Neural Network Auto Focus for Moving Objects

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing auto focus (AF) systems in digital cameras struggle to accurately focus on moving objects due to mechanical delays and reliance on slower hardware, leading to incorrect predictions and reduced image quality, especially when using heavier lenses or capturing objects closer to the camera.

Innovation Solution

An electronic device and method that generates focal codes based on depth information of moving objects, allowing for precise prediction and focus adjustment, independent of AF hardware, and captures images with improved focus times and quality.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional AF systems rely on mechanical hardware and object tracking techniques, then they can capture moving objects, but the mechanical delay and slower hardware result in incorrect focus predictions and reduced image quality

Engineering Contradiction:

Improvefocus prediction accuracyVSAvoidmechanical delay

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent replaces the traditional mechanical AF system with a computational approach using a neural network model. The neural network predicts future object positions and calculates focal codes without relying on mechanical movement, thereby eliminating mechanical delay and improving focus prediction accuracy for moving objects

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system performs preliminary actions by using the neural network to predict future object positions before the actual image capture. This allows the AF system to pre-calculate the appropriate focal code based on predicted object location, ensuring accurate focus is achieved by the time the image is captured

Inventive Principle:

Principle #10Preliminary action

2Measurement precision

If heavier lenses (>85mm) are employed to capture moving objects, then image quality can be improved, but the impact of lack of focus becomes more pronounced due to slower AF hardware

Engineering Contradiction:

Improveimage qualityVSAvoidfocus accuracy

Core Design Contradiction:

Measurement precisionVSReliability

Solution Approach 1:

By replacing the mechanical AF system with a neural network-based computational system, the patent eliminates the bottleneck of slow AF hardware. This allows heavier lenses to be used effectively since the computational system can rapidly calculate focal codes without being constrained by mechanical response time, maintaining both image quality and focus accuracy

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Measurement precision

If the time period between shutter press and image capture is minimized, then focus prediction accuracy improves, but existing AF systems allow too much time for object movement fluctuations due to hardware limitations

Engineering Contradiction:

Improveposition prediction accuracyVSAvoidobject movement fluctuation time

Core Design Contradiction:

Measurement precisionVSDuration of action of moving object

Solution Approach 1:

The neural network-based system replaces mechanical processing with computational processing, which operates much faster. This allows the system to minimize the time between shutter press and image capture while accurately predicting object position, as the neural network can rapidly process input data and output focal codes without being constrained by mechanical response time

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

4Ease of operation

If object tracking techniques are heavily relied upon, then moving objects can be tracked, but slight camera movement leads to significant changes in focus when objects are closer to the camera

Engineering Contradiction:

Improveobject tracking capabilityVSAvoidfocus stability

Core Design Contradiction:

Ease of operationVSMeasurement precision

Solution Approach 1:

The patent incorporates feedback mechanisms where the neural network continuously receives input data including object position, camera position, and depth information. This feedback loop allows the system to dynamically adjust focal codes based on real-time conditions, compensating for camera movement and maintaining stable focus on moving objects

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system transitions from two-dimensional image plane tracking to three-dimensional spatial tracking by incorporating depth information from depth maps. This additional dimension allows more accurate prediction of object position and camera-relative movement, improving focus stability for close objects

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

Data Source

PatentUS9973681B2Method and electronic device for automatically focusing on moving object

Publication Date: 2018.05.15 SAMSUNG ELECTRONICS CO LTD

AI summary

A method of an electronic device for automatically focusing on a moving object is provided. The method includes generating, by a processor, at least one focal code based on information comprising depth information of the moving object obtained using at least one previous position of the moving object, focusing, by the processor, on at least one portion of the moving object based on the at least one focal code, and capturing, by a sensor, at least one image of the moving object comprising the at least one portion of the moving object.