Vehicle Detection Using Direction-Based Reference Image Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image recognition technologies for detecting surrounding vehicles face challenges in accuracy and computational efficiency, particularly when the relative angle between vehicles is high, leading to difficulties in recognizing vehicle types and distinguishing them from other moving objects.
Innovation Solution
A vehicle determination apparatus that identifies the traveling direction of surrounding vehicles using a direction identification unit, acquires image feature amounts from reference images corresponding to the identified direction, and compares them with computed image features to determine the presence of surrounding vehicles, thereby reducing computational load and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large amount of reference images are stored in the database for image recognition, then the detection accuracy of surrounding vehicles is improved, but the computational amount increases and processing speed is reduced
Solution Approach 1:
The patent segments the large database of reference images into multiple sub-databases based on vehicle types (e.g., sedans, trucks, buses). The image recognition apparatus first identifies the vehicle type using a simplified method, then selects only the corresponding sub-database for detailed recognition. This segmentation reduces the computational load while maintaining detection accuracy by focusing resources on relevant vehicle categories.
Solution Approach 2:
The patent implements a preliminary vehicle type identification step before performing detailed image recognition. By first categorizing the vehicle into a specific type using a quick assessment method, the system prepares the appropriate sub-database in advance, avoiding the need to process all reference images and thereby improving processing speed without sacrificing detection accuracy.
2Reliability
If the degree of skew between the vehicle and surrounding vehicle is high, then the optical flow process is used for detection, but the identification of vehicle type and distinction from other moving objects becomes difficult
Solution Approach 1:
The patent implements a dynamic switching mechanism that adapts the recognition method based on the degree of skew. When the skew angle is small, detailed image recognition is performed for accurate vehicle type identification. When the skew angle is large, the system switches to optical flow-based detection to maintain reliable surrounding vehicle detection. This dynamic adaptation resolves the contradiction by optimizing the method based on real-time conditions.
Solution Approach 2:
The patent changes the recognition parameter (method selection) based on the skew angle parameter. By monitoring the degree of skew and adjusting the recognition approach accordingly, the system maintains both detection reliability and vehicle type identification accuracy across different viewing conditions.
3Adaptability or versatility
If image recognition is performed using photographed images with high relative angles, then the detection covers more directions, but the detection rate is reduced due to significant differences in image information
Solution Approach 1:
The patent segments the detection process into two stages: first, a broad screening phase that covers all directions to ensure comprehensive detection coverage; second, a detailed recognition phase that applies only to regions where vehicles are suspected to exist. This segmentation allows the system to maintain wide detection coverage while achieving high detection rates by concentrating computational resources on relevant areas.
Solution Approach 2:
The patent performs preliminary detection to identify regions containing surrounding vehicles before conducting detailed image recognition. This preliminary action filters out irrelevant areas, allowing the system to maintain high detection coverage while improving the detection rate by focusing detailed analysis only on regions where vehicles are likely to be present.
Data Source
AI summary
A direction identification unit (21) identifies a traveling direction in which a surrounding vehicle travels in a partial region of a region indicated by image information (42) obtained by photographing using a camera (31). A feature amount acquisition unit (22) acquires a reference feature amount (41) being a feature amount computed from a reference image corresponding to the identified traveling direction. A vehicle determination unit (23) computes an image feature amount being a feature amount of the image information of the partial region and compares the computed image feature amount with the acquired reference feature amount (41), thereby determining whether the surrounding vehicle is present in the partial region.