Object Recognition Using 3D Shape Approximation in Cluttered Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection methods in industrial robots face challenges in accurately identifying target objects in images with backgrounds or overlapping objects, leading to reduced accuracy and inefficiencies in tasks like order picking due to the need for multiple templates and complex annotations.
Innovation Solution
An object recognition device and method that utilizes an image acquisition unit, a three-dimensional shape approximation determination unit, an image region estimation unit, and a region selection unit to identify and select the smallest area region within an image, approximating it to three-dimensional shapes, thereby enhancing detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional object detection methods are used to detect objects in images with backgrounds or overlapping objects, then the detection process is simple, but the object detection accuracy decreases
Solution Approach 1:
The patent transforms the 2D image detection problem into a 3D shape approximation problem. By estimating whether rectangular regions can be approximated to 3D shapes and selecting regions with smallest areas, the system adds dimensional reasoning to improve detection accuracy in complex scenes with backgrounds and overlapping objects.
Solution Approach 2:
The patent segments the image processing into distinct steps: acquiring image data, determining 3D shape approximability of rectangular regions, cutting out estimation regions, and selecting the smallest area region. This segmentation allows each step to be optimized independently, improving overall detection accuracy without excessive complexity.
2Adaptability or versatility
If multiple templates are prepared for objects with different postures, then the detection coverage is improved, but the preparation time and cost increase
Solution Approach 1:
The patent creates a universal detection method that works for objects with different postures without requiring separate templates. The 3D shape approximation determination and smallest area region selection provide a single, posture-invariant approach that adapts to various object orientations and configurations.
Solution Approach 2:
Instead of changing templates based on posture, the patent changes the detection parameters by evaluating 3D shape approximability and area metrics. This parameter-based approach maintains versatility across different object postures while avoiding the time-consuming template preparation process.
Data Source
AI summary
Provided is an object recognition device for detecting an object in an image, with which it is possible to detect, with high accuracy, a target object even in an image that has a background and other objects shown therein. This object recognition device is characterized by comprising: an image acquisition unit which acquires a first image composed of two-dimensional pixels; a three-dimensional shape approximation determination unit which determines whether image information of a prescribed rectangular region within the first image can be approximated by prescribed three-dimensional shape information; an image region estimation unit which clips out the rectangular region as a first estimation region on the basis of a determination result provided by the three-dimensional shape approximation determination unit; and a region selection unit which selects, from among a plurality of the first estimation regions, a first estimation region that has the smallest area.