Peripheral Mosaicking for Partial Vehicle Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current driving assistance systems face challenges in rapidly processing images to detect vehicles that are partially or suddenly appearing in the forward view of a moving vehicle, leading to delayed obstacle detection and information provision to the driver.
Innovation Solution
An image processing apparatus and method that extracts peripheral regions corresponding to a target object size, modifies these regions to change the viewpoint, creates a mosaicking image by stitching them together, and uses a machine learning algorithm to detect objects, including vehicle wheels, from the mosaicking image, thereby reducing calculation and function call requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the entire forward view image is processed for object detection, then detection coverage is improved, but processing time increases
Solution Approach 1:
The patent divides the forward view image into multiple peripheral regions corresponding to different angular ranges. Instead of processing the entire image, the system segments it into manageable parts and processes only relevant regions, reducing overall processing time while maintaining detection coverage for vehicles appearing from side directions.
Solution Approach 2:
The patent extracts only the necessary peripheral regions from the full image that are likely to contain partially visible vehicles. By taking out and processing only these specific regions rather than the entire image, the system reduces calculation amount and processing time while maintaining effective detection coverage.
2Measurement precision
If traditional object detection is applied to the full image, then all objects can be detected, but calculation amount increases
Solution Approach 1:
The patent segments the image processing task by identifying and processing only peripheral regions that are likely to contain partially visible vehicles. This segmentation reduces the quantity of pixels and data requiring calculation while maintaining detection accuracy for the target objects in those regions.
Solution Approach 2:
The patent applies different processing strategies to different regions of the image. Peripheral regions are extracted and processed with specific viewpoint modifications, while the center region is handled differently. This local quality approach optimizes calculation amount by applying intensive processing only where needed for detecting partially visible vehicles.
3Measurement precision
If the viewpoint is not modified for peripheral regions, then processing is simpler, but partially visible vehicles cannot be detected
Solution Approach 1:
The patent performs preliminary viewpoint modification on peripheral regions before object detection. By pre-processing these regions to change the viewpoint and make partially visible vehicles more prominent, the system enables accurate detection without requiring overly complex detection algorithms, thus managing processing complexity effectively.
Solution Approach 2:
The patent changes the viewpoint dimension for peripheral regions through image processing transformations. This dimensional change in how the peripheral regions are viewed makes partially visible vehicles more detectable, improving detection accuracy while keeping the processing complexity manageable through standardized transformation operations.
4Measurement precision
If multiple function calls are made for object detection, then comprehensive detection is achieved, but processing time increases
Solution Approach 1:
The patent segments the detection process into: (1) extracting peripheral regions, (2) modifying viewpoints of these regions, (3) stitching them into a mosaicking image, and (4) performing a single object detection function call on the stitched image. This segmentation reduces the number of function calls from multiple separate detections to one comprehensive detection, significantly reducing processing time while maintaining detection comprehensiveness.
Solution Approach 2:
The patent merges multiple peripheral region images into a single mosaicking image through stitching. By combining these regions into one unified image structure, the system enables a single object detection function call to cover all peripheral areas, reducing the number of function calls and processing time while maintaining comprehensive detection coverage.
Data Source
AI summary
Provided are an image processing apparatus and method, and a driving assistance system employing the same. The image processing apparatus includes a peripheral region extracting unit extracting, from an image, peripheral regions corresponding to a size of a target object determined in advance, a modifying unit modifying the peripheral regions to allow a viewpoint for the peripheral regions to be changed, a mosaicking image creating unit creating a mosaicking image by stitching the modified peripheral regions together, and an object detection unit detecting an object including a part of the target object from the mosaicking image.