Yubo - Revision history

imported>Student221: /* Introduction */

2019-12-19T23:36:51Z

Introduction

← Older revision		Revision as of 23:36, 19 December 2019
Line 8:		Line 8:

	This project attempts to provide a virtual focuser implementation and a basic autofocus algorithm to evaluate. There are many different focuser types, many of which have been covered in the lectures. There are many benefits to creating a virtual focuser. First, it allows the tweaking of physical focuser characteristics, without having to rely on an abundance of physical focuser modules. Second, it abstracts away the reliance on hardware when evaluating autofocus algorithms. This enables camera software developers to perform system and regression tests on a cloud environment without having to set up a static and controlled scene.		This project attempts to provide a virtual focuser implementation and a basic autofocus algorithm to evaluate. There are many different focuser types, many of which have been covered in the lectures. There are many benefits to creating a virtual focuser. First, it allows the tweaking of physical focuser characteristics, without having to rely on an abundance of physical focuser modules. Second, it abstracts away the reliance on hardware when evaluating autofocus algorithms. This enables camera software developers to perform system and regression tests on a cloud environment without having to set up a static and controlled scene.


	The github for the project can be found at: https://github.com/YuboHan/FocusSimulator		The github for the project can be found at: https://github.com/YuboHan/FocusSimulator

imported>Student221: /* Introduction */

2019-12-19T23:36:36Z

Introduction

← Older revision		Revision as of 23:36, 19 December 2019
Line 8:		Line 8:

	This project attempts to provide a virtual focuser implementation and a basic autofocus algorithm to evaluate. There are many different focuser types, many of which have been covered in the lectures. There are many benefits to creating a virtual focuser. First, it allows the tweaking of physical focuser characteristics, without having to rely on an abundance of physical focuser modules. Second, it abstracts away the reliance on hardware when evaluating autofocus algorithms. This enables camera software developers to perform system and regression tests on a cloud environment without having to set up a static and controlled scene.		This project attempts to provide a virtual focuser implementation and a basic autofocus algorithm to evaluate. There are many different focuser types, many of which have been covered in the lectures. There are many benefits to creating a virtual focuser. First, it allows the tweaking of physical focuser characteristics, without having to rely on an abundance of physical focuser modules. Second, it abstracts away the reliance on hardware when evaluating autofocus algorithms. This enables camera software developers to perform system and regression tests on a cloud environment without having to set up a static and controlled scene.

			The github for the project can be found at: https://github.com/YuboHan/FocusSimulator

	== Related Lecture Material ==		== Related Lecture Material ==

imported>Student221: /* Results */

2019-12-13T09:59:39Z

Results

← Older revision		Revision as of 09:59, 13 December 2019
Line 235:		Line 235:
	We can vary the barrel speed from 1/10 of the barrel length to the full barrel length per capture, to see how long the algorithm takes to converge.		We can vary the barrel speed from 1/10 of the barrel length to the full barrel length per capture, to see how long the algorithm takes to converge.

	[[File:YuboBarrelSpeedGraph.png]]		[[File:YuboBarrelSpeedGraph.png\|500px]]

	From the graph, we can see that there is no direct correlation between captures to converge and barrel speed, although there is a slight downwards trend as barrel speed increases. The random spread of the convergence speed is most likely due to luck: if the algorithm can hit the highest focus score without having to adjust, then it will converge quickly. Otherwise, it may take time to fine tune itself. The general downwards trend of the graph is expected: the faster the algorithm can finish the sweep, the faster it can converge. Using this logic, it should be expected that points above 0.005 m/capture should converge in similar captures, as it takes all barrel speeds the same time to sweep the barrel (2 captures). However, the data is too random to draw concrete conclusions.		From the graph, we can see that there is no direct correlation between captures to converge and barrel speed, although there is a slight downwards trend as barrel speed increases. The random spread of the convergence speed is most likely due to luck: if the algorithm can hit the highest focus score without having to adjust, then it will converge quickly. Otherwise, it may take time to fine tune itself. The general downwards trend of the graph is expected: the faster the algorithm can finish the sweep, the faster it can converge. Using this logic, it should be expected that points above 0.005 m/capture should converge in similar captures, as it takes all barrel speeds the same time to sweep the barrel (2 captures). However, the data is too random to draw concrete conclusions.

imported>Student221: /* Results */

2019-12-13T09:58:59Z

Results

← Older revision		Revision as of 09:58, 13 December 2019
Line 245:		Line 245:
	We can also see the relationship between the scene distance vs. convergence speed by varying the scene distance from 0.5m to 1000m.		We can also see the relationship between the scene distance vs. convergence speed by varying the scene distance from 0.5m to 1000m.

	[[File:YuboSceneDist1.png]]		[[File:YuboSceneDist1.png\|500px]]

	From the information in figure 2, the convergence speed above 100m appears to be the same. We can now zoom in to where the scene distance is closer to the camera.		From the information in figure 2, the convergence speed above 100m appears to be the same. We can now zoom in to where the scene distance is closer to the camera.

	[[File:YuboSceneDist2.png]]		[[File:YuboSceneDist2.png\|500px]]

	The results are mostly expected. Since the focuser always start sweeping from a distance closest to the center, the start position is always focused at infinity (lens is at focal point). This means that if the focus is at infinity, then the lens need to sweep always to the end before returning. If the object is at a macro distance, once the sweep is done the focuser will already be close to the object, thus making convergence faster.		The results are mostly expected. Since the focuser always start sweeping from a distance closest to the center, the start position is always focused at infinity (lens is at focal point). This means that if the focus is at infinity, then the lens need to sweep always to the end before returning. If the object is at a macro distance, once the sweep is done the focuser will already be close to the object, thus making convergence faster.

imported>Student221: /* Related Lecture Material */

2019-12-13T09:58:06Z

Related Lecture Material

← Older revision		Revision as of 09:58, 13 December 2019
Line 23:		Line 23:
	A neat in-class demonstration was shown to demonstrate the idea of phase detection autofocus:		A neat in-class demonstration was shown to demonstrate the idea of phase detection autofocus:

	[[File:lecture2PhaseDetection.png]]		[[File:lecture2PhaseDetection.png\|500px]]

	An image is generally more focused if it contains higher frequency components. As shown in lecture 1, this project will utilize the Discrete Fourier Transform of the image. We will implement an algorithm to weigh the DFT such that the final 'focus score' will be higher, if the image contains more higher frequency signals.		An image is generally more focused if it contains higher frequency components. As shown in lecture 1, this project will utilize the Discrete Fourier Transform of the image. We will implement an algorithm to weigh the DFT such that the final 'focus score' will be higher, if the image contains more higher frequency signals.

imported>Student221: /* Related Lecture Material */

2019-12-13T09:57:41Z

Related Lecture Material

← Older revision		Revision as of 09:57, 13 December 2019
Line 11:		Line 11:
	== Related Lecture Material ==		== Related Lecture Material ==

	Lecture 1 defines a scene to be in focus when it satisfies the lens maker equation (thin lens approximation):		Lecture 1 defines a scene to be in focus when it satisfies the lens maker equation (thin lens approximation).
	[[File:lecture1Lensmaker.png\|500px\|thumb\|right\|Focus Lecture Slide]]
			[[File:lecture1Lensmaker.png\|500px\|thumb\|right\|'''Figure 1:''' Focus Lecture Slide]]

	When the object is out of focus, the 'blurriness' of the scene is affected by the aperture size. The area of the blur itself is known as the circle of confusion.		When the object is out of focus, the 'blurriness' of the scene is affected by the aperture size. The area of the blur itself is known as the circle of confusion.

	[[File:lecture1CoC.png]]		[[File:lecture1CoC.png\|500px\|thumb\|right\|'''Figure 2:''' Circle of Confusion Lecture Slide]]

	There are many ways to adjust focus. In lecture one, we discussed the light field camera. With this method, the camera is able to adjust focus after the image has taken, and thus does not require an autofocus algorithm to be run while we are taking the image. Lecture two covers two more techniques in adjusting focus by physically moving the lens towards and away from the sensors: VCM, and MEMs technology. The lectures also mention a method of autofocusing called Phase Detection. This project however, will be implementing another method of auto-focus called Contrast detection. The main difference is that phase detection utilizes the phase of incoming rays, whereas contrast detection only adjusts focus based on pixel sensor information.		There are many ways to adjust focus. In lecture one, we discussed the light field camera. With this method, the camera is able to adjust focus after the image has taken, and thus does not require an autofocus algorithm to be run while we are taking the image. Lecture two covers two more techniques in adjusting focus by physically moving the lens towards and away from the sensors: VCM, and MEMs technology. The lectures also mention a method of autofocusing called Phase Detection. This project however, will be implementing another method of auto-focus called Contrast detection. The main difference is that phase detection utilizes the phase of incoming rays, whereas contrast detection only adjusts focus based on pixel sensor information.

imported>Student221: /* Related Lecture Material */

2019-12-13T09:56:52Z

Related Lecture Material

← Older revision		Revision as of 09:56, 13 December 2019
Line 12:		Line 12:

	Lecture 1 defines a scene to be in focus when it satisfies the lens maker equation (thin lens approximation):		Lecture 1 defines a scene to be in focus when it satisfies the lens maker equation (thin lens approximation):
	[[File:lecture1Lensmaker.png\|500px]]		[[File:lecture1Lensmaker.png\|500px\|thumb\|right\|Focus Lecture Slide]]

	When the object is out of focus, the 'blurriness' of the scene is affected by the aperture size. The area of the blur itself is known as the circle of confusion.		When the object is out of focus, the 'blurriness' of the scene is affected by the aperture size. The area of the blur itself is known as the circle of confusion.

imported>Student221: /* Related Lecture Material */

2019-12-13T09:56:06Z

Related Lecture Material

← Older revision		Revision as of 09:56, 13 December 2019
Line 12:		Line 12:

	Lecture 1 defines a scene to be in focus when it satisfies the lens maker equation (thin lens approximation):		Lecture 1 defines a scene to be in focus when it satisfies the lens maker equation (thin lens approximation):
			[[File:lecture1Lensmaker.png\|500px]]
	[[File:lecture1Lensmaker.png]]

	When the object is out of focus, the 'blurriness' of the scene is affected by the aperture size. The area of the blur itself is known as the circle of confusion.		When the object is out of focus, the 'blurriness' of the scene is affected by the aperture size. The area of the blur itself is known as the circle of confusion.

imported>Student221: /* Appendix A - Robustness Test Results */

2019-12-13T09:54:01Z

Appendix A - Robustness Test Results

← Older revision		Revision as of 09:54, 13 December 2019
Line 274:		Line 274:

	== Appendix A - Robustness Test Results ==		== Appendix A - Robustness Test Results ==

			See https://docs.google.com/document/d/1dbuxbXM_8jDIH43FmakniGnJeYNKMcgXWWu7I-0OnQM/edit?usp=sharing for Appendix A

imported>Student221: /* References */

2019-12-13T09:52:59Z

References

← Older revision		Revision as of 09:52, 13 December 2019
Line 270:		Line 270:

	2. “android.hardware.camera2.CaptureRequest,” Android Developer Documentation. [Online]. Available: https://developer.android.com/reference/android/hardware/camera2/CaptureRequest.html#CONTROL_AF_MODE. [Accessed: 03-Dec-2019].		2. “android.hardware.camera2.CaptureRequest,” Android Developer Documentation. [Online]. Available: https://developer.android.com/reference/android/hardware/camera2/CaptureRequest.html#CONTROL_AF_MODE. [Accessed: 03-Dec-2019].

			Additional references from the internal Nvidia Camera Architecture designs, brownbags, presentations used but not referenced for NDA purposes.

	== Appendix A - Robustness Test Results ==		== Appendix A - Robustness Test Results ==