Vehicle Surrounding Image Stitching via Floor-Wall Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current 360-degree vehicle around view monitoring (AVM) systems provide top-view images that do not accurately reflect the surrounding scenery, making it difficult for drivers to determine the relative position of the vehicle and obstacles, especially in parking situations or narrow passages.
Innovation Solution
The method involves capturing images from multiple cameras, transforming them into top-view images, identifying intervals of consecutive empty pixels to divide the images into floor and wall side images, and then stitching these images to generate a synthetic image that clearly reveals the vehicle's surroundings, using an electronic apparatus with modules for image capturing, transformation, division, and stitching.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a 360-degree AVM system captures images from four directions and transforms them into a top-view image, then the driver can determine the relative position between the vehicle and obstacles, but the top-view image cannot reflect the actual appearance of the scenery or obstacles and is inconsistent with the scene actually viewed by the driver
Solution Approach 1:
The patent divides the transformed top-view image into multiple sub-regions (floor side images and wall side images) based on detected intervals of consecutive empty pixels. This segmentation allows different parts of the image to be processed and combined differently, preserving both the top-view spatial relationship information and the actual appearance information of obstacles and scenery.
Solution Approach 2:
The patent combines the divided floor side images and wall side images to generate a synthetic image that integrates both the top-view spatial layout and the realistic appearance of surrounding objects. This merging process creates an enhanced image that maintains geometric accuracy while restoring visual fidelity.
2Area of stationary object
If the top-view image is generated by processing images from multiple cameras, then the driver can see the surrounding area, but the image does not clearly reveal the scenery surrounding the vehicle
Solution Approach 1:
The patent applies different processing strategies to different regions of the image. By identifying intervals of empty pixels and dividing the image accordingly, the system applies appropriate image combination techniques to floor side regions and wall side regions separately, optimizing the clarity and representativeness of each local area while maintaining overall coverage.
Data Source
AI summary
A method and an apparatus for processing surrounding images of a vehicle are provided. In the method, plural cameras disposed on the vehicle are used to capture images of plural perspective views surrounding the vehicle. The images of the perspective views are transformed into images of a top view. An interval consisted of at least a preset number of consecutive empty pixels is found from one column of pixels in each image of the top view, and the images of the perspective views and the top view are divided into floor side images and wall side images according to the height of the interval in the image. The divided floor side images and wall side images are stitched to generate a synthetic image surrounding the vehicle.