Object Key Point Detection Using Polar Coordinate Axes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional object detection methods struggle to accurately detect key points on an image that are not the boundary of an object, and existing deep learning techniques like ExtremeNet and YOLO do not effectively identify feature points.
Innovation Solution
An object detection method and device that utilizes a machine-learning model, specifically a convolutional neural network, to estimate key point candidates by projecting points onto defined coordinate axes in a polar coordinate system set for an object model, considering the object's shape, and detects key points based on these projections.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional object detection methods (ExtremeNet, YOLO) are used to detect objects, then detection speed and basic object localization are improved, but the ability to detect key points that are not object boundaries deteriorates
Solution Approach 1:
The detection process is segmented into two distinct stages: first detecting object boundaries using conventional methods, then detecting key points within those boundaries using the polar coordinate system approach. This segmentation allows each stage to optimize for its specific task, maintaining detection speed while improving key point accuracy
Solution Approach 2:
The invention transforms the key point detection problem from a 2D Cartesian coordinate system to a polar coordinate system centered on the object. This dimensional transformation allows the detection algorithm to naturally identify radial key points (such as tips of body parts, protrusions, or characteristic features) by searching along polar axes at defined angles, thereby improving key point detection capability without sacrificing overall detection efficiency
2Measurement precision
If four boundary end points are detected to form a bounding box, then object position accuracy is improved, but detection of feature points inside the object deteriorates
Solution Approach 1:
The method performs preliminary detection of object boundaries and bounding boxes first, establishing a foundation for subsequent key point detection. By pre-defining the object region and its boundaries, the system creates a structured framework that guides the subsequent polar coordinate-based key point search, ensuring both position accuracy and feature point detection
Solution Approach 2:
The polar coordinate system acts as an intermediary framework that connects the bounding box detection results with the key point detection task. By transforming coordinates and defining polar axes relative to the detected object center and boundaries, the system bridges the gap between boundary-based detection and feature point identification, preserving both position accuracy and feature information
Data Source
AI summary
An object detection method includes a key point estimation step of estimating key point candidates for each object in an image; and a detection step of detecting key points for each object based on the estimated key point candidates. Considering an object model that models shape of an object, the key points are points that satisfy a defined condition among points indicating a boundary of the object model that are projected onto defined coordinate axes. The defined coordinate axes have an origin at a geometric center of the object model and each forms a defined angle relative to a polar axis in a polar coordinate system set for the object model.