Multi-Camera Ball Location via Segmentation and Triangulation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current object detection methods are inadequate for real-time detection and tracking of small balls in multi-camera super high-resolution videos, particularly due to the ball's small size, poor color or texture, motion blur, and occlusion by players, which limits their accuracy and efficiency.
Innovation Solution
A multi-camera solution using 12 super-high resolution cameras with a customized object detection scheme like YOLO and tracking-by-detection, combined with 3D triangulation and bundle adjustment for accurate ball location, and RANSAC for cross-validation to mitigate false and miss detections.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If general object detection methods using CNNs are applied, then rich feature extraction is achieved, but the methods cannot be applied for small size object detection
Solution Approach 1:
The patent segments the super high-resolution video frames into multiple sub-regions or tiles, allowing the detection system to process small ball objects more effectively by dividing the large image into manageable sections that can be analyzed with appropriate detection algorithms
2Measurement precision
If sliding window or image partition method is used for small size object detection, then general object location is resolved, but full technical solution for multi-cam super high-resolution video tracking is not provided
Solution Approach 1:
The patent merges multiple detection results from different cameras through multi-cam build logic, combining individual camera detections into a unified multi-camera detection system that provides comprehensive ball location and tracking across the entire sports field
Solution Approach 2:
The patent transitions from 2D detection in individual camera frames to 3D spatial localization by integrating data from multiple cameras at different positions, adding the depth dimension to achieve accurate ball location in three-dimensional space
3Measurement precision
If multiple super high-resolution cameras are used, then ball detection accuracy is improved, but processing complexity and computational load increase
Solution Approach 1:
The patent performs preliminary detection in each individual camera before combining results, allowing each camera to independently identify potential ball locations which are then validated and integrated in the multi-cam build stage, reducing the overall computational burden
4Measurement precision
If detection is performed on the entire frame, then comprehensive coverage is achieved, but processing time increases for real-time detection
Solution Approach 1:
The patent divides the large super high-resolution video frames into multiple smaller sub-regions or tiles, enabling parallel processing of different regions and reducing the computational time required for each detection operation while maintaining comprehensive coverage through aggregation of all sub-region results
Data Source
AI summary
A multi-camera architecture for detecting and tracking a ball in real-time. The multi-camera architecture includes network interface circuitry to receive a plurality of real-time videos taken from a plurality of high-resolution cameras. Each of the high-resolution cameras simultaneously captures a sports event, wherein each of the plurality of high-resolution cameras includes a viewpoint that covers an entire playing field where the sports event is played. The multi-camera architecture further includes one or more processors coupled to the network interface circuitry and one or more memory devices coupled to the one or more processors. The one or more memory devices includes instructions to determine the location of the ball for each frame of the plurality of real-time videos, which when executed by the one or more processors, cause the multi-camera architecture to simultaneously perform one of a detection scheme or a tracking scheme on a frame from each of the plurality of real-time videos to detect the ball used in the sports event and perform a multi-camera build to determine a location of the ball in 3D for the frame from each of the plurality of real-time videos using one of detection or tracking results for each of the cameras.