Robotic Grasp Point Detection Using Surface Normal Scattering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing robotic systems struggle with generalizing object picking tasks across different scenes, shapes, and appearances due to challenges in image pre-processing and neural network training, leading to inefficiencies in identifying optimal gripping locations.
Innovation Solution
A method involving the calculation of surface normal vector scattering measures, such as standard deviation, is applied to pre-process depth images, enhancing the input data for neural networks to improve the detection of gripping locations, enabling better generalization and accuracy in picking up objects with varying appearances.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional image processing methods are used for robotic picking, then the system can handle simple objects, but it fails to generalize to new scenes, backgrounds, shapes, and appearances
Solution Approach 1:
The patent applies preliminary action by performing pre-processing of depth images to calculate surface normal vectors and their scattering measures before feeding data to the neural network. This pre-computation of geometric features enables the system to generalize better to new objects and scenes while maintaining reliable detection accuracy, as the preprocessing extracts invariant geometric properties that transcend appearance variations
2Adaptability or versatility
If complex pre-processing steps are applied to highlight relevant aspects, then generalization improves, but computational complexity and processing time increase
Solution Approach 1:
The patent applies the extraction principle by isolating and computing only the most relevant geometric feature - the scattering measure of surface normal vectors - from the depth image. This selective extraction of critical geometric information improves generalization capability while avoiding the computational burden of complex full-image pre-processing, as it focuses processing on invariant surface properties rather than entire image scenes
3Measurement precision
If more comprehensive image data is used for training, then detection accuracy improves, but training time and computational resources increase
Solution Approach 1:
The patent applies parameter changes by transforming the raw depth image data into a different parameter space - specifically, computing surface normal vectors and their scattering measures. This parameter transformation creates a more informative feature representation that improves detection accuracy while requiring less training data and time, as the geometric parameters directly capture object shape characteristics that are invariant to appearance variations
Data Source
AI summary
A method for picking up an object by means of a robotic device. The method includes obtaining at least one depth image of the object; determining, for each of a plurality of points of the object, the value of a measure of the scattering of surface normal vectors in an area around the point of the object; supplying the determined values to a neural network configured to output, in response to an input containing measured scattering values, an indication of object locations for pick-up; determining a location of the object for pick-up from an output which the neural network outputs in response to the supply of the determined values; and controlling the robotic device to pick up the object at the determined location.