Spherical Image Feature Extraction via Solid Angle Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for image feature extraction and object detection in panoramic images, such as equirectangular images, face challenges with distortion, particularly near poles, making it difficult to maintain shape and straight lines, and require inefficient projection transformations or complex sample collection.
Innovation Solution
Defining a combination of solid angles along specific directions for a spherical image to determine values that evenly distribute surface areas of spherical segments, allowing for the generation of an image feature template for efficient feature extraction and object detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Shape
If projection transformation is conducted to correct panoramic image distortion, then shape preservation is improved, but processing efficiency deteriorates due to the non-existence of perfect transformation approaches and the need for multiple transformations at different positions
Solution Approach 1:
The spherical image is divided into multiple spherical segments based on solid angle calculations. Each segment is processed independently to extract image features, avoiding the need for complex global projection transformations while maintaining shape accuracy in each local region.
Solution Approach 2:
The patent applies different processing approaches to different regions of the spherical image based on their local characteristics. By using solid angle-based segmentation, each spherical segment maintains its local geometric properties, providing shape preservation where needed without requiring uniform transformation across the entire image.
2Shape
If projection transformation is conducted to correct panoramic image distortion, then shape preservation is improved, but device complexity increases due to the need for multiple transformation operations
Solution Approach 1:
The image is segmented into spherical segments using solid angle calculations, which provides a mathematically elegant and computationally efficient way to handle distortion correction without requiring multiple sequential transformation operations.
Solution Approach 2:
The patent transitions from traditional 2D planar projection transformations to 3D spherical coordinate-based segmentation. By working in the spherical dimension and using solid angles, the method simplifies the mathematical operations required for distortion correction while maintaining shape fidelity.
3Measurement precision
If samples are collected at different positions for training, then object detection accuracy is improved, but process complexity increases due to the complicated and difficult sample collection process
Solution Approach 1:
Instead of manually collecting samples at different positions, the patent automatically segments the spherical image into spherical segments based on solid angle calculations. This automated segmentation process replaces the complex manual sample collection process while still providing diverse training data from different regions of the image.
Solution Approach 2:
The system performs self-service by automatically generating training segments from the input spherical image through solid angle-based segmentation. This eliminates the need for external manual sample collection and annotation processes, reducing complexity while maintaining the diversity and quality of training data.
4Shape
If uniform division of spherical image is achieved through solid angle calculation, then distortion reduction is improved, but calculation complexity increases
Solution Approach 1:
The patent embraces the spherical nature of the panoramic image by using solid angle calculations specifically designed for spherical geometry. This approach naturally handles the curvature of the spherical surface, providing uniform division that reduces distortion while being mathematically elegant and computationally efficient for spherical data.
Data Source
AI summary
Disclosed is an image feature extraction method including a step of defining a combination of at least two kinds of solid angles along at least two directions, of an input spherical image; a step of determining respective values of the combination of the at least two kinds of solid angles, so that surface areas of spherical crowns of spherical segments, which are obtained by dividing the spherical image by the respective values of the combination of the at least two kinds solid angles, have a same value; and a step of generating, by utilizing the respective values of the combination of the at least two kinds of solid angles, an image feature template so as to conduct image feature extraction.