Panoramic Image Landmark Recognition via Projection and Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing landmark recognition models for panoramic images suffer from low efficiency and accuracy due to distortion, requiring manual comparison with large databases, which is time-consuming and prone to errors.
Innovation Solution
A method involving projection transformation, semantic segmentation, distortion correction, and landmark recognition in a panoramic image to automatically identify landmarks by generating a projection image, determining landmark and road surface regions, correcting distortion, and recognizing landmarks in the corrected region.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing landmark recognition models are used on panoramic images, then recognition can be performed, but recognition accuracy deteriorates due to distortion
Solution Approach 1:
The panoramic image is segmented into multiple regions including a first region containing the landmark and a second region containing the road surface. This segmentation allows selective processing of different areas to mitigate distortion effects on landmark recognition while preserving contextual information from other regions.
Solution Approach 2:
The patent transforms the panoramic image from its original distorted dimensional representation to an unfolded or corrected dimensional representation. This dimensional transformation removes the distortion effects while maintaining the spatial relationships necessary for accurate landmark recognition.
2Measurement precision
If manual comparison with large databases is used for landmark recognition, then recognition can be performed, but processing efficiency deteriorates
Solution Approach 1:
The system performs preliminary actions by pre-processing the panoramic image through segmentation and distortion correction before landmark recognition. This preliminary processing creates optimized input data that enables faster and more accurate recognition without requiring exhaustive manual comparison with entire databases.
Solution Approach 2:
The patent extracts the landmark region from the panoramic image by segmenting it into distinct regions. This extraction isolates the relevant landmark information from the rest of the image, allowing recognition algorithms to focus only on the extracted region rather than processing the entire panoramic image or comparing it with complete databases.
3Measurement precision
If semantic segmentation is performed to separate landmark and road surface regions, then recognition accuracy improves, but processing complexity increases
Solution Approach 1:
The panoramic image is segmented into multiple regions including a first region containing the landmark and a second region containing the road surface. This segmentation allows selective processing of different areas to mitigate distortion effects on landmark recognition while preserving contextual information from other regions.
Data Source
AI summary
Disclosed are a method and apparatus for recognizing a landmark in a panoramic image. The method includes steps of performing projection transformation on the panoramic image so as to generate a projection image; conducting semantic segmentation on the projection image so as to determine a landmark region and a road surface region; correcting distortion in the landmark region so as to produce a corrected landmark region; and recognizing the landmark in the corrected landmark region.