|
Ultra fast cnn based Hardware Computing Platform Concepts for adas visual Sensors and Evolutionary Mobile RobotsBog'liq Alireza FasihContents
List of Figures ......................................................................................................................................................... vi
List of Abbreviations ............................................................................................................................................. 1
1.
Introduction .................................................................................................................................................... 5
1.1
Motivation and general context ..................................................................................................... 5
1.2
Research questions and objectives of the thesis .................................................................... 7
1.3
Summary of the key contributions of the thesis .................................................................. 15
1.3.1
Scientific significance of the thesis
.................................................................................. 15
1.3.2
Practical significance of the thesis
................................................................................... 16
1.4
List of publications in the frame of this thesis ..................................................................... 16
2.
Requirements of ADAS concerning real-time computing for the image processing
based Sensors ........................................................................................................................................................ 18
2.1
Context and Motivation ................................................................................................................. 18
2.2
State of the art of real-time ADAS platforms ......................................................................... 20
2.3
Contribution for real-time ADAS ............................................................................................... 21
3.
Major limitations of traditional high performance computing concepts ............................ 23
3.1
Motivation and general context .................................................................................................. 23
3.2
State of the art in traditional processing method ............................................................... 24
3.3
Contribution of Ideal ADAS architecture ................................................................................ 25
4.
Potential of Neurocomputing including Cellular Neural Networks for ultrafast image
processing ............................................................................................................................................................... 30
4.1
Context and Motivation ................................................................................................................. 30
4.2
A survey of the related state-of- the art on image processing based on ANN. ........ 33
4.3
Contribution to an image processing platform for ADAS ................................................ 35
5.
CNN template calculation schemes with a particular focus on the learning/training
based approach through Genetic Algorithms .......................................................................................... 39
5.1
Introduction ........................................................................................................................................ 39
5.2
Genetic algorithm based template optimization for a vision system ......................... 41
5.2.1
General background .............................................................................................................. 42
5.2.2
Principles of Cellular Neural Network ........................................................................... 43
5.2.3
Genetic algorithms ................................................................................................................. 45
iv
5.2.4
Initial population for the genetic algorithm ................................................................ 45
5.2.5
Selection theory in the genetic algorithms................................................................... 46
5.2.6
Reproduction in the genetic algorithms ........................................................................ 46
5.2.7
Crossover and mutation in the genetic algorithms .................................................. 47
5.2.8
Fitness function in the genetic algorithm ..................................................................... 47
5.2.9
Obstacle detection through the developed concept ................................................. 48
5.3
Experimental results ....................................................................................................................... 52
5.4
Real-
time computing issues for the genetic algorithm based CNN template’s
calculations ........................................................................................................................................................ 55
6.
Emulation of analog computing on FPGA ........................................................................................ 58
6.1
Introduction ........................................................................................................................................ 58
6.2
New computational modeling for solving higher order ODE’s based on FPGA
...... 59
6.3
General background ........................................................................................................................ 59
6.4
HDL Description and system architecture for the analog computing emulation
concept of FPGA ............................................................................................................................................... 61
6.5
The “Digital Differential Analyzer” method
........................................................................... 63
6.6
Integration of hardware and software .................................................................................... 64
6.7
Experimental results ....................................................................................................................... 65
6.7.1
Future work .............................................................................................................................. 68
6.8
Concluding remarks ........................................................................................................................ 68
7.
Implementation of CNN on FPGA ........................................................................................................ 69
7.1
Introduction ........................................................................................................................................ 69
7.2
A framework for FPGA based real-time machine vision: direct convolution versus
CNN 70
7.3
Introduction to video processing platform............................................................................ 70
7.4
System Architecture ........................................................................................................................ 71
7.5
Hardware Platform Specification .............................................................................................. 72
7.6
Electronic System Level and High Level Direct Convolution Modeling ..................... 74
7.7
Modeling cellular neural network by DDA ............................................................................ 75
7.8
CNN Emulation Architecture ....................................................................................................... 77
7.9
Hardware and software integration ......................................................................................... 79
7.10 Direct convolution vs. CNN
–
performance comparison .................................................. 81
v
7.11 Conclusion ........................................................................................................................................... 82
8.
Implementation of CNN on GPU ........................................................................................................... 83
8.1
CNN Based High Performance Computing for Real Time Image Processing on GPU
84
8.2
Introduction ........................................................................................................................................ 84
8.3
Theory of Parallel Computing ..................................................................................................... 86
8.4
System Design and Architecture of CNN................................................................................. 88
8.5
System Diagram ................................................................................................................................ 90
8.6
Methodology of system design using OpenCL ...................................................................... 91
8.7
Conclusion ........................................................................................................................................... 94
9.
Cellular Neural Networks for Controlling an Unstructured Robot ........................................ 96
9.1
Introduction ........................................................................................................................................ 96
9.2
Cellular Neural Networks-Based Genetic Algorithm for Optimizing the Behavior of
an Unstructured Robot ................................................................................................................................. 97
9.3
Introduction ........................................................................................................................................ 98
9.4
Using genetic algorithms for CNN template optimization ............................................... 99
9.2
Training algorithm and simulation results ......................................................................... 104
9.3
Conclusion ........................................................................................................................................ 113
10.
Conclusion and Outlook ................................................................................................................... 115
References ........................................................................................................................................................... 119
|
| |
