribution of distance information of objects located around the measurement point. A subgroup generator horizontally divides the range image into a number of column regions of prescribed width, and for each column region, generates one or more subgroups that extend continuously in a vertical direction of the range image and fall within a given range of distance information. A continuity determiner determines, for each of the subgroups, whether or not there is distance continuity between the subgroup and its horizontally adjacent subgroup in the range image. A merger merges together horizontally continuous-in-distance subgroups. An object detector detects an object in each of regions of the range image corresponding to the respective merged groups.