High-speed autofocus control using piezoelectric actuation and LSTM networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image-based autofocus methods are insufficient for high-speed inline inspection in manufacturing processes due to slow autofocus times, especially in applications like roll-to-roll flexible electronics printing, where vibrations and micro/nanoscale patterns require fast and continuous image-based autofocus.
Innovation Solution
A method for high-speed autofocus control that involves obtaining multiple images of a target, determining focus measure data, and adjusting the camera focus based on direct Gaussian mean calculations and adaptive step sizes determined by the focus measure curve standard deviation, along with the use of a piezoelectric motion stage controlled by a long short-term memory (LSTM) network for model predictive control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If step motor-based autofocus is used, then the system is simple and reliable, but the autofocus time is too slow and consumes excessive time
Solution Approach 1:
The patent replaces the mechanical step motor system with a piezoelectric actuator system. The piezoelectric actuator uses electro-mechanical conversion to achieve faster focus adjustment without the mechanical inertia and step limitations of traditional motors, directly resolving the contradiction between reliability and speed.
Solution Approach 2:
The patent changes the operating parameters by using piezoelectric materials that respond to voltage changes with precise dimensional adjustments. This allows the system to achieve the same focusing function with dramatically reduced response time by changing from mechanical rotation to electro-mechanical deformation.
2Measurement precision
If traditional image-based autofocus methods are used, then focusing accuracy can be maintained, but the number of images required is high and processing time increases
Solution Approach 1:
The patent applies preliminary action by using the LSTM network to predict the optimal focus position before actually capturing multiple images for analysis. The neural network pre-processes the information from fewer images to determine the focal point, reducing the total number of images needed while maintaining accuracy.
Solution Approach 2:
The patent introduces an intermediary computational layer (LSTM network) between image capture and focus determination. This intermediary processes the essential focusing information from a reduced set of images, acting as a mediator that maintains measurement precision while reducing the burden on the imaging system.
3Speed
If fast piezoelectric actuators are used for high-speed autofocus, then autofocus speed improves, but system complexity and control difficulty increase
Solution Approach 1:
The patent implements feedback control where the LSTM network continuously monitors image sharpness metrics and adjusts piezoelectric actuator commands in real-time. This closed-loop feedback system manages the complexity of controlling fast piezoelectric actuators by using intelligent algorithms to adapt to system variations and maintain optimal performance.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The proposed method significantly reduces autofocus time by 22% compared to Adaptive Hill-Climbing methods and requires 80% fewer images for focusing, while maintaining satisfactory accuracy as measured by root-mean-square error, and performs well in both well-lit and low-light conditions.
Implementation Method 1
a piezoelectric controlled motion stage configured to position a camera relative to a target
Implementation Method 2
determine a control input for the piezoelectric controlled motion stage using a long short-term memory (LSTM) backpropagation network trained to minimize a cost function
Data Source
AI summary
Various examples are provided related to high-speed autofocus control. In one example, a method includes obtaining a first image of a target with a camera; adjusting focus of the camera by a specified AF bin step size; obtaining a second image of the target; adjusting focus of the camera by the specified AF bin step size; obtaining a third image of the target; determining an optimal focus using data of the second and third images; and adjusting focus of the camera to the optimal focus. In another example, a method includes generating an input vector comprising a sequence of input-output pairs associated with a piezoelectric controlled motion stage that can position a camera relative to a target; determine a control input for the motion stage using a LSTM backpropagation network trained to minimize a cost function over a defined prediction horizon; and applying the control input to the motion stage.