Video Processing for Rotary Cameras Using Spherical Coordinate Transformation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Rotary video capturing devices, such as pan-tilt surveillance cameras, face challenges in object detection due to changing fields of view, leading to invalid or incorrect results when the background image is not stable across different view positions.
Innovation Solution
A video processing method and system that involves receiving video frames with field-of-view information, performing coordinate transformations to obtain panoramic background pixels, and comparing these with original pixels to determine foreground or background, enabling effective foreground detection and background learning-and-updating.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Difficulty of detecting and measuring
If background subtraction is used for object detection, then foreground detection capability is improved, but detection accuracy deteriorates when the rotary device moves to new view positions
Solution Approach 1:
The patent transforms the problem from a 2D image coordinate system to a 3D spherical coordinate system by introducing field-of-view information (horizontal and vertical angles). This dimensional expansion allows the system to map pixels from different view positions to corresponding locations in a unified spherical coordinate space, enabling accurate background subtraction even when the rotary device moves to new positions.
Solution Approach 2:
The patent changes the coordinate parameters from simple (x, y) image coordinates to (pan, tilt, zoom) field-of-view parameters. By transforming pixel coordinates using these parameters and the intrinsic matrix, the system can accurately locate corresponding pixels across different views, maintaining detection accuracy despite changes in view position.
2Measurement precision
If a panoramic background image is created through coordinate transformation, then detection accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent performs coordinate transformation and constructs the panoramic background image in advance, before actual object detection is needed. This preliminary action allows the system to pre-process and store the background information in a unified coordinate system, so that subsequent detection operations only require simple pixel-by-pixel comparison rather than repeated complex transformations.
Solution Approach 2:
The patent creates a transformed copy of the current view as a panoramic background image in spherical coordinates. This copy serves as a reference that can be directly compared with future views after similar transformations, avoiding the need to re-process original images and reducing computational burden during real-time detection.
3Adaptability or versatility
If background images are continuously updated for each view position, then adaptability is improved, but system stability deteriorates due to lack of convergence
Solution Approach 1:
The patent creates a universal panoramic background image in spherical coordinates that can serve as a reference for multiple different view positions. Instead of maintaining separate background images for each view, this single unified background structure handles all views through coordinate transformation, providing both adaptability to new views and stability through a consistent reference framework.
Solution Approach 2:
The system continuously updates the panoramic background image by comparing new transformed views with the existing background, using the detected foreground objects to refine and update the background model. This feedback mechanism allows the background to adapt to gradual changes in the environment while maintaining stability through iterative convergence.
Data Source
AI summary
A video processing method and a video processing system are provided. The video processing method is applied to a rotary video capturing device. The video processing method includes the following steps. First, at least one video frame corresponding to at least one piece of field-of-view (FOV) information is received, and a plurality of first coordinates corresponding to a plurality of first pixels of the at least one video frame is obtained. Coordinate transformation procedure is performed on the first coordinates according to the at least one piece of FOV information to obtain a plurality of second coordinates. A plurality of second pixels in a panoramic background image corresponding to the second coordinates is obtained. The first pixels are compared with the second pixels to determine whether the first pixels belong to a foreground or a background, to output comparison results.