Dynamic Image Processing Area Adjustment for Pedestrian Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle periphery monitoring devices face challenges in accurately setting image processing areas for extracting monitoring objects due to varying radar beam profiles, which can result in incomplete capture of object images, especially when objects are difficult to reflect radar beams.
Innovation Solution
A vehicle periphery monitoring device and method that includes an image processing target area setting portion, feature region extracting portion, and object type discriminating portion to dynamically adjust the image processing area based on distance data from a radar and luminance distribution, ensuring the inclusion of the entire monitoring object, particularly pedestrians, by correcting the image processing area using feature region and luminance projections.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If the image processing area is determined based on radar distance information and beam profile, then the image processing time is reduced by limiting the processing area, but the completeness of object image capture deteriorates when objects are difficult to reflect radar beams
Solution Approach 1:
The image processing area is dynamically adjusted based on object type. The system first performs initial image processing in a radar-based area, then determines object type using extracted features. Based on the object type determination, the system dynamically expands or contracts the image processing area to ensure complete object capture while maintaining processing efficiency.
Solution Approach 2:
The system uses feedback from feature extraction and object type determination to correct and adjust the image processing area. The extracted features and determined object type provide feedback that guides the dynamic adjustment of the processing area boundaries, ensuring that the final area encompasses the complete object while optimizing processing time.
2Reliability
If the image processing area is expanded to ensure complete object capture, then the object image capture completeness is improved, but the image processing time increases due to processing larger areas
Solution Approach 1:
The image processing is segmented into multiple stages: initial processing in a radar-based area, feature extraction, object type determination, and then selective expansion of the processing area based on object type. This segmentation allows the system to process only necessary areas at each stage, maintaining efficiency while ensuring completeness.
Solution Approach 2:
The image processing area is dynamically adjusted based on object type. The system first performs initial image processing in a radar-based area, then determines object type using extracted features. Based on the object type determination, the system dynamically expands or contracts the image processing area to ensure complete object capture while maintaining processing efficiency.
3Device complexity
If the image processing area is determined using fixed beam profile assumptions, then the processing method is simplified, but the accuracy of object type discrimination deteriorates due to varying radar reflection characteristics
Solution Approach 1:
The system dynamically adjusts the image processing area based on determined object type, rather than using fixed beam profile assumptions. This dynamic adjustment allows the processing area to adapt to different object types and their varying radar reflection characteristics, improving discrimination accuracy while maintaining reasonable processing complexity.
Solution Approach 2:
The system uses its own extracted features and determined object type to automatically adjust the image processing area. This self-service mechanism eliminates the need for complex pre-programmed beam profile assumptions for different object types, as the system autonomously adapts based on real-time feature analysis.
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
Improves the accuracy of object type discrimination by ensuring the entire image portion of the monitoring object is included in the image processing area, enhancing the detection and classification of pedestrians and other objects in real-time.
Implementation Method 1
a radar mounted in a vehicle and configured to detect a distance between the vehicle and an object located in a first monitoring range of a periphery of the vehicle
Implementation Method 2
an imaging unit mounted in the vehicle and configured to photograph a second monitoring range overlapped with the first monitoring range
Data Source
AI summary
A vehicle periphery monitoring device is provided with an image processing target area setting portion for setting an image processing target area (61) that may include an image portion of an monitoring object in a captured image (Im2) on the basis of a distance between an object and a vehicle; a feature region extracting portion for extracting a feature region with a feature amount of a head of a pedestrian in a search area (64) based on the image processing target area (61); an image processing target area correcting portion for correcting a range of the image processing target area from (61) to (62) on the basis of the position of a feature region (60a), and an object type discriminating portion for discriminating a type of real space monitoring object corresponding to the image portion included in the corrected image processing target area (62).