Real-Time Vehicle Detection Using Segmented Candidate Regions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current vehicle detection methods for forward collision warning systems are computationally intensive, making real-time detection at full VGA resolution challenging on embedded processors, and often require undersampling or additional processors, which compromise reliability or increase cost and complexity.
Innovation Solution
The method employs a dual-path approach using Sobel edge maps and summed area tables, combined with run-length encoding and Haar templates for symmetry and corner detection, to efficiently identify vehicles in real-time at VGA resolution on a single dual-core DSP, optimizing processor cycles and reducing reliance on all detection features for improved reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If histogram of gradients and SVM classifier are used for object detection, then detection reliability is improved, but computational complexity increases making real-time detection difficult
Solution Approach 1:
The patent segments the detection process into multiple stages: candidate region generation, feature extraction, and classification. By dividing the image into candidate regions first and then applying computationally intensive SVM classification only to these regions rather than the entire image, the system maintains high detection reliability while reducing overall computational complexity to enable real-time processing.
Solution Approach 2:
The patent performs preliminary actions by generating candidate object regions and extracting features before applying the SVM classifier. This preliminary processing filters out non-relevant areas, so that the computationally intensive classification step operates only on a small subset of candidate regions, thereby maintaining reliability while reducing real-time computational burden.
2Measurement precision
If full VGA resolution processing is performed, then detection precision is improved, but processing speed decreases making real-time detection challenging
Solution Approach 1:
The patent segments the full VGA resolution image into multiple candidate regions of interest. By applying detection algorithms only to these segmented regions rather than processing the entire high-resolution image, the system maintains detection precision on relevant objects while significantly improving processing speed to achieve real-time performance.
Solution Approach 2:
The patent applies partial action by processing only the necessary candidate regions rather than the entire image. This selective processing approach maintains sufficient detection precision for safety-critical applications while reducing the overall computational load to enable real-time operation at VGA resolution.
Data Source
AI summary
Objects are detected in real-time at full VGA 30 frame per second resolution. A preprocessor performs run-length encoding (RLE) and generates a summed area table (SAT) of an image. The RLE and SAT are used to identify candidate objects and to iteratively refine their boundaries. A histogram of gradients (HoG) and support vector machine (SVM) then reliably classify the object. The method may be part of an advanced driver assistance system (ADAS).