Automated Tuning and Calibration of a Computer Vision System
AssigneeGriffyn Robotech Pvt. Ltd.
InventorsAmit Anil Mahajan, Samir Shriram Bagalkote
Life of this patent
Estimating the term…
Abstract
A system and method for autocalibration of the camera and lighting components of an inspection and cosmetic grading machine is provided. Camera and lighting assemblies capture images of an object and create a 2D composite image which is processed by an image processing module with a deep learning machine algorithm to detect surface defects in the object and may measure the depth associated with light intensity. Precisely calibrated lights and vision system are required. Over time, environmental factors and component vibration and movement dislocate the components causing both lighting and camera anomalies. Automated systems are proposed that allow automated calibration of these systems without human intervention.
Claims6
1-2. (canceled)
A method for autocalibration of a computer vision system, comprising:
- placing a test object in an object holder of a computer vision system;
- acquiring images of a test object, using the computer vision system and image processing algorithms;
- analyzing the acquired images using trained machine learning algorithms and optimization methods, where the optimization process computes a deviation in the processed image with respect to a reference image stored in computer memory;
- generating an actuation signal for tuning the vision system parameters;
- setting tunable parameters;
- iterating the image acquisition and analysis with all test objects, automatically training the deep learning model with the results to perform fine tuning of the parameters to ideal settings;
- and
- automatically tuning the features to the parameters representing the best possible configuration in comparison with the reference image.
A system for autocalibration of a computer vision system, comprising:
- a motorized motion control unit, capable of integration with a computer vision environment and operatively configured to control the position and features of components of a computer vision system, wherein the components include the object manipulation, lighting and lens components;
- a computing device comprising memory and a processor, further comprising deep-learning, optimization and image processing algorithms, and measurement tool instructions stored in memory which when executed by the processor perform the following steps:
- (i) instruct the motion control unit to make adjustments to the computer vision system components, (ii) collect measurements and data of the adjustments and quality of resulting captured images, (iii) present the data to the deep-learning algorithm to determine and automatically set the optimal component features for vision tuning;
- and
- a user interface operatively configured to assist a user in viewing and maintaining the motion control unit and computing device.
Dependencies are inferred from the claim text.