Object Pose Estimation via Local Coordinate Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current object pose obtaining methods are limited in robustness due to factors like light intensity and occlusion, as they rely on the entire image for pose estimation, which can lead to inaccurate results.
Innovation Solution
The method extracts an image block of the object, generates a local coordinate system, obtains 2D projection key points from a 3D model, converts these points into the local coordinate system, and determines the object pose using direction vectors between pixel points and prediction key points, thereby improving robustness by considering occlusion and light intensity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the entire image is used for pose estimation, then more information is available for analysis, but the accuracy and robustness deteriorate due to light intensity and occlusion factors
Solution Approach 1:
The patent segments the image into multiple image blocks, each corresponding to a different object or region. By processing each block separately rather than the entire image, the method isolates the object of interest from distracting background elements, light variations, and occlusions, thereby improving pose estimation accuracy while reducing the impact of harmful environmental factors.
Solution Approach 2:
The patent extracts the relevant object region from the full image by generating image blocks that contain only the object of interest. This extraction process removes unnecessary background information and harmful factors such as varying light intensity and occlusions from other objects, allowing the pose estimation algorithm to focus solely on the target object and improve measurement precision.
2Reliability
If image blocks are extracted and processed separately, then robustness against occlusion and light intensity improves, but the complexity of the processing system increases
Solution Approach 1:
The patent divides the complex task of pose estimation into simpler sub-tasks by segmenting the image into multiple blocks. Each block is processed independently through the same pipeline (coordinate system generation, key point detection, pose calculation), which simplifies the overall system architecture while improving reliability. The modular approach makes the system more robust to occlusion and lighting variations without requiring complex adaptive logic.
Solution Approach 2:
The patent applies local quality by generating a dedicated local coordinate system for each image block rather than using a global coordinate system. This allows each block to be processed with optimized parameters and transformations tailored to its specific characteristics, improving robustness against local variations in occlusion and lighting while maintaining a relatively simple overall processing framework.
Data Source
AI summary
The present disclosure provides an object pose obtaining method, and an electronic device, relates to technology fields of image processing, computer vision, and deep learning. A detailed implementation is: extracting an image block of an object from an image, and generating a local coordinate system corresponding to the image block; obtaining 2D projection key points in an image coordinate system corresponding to a plurality of 3D key points on a 3D model of the object; converting the 2D projection key points into the local coordinate system to generate corresponding 2D prediction key points; obtaining direction vectors between each pixel point in the image block and each 2D prediction key point, and obtaining a 2D target key point corresponding to each 2D predication key point based on the direction vectors; and determining a pose of the object according to the 3D key points and the 2D target key points.