• Share
  • Email
  • Embed
  • Like
  • Save
  • Private Content
Towards Embedded Computer Vision邁向嵌入式電腦視覺
 

Towards Embedded Computer Vision邁向嵌入式電腦視覺

on

  • 1,132 views

My slides for acamedia talk about embedded vision in 2010. Some of our research results are also presented in this presentation.

My slides for acamedia talk about embedded vision in 2010. Some of our research results are also presented in this presentation.
Few slides have chinese characters.

Statistics

Views

Total Views
1,132
Views on SlideShare
1,129
Embed Views
3

Actions

Likes
0
Downloads
48
Comments
0

1 Embed 3

http://www.linkedin.com 3

Accessibility

Categories

Upload Details

Uploaded via as Adobe PDF

Usage Rights

CC Attribution-NonCommercial-NoDerivs LicenseCC Attribution-NonCommercial-NoDerivs LicenseCC Attribution-NonCommercial-NoDerivs License

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment

    Towards Embedded Computer Vision邁向嵌入式電腦視覺 Towards Embedded Computer Vision邁向嵌入式電腦視覺 Presentation Transcript

    • 本著作採用創用CC 「姓名標示」授權條款台灣3.0版 Towards Embedded Computer Vision Wang, Yuan-Kai(王元凱)Electronic Engineering Department,Fu Jen Univ. (輔仁大學電機工程系) Email: ykwang@mails.fju.edu.tw URL: http://www.ykwang.tw 2010/05/14
    • 王元凱 Towards Embedded Computer Vision p. 2 Contents 1. Embedded Systems 2. Embedded Computer Vision 3. Entertainment Robot (CPU) 4. Embedded Vision Sensor (CPU) 5. Portable Vision Device (DSP) 6. Smart Video Surveillance (FPGA) 7. Summary & OutlookFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 3 1. Embedded SystemsFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 4 Evolution of Computer Past Now Future • Embedded System is a computer that is • Special-purpose • Light, Thin, Short, Small ⇒ Limited resourcesFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 5 Embedded Systems "Without" Sensors 資料來源:資策會MIC ITIS計畫整理Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 6 Embedded Systems "With" Sensors Wii Roomba GPS Exoskeleton NavigationFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 7 Embedded Systems "With" Image Sensors DARPA Augmented Grand RealityChallenge Surface IntelligentComputing RobotFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 8 2. Embedded Computer Vision (ECV)Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 9 What Is ECV  Embedded compute vision  Implements computer vision algorithms on low-cost, low-power, constrained hardware  Constrained hardware  Low-speed CPU  Low capacity memory  No floating-point (FPU)  Low-resolution image sensorFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 10 Embedded Computer Vision  Embedded System + Camera + Computer Vision Algorithm Image Image Image Capturing Processing RecognitionFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 11 Why Smart Camera (1/2)  Front-end processing  An example for video surveillance Classical stationary camera Smart cameraIOImage Inc.Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 12 Why Smart Camera (2/2)  In-node processing: Vision sensor network  Distributed vision system  Camera networks  Use multiple cameras to analyze the scene  Benefit  Less problems with occlusion  Challenge  Distributed processing and reasoningFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 13 International Activities (1/2)  Special conferences  IEEE Int. Conf. Distributed Smart Cameras  Special journal issues  IEEE Journal of Selected Topics in Signal Processing, vol. 2, no. 4, Aug. 2008  EURASIP Journal on Embedded Systems,  Short courses in important CV conferences  CVPR07&08: Distributed vision processing in smart camera networks  ESC07: Embedded CV and smart cameras  ICASSP09: Distributed processing in smart camerasFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 14 International Activities (2/2)  Research projects and Lab.  Princeton Univ./Georgia Tech.: Embedded Systems Lab., Wayne Wolf  Stanford Univ. Wireless sensor networks Lab.  UCLA, CMU, MIT  Delft Univ. of Technology SmartCam Project  Graz Univ. of TechnologyFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 15 Three Ways for ECV  CPU (Central Processing Unit)  ARM, PowerPC  DSP (Digital Signal Processor)  TI, ADI, NXP  FPGA (Field Programmable Gate Array)  Altera, XilinxFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 16 Embedded CPU  Embedded CPU = Low-power CPU  ARM  Drawbacks of embedded CPU for computer vision  No FPU, usually fixed-point  Speed: 60MHz ~ 600MHz  Therefore it is usually developed for (video) sensor networksFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 17 Embedded CPU: MeshEye  Stanford MeshEye (http://wsnl.stanford.edu/smartcam.html)  ARM 7 (55MHz), ZigBee node  3 image sensors "MeshEye:a hybrid-resolution  30x30 grayscale x 2 smart camera intelligent in distributed mote for applications  640x480 color x 1 surveillance", IPSN-SPOTS, 2007 Object detectionFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 18 Embedded CPU: CMUCam  CMU CMUcam3  ARM7  ECV applications  Robotic vision, color tracking, histogram processing, face detection (http://www.cmucam.org)Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 19 DSP  DSP is good for signal processing  SIMD structure for filtering processing  However, Computer Vision needs extreme DSP + video port  Media processors: powerful DSP  Parallelism: VLIW  Faster memory, DMA  Wide data bus  ECV applications  Face detection, face recognition, license plate recognition, vehicle tacking,Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 20 DSP : TRICam  "Visual surveillance on DSP-based embedded platform," Graz Univ. of Technology, 2008(Phd. dissertation)  TI C6414 (600MHz)  Applications: Adaboost face detection, vehicle detection, license plate detectionFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 21 FPGA (1/3)  Customizable hardware for parallelism  Reconfigurable computing  Flexible hardware design by HDL codes on FPGA.Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 22 FPGA (2/3)  "Hardware, Design and Implementation Issues on a FPGA-Based Smart Camera," ICDSC, 2007Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 23 FPGA (3/3)  ECV application: object tracking  Template matching by MAE (Maximum Absolute Error)Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 24 Hybrid: MCU+CPLD  UCLA Cyclops http://research.cens.ucla.edu/projec ts/2007/Multiscaled_Actuated_Sens  MCU: Atmega128 ing/Cyclops/  CPLD: image processing  ECV app.: Hand gesture recognitionFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 25 Hybrid: CPU+DSP (1/2)  "Distributed Embedded Smart Cameras for Surveillance Applications," IEEE Computer, 2006. Developed for traffic surveillanceFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 26 Hybrid: CPU+DSP (2/2)Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 27 SOC - Xetal  Philips: Xetal SIMD processor  InCa  WiCa  Xetal3 + 8051  Stereo sensors (640x480)  50 GOPS @ 600mWatt  ZigBee node  C++ programming  Used by Stanford Univ., Delft University of Technology  ECV Applications: edge detection, face detection, hough transform, gesture recognition, depth estimation, ...Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 28 Challenges for ECV  Algorithm refinement  Parallel computation  Function partition, Multi-threading  Stream processing  Memory hierarchy optimization  Hardware design  Pipeline, SIMD, board design  Optimized programming skills  Fixed-point arithmetic  Memory management  Intrinsic commandsFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 29 Our Current Results  Entertainment robot (CPU)  Sony AIBO robot with 64-bit RISC CPU  Application: Game playing, Face detection/recognition, Facial expression recognition  Smart vision sensor: (CPU)  Self-made sensor with ARM7  Application : Face detection, Robot  Portable vision device (DSP)  Self-made device with Dualcore DSP  Application: gesture recognition  Smart video surveillance (FPGA)  Background subtraction with FPGA  Application: video surveillanceFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 30 3. Entertainment Robot CPU SolutionFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 31 Two AIBO Models  Hardware ERS-7  CPU: 64-bit RISC 576 MHz/192Mhz  RAM: 64MB/32MB  Flash: 32MB  20 motors  Camera:  CMOS sensor  Resolution: 280 × 160 ERS220A  10 fpsFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 32 SensorsFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 33 Software  Operating System  Aperios (Embedded Linux)  Development tools  C++  GCC 3.4.4 on Linux  Libraries: OpenR, TekkotsuFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 34 Our AIBO Pet  Rolling Dice  Face Detection  Face Recognition  Facial Expression RecognitionFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 35 Rolling Dice (1/3)Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 36 Rolling Dice (2/3)  Detect the dice by color detection  Using Gaussian mixture model (GMM) and EM algorithm to model colors p( x | c) = 1 N 1 ( − ( x − mi )T ∑i−1 ( x − mi )) ∑ω i =1 i 2π ∑i 1/2 e 2 Original Detection image result GMM Morphology+CCLFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 37 Rolling Dice (3/3)Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 38 Face Detection (1/2)Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 39 Face Detection (2/2)  Algorithm: Adaboost face detection  Proposed by Viola and Jones in 2001  Cascaded weak classifiersFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 40 Face Recognition (1/3)Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 41 Face Recognition (2/3)  Eigenface approach (PCA)Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 42 Face Recognition (3/3)Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 43 Facial Expression Recognition 1/3Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 44 Facial Expression Recognition 2/3Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 45 Facial Expression Recognition 3/3  3 expressions  Happy, Surprise, Angry  Video-based method  Feature: optical flow, common vector flow  Classifier: Hidden Markov model  Well done in small-resolution images  140 * 120 (~ QCIF, 176 * 144)Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 46 4. Embedded Vision Sensor CPU SolutionFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 47 FJUCam  It is a self-made camera  Camera module + embedded system  3S: Small, Smart, Sensing  Components of the FJUCam  ARM7 TDMI  32-bit, 60MHz, 64KB RAM, 128KB ROM  CMOS sensor: OV6620 (CIF 50fps)  CIF(352x288), QCIF, 8-bit RGBFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 48 What Is FJUCam (1/3)• Weight: 35gm• Power sources: •Size: • 5V DC current 6 x 4.5 x 5 (cm) • 5V Cell Battery (W x H x D)• Power consumption: 1W Three Modules 1. Main board, 2. Lens module 3. Storage moduleFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 49 What Is FJUCam (2/3)Power Main Board Button/ISP Serial portSwitch ARM7 MicrocontrollerFrame buffer/FIFO Front Side Other interfaces: RS232x2, SPI, I2C, GPIO Back Side 49Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 50 What Is FJUCam (2/3) Lens Module Storage ModuleFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 51 Software  Development environment  C Language  PC Windows + Cygwin + GCC  cc3 library (open source developed by CMU)Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 52 Face Detection  The Adaboost algorithm for face detection  Proposed by Viola and Jones in 2001  Cascaded weak classifiers(21 cascades)  Algorithm refinement  Reduced to 5 cascades  Fixed-point arithmetic  Stream processing for only 64KB memory utilization Image FJUCam Display Face DetectionFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 53 Cyclops Robot  Color trackingFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 54 5. Portable Vision Device: X-EYE DSP SolutionFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 55 Goal of X-Eye 55Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 56 Components 觸控面板 顯示器 移動電源 自製 Camera USB 外殼 USB 筆電 連接線 Hub BeagleBoard 微投影機 SD卡 USB-WIFI卡 鍵盤 USB-RS232 讀卡機Fu Jen University 控制線 Department of Electrical Engineering 滑鼠 Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 57 1st Generation PrototypeFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 58 Photo Mode Capture Command: capture images Switch Command: Mode switchFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 59 Manage Mode Original Photos Next Command Previous Command Switch Command 2(to photo)Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai 59
    • 王元凱 Towards Embedded Computer Vision p. 60 Hardware (1/2) USER RESET OMAP3530 ProcessorPeripheral I/O •600MHz Cortex-A8 •NEON+VFPv3•USB Host •16KB/16KB L1 •256KB L2•JTAG •430MHz C64x+ DSP •32K/32K L1•DVI-D video out •48K L1D •32K L2•S-Video out •Power VR SGX GPU •64K on-chip RAM•SD/MMC+ POP Memory •256MB LPDDR RAM•Stereo in/out •256MB NAND flash•RS-232 serial1•Alternate power•USB 2.0 HS OTG 7.6 cmFu Jen University 60 Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 61 Hardware (2/2)Fu Jen University Department of Electrical Engineering 2010.04.25 Wang, Yuan-Kai 61
    • 王元凱 Towards Embedded Computer Vision p. 62 Software (1/2)Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 63 Software (2/2) 軟體名稱 版本 功用簡述 Gesture Command 1.0 手勢辨識 Module OpenCV 1.0 影像處理 FFMpeg 0.5.1 視(音)訊邊解碼 QT 4.6.2 視窗介面 WinXP下安裝Linux VMWare 6.5.3 工具 Ubuntu (host) 9.04 安裝交叉編譯環境 Ubuntu (client) 9.10 BB上的filesystem Kernel (client)Fu Jen University 2.6.29 Department of Electrical Engineering BB上的kernel 2010.04.25 X-Eye Wang, Yuan-Kai 63
    • 王元凱 Towards Embedded Computer Vision p. 64 Algorithm (1/2)  Gaussian Mixture Model (GMM)  Model four colors 1 N 1 ( − ( x − mi )T Covi−1 ( x − mi )) p ( x | c) = ∑ ωi e 2 i =1 2π || Covi ||1/2  Expectation Maximization (EM)  Parameter estimation of GMM E Step M Step N 1 ∑ E(z N 1 ω p( x j | m , C ) ωit +1 = t +1 ) m = ∑ E(z t t t )x j E ( zij ) = Nω t +1 i i i ij i ij i N j =1 i j =1 ∑ ω tp p( x j | mtp , C tp ) =C t +1 1 t +1 ∑ N E ( zij )[( x j − mit +1 )( x j − mit +1 )T ] p =1 N ωi i j =1Fu Jen University 64 Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 65 Algorithm (2/2)  Color Identification  c ˆ = arg max P( x | c j ), j 1 ~ k cj  Performance optimization by Look Up Table (LUT) for real-time  Gesture Recognition  Four gestures: capture, switch, next, previous,Fu Jen University 65 Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 66 6. Smart Video Surveillance FPGA SolutionFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 67 Our Experience Background subtraction, ... • 2.8 GHz Intel CPU • Software: C/C++ FPGA • Frame rate: 10 fpsFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 68 Background Subtraction Current Frame Bk + 1 BackgroundM k +1 ( x, y ) P k+1 Image Update Background Image= Pk +1 ( x, y ) − Bk ( x, y ) - Bk M k + 1 Bk +1 ( x, y ) = αBk ( x, y ) + (1 − α ) Pk +1 ( x, y ) Post Processing Motion Object Image Speed up by (1) Circuit design, (2) ParalizationFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 69 Background Subtraction by FPGA (1/3)  Parallelism: 7-level pipeline  SIMD with stream processingFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 70 Background Subtraction by FPGA (2/3)  Hardware: Altera Cyclone II 2C35  Design: Verilog HDL with Quartus II Tools Background New Frame Result Frame rate • Background module : 368 fps • Whole system : 51 fpsFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 71 Background Subtraction by FPGA (3/3)  Comparisons  PC: 2.8GHz CPU, C implementation  FPGA can speed up 500 times 2.8G 51 CPU FPGA 25M 10 Clock(Hz) FPSFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 72 7. Summary and Future Research DirectionsFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 73 Summary (1/2)  Embedded CPU is not appropriate for computer vision  Although CPU has great flexibility and programming environment  Its architecture is interrupt-based  Designed for I/O-process usage  Not for data-intensive computing, such as DSP and image/video processingFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 74 Summary (2/2)  High-performance processor is necessary for computer vision  Clock rate is not the crucial point  But SIMD and algorithm parallelism doFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 75 Questions  Embedded compute vision  Low-cost, low-power, Constrained minimal hardware Resource  High-Performance Contradiction computer vision  Fast speed without cost, Abundant power, and hardware Resource constraints From contradiction to convergence ?! Yes by multicoreFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 76 Challenges (1/2)  Algorithm decomposition  Function decomposition  Partition serial part and parallel part  Data flow analysis and data dependency analysis  Parallelism  Loop unrolling  Multithreading  PipelineFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 77 Challenges (2/2)  Performance analysis method  For efficiency improvement  Implementation efforts  Choose a good embedded platform for computer vision  Software issues  Hardware issues  Programming skills  Multi-threadingFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 78 Future Research Directions  Multicore framework  DSP + CPU : FJUCam2  Advantage: Using C language  Challenge: algorithm parallelism  FPGA + CPU:  Advantage  Reconfigurable multicore  Less Verilog, more C  Challenge: hardware/software co-design  GPGPU  Advantage:  240~512 cores  Using C language  Challenge: algorithm parallelismFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 79 FJUCam2 (1/2)  Next generation FJUCam (MPSoC)  Adopt multicore technology  DSP(Media) + CPU 99/6 Image RAM Resolution 128MB FJUCam2 VGA 98/2 64KB FJUCam1 CIF Processor 60MHz 600MHzFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 80 FJUCam2 (2/2)  Algorithms going to be developed for FJUCam2  Color tracking  Gesture recognition  Face tracking and recognition  Event detection  Video summarization  Sleep monitoring  Distributed vision processingFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 81 Reconfigurable Multi-Core  FPGA + CPU + Linux Hardware Software DE2_70 FPGA Chip CMOS Nios II Processer Controller Background RGB to Y Subtraction Avalon Bus RAW Flash SDRM1 SSRAM DM9000A to Morphology Controller Controller Controller Controller RGB CMOS CCD SDRAM0 Flash SDRAM1 SSRAM DM9000A Capture Internet VGA PCFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 82 The End Free for Any QuestionsFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 83 Our Development Boards Xscale270 Beagle board TI DSP DaVinci 6446 Altera DE2-70 Celoxica RC10+DKFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 84 Robotic VisionFu Jen University Department of Electrical Engineering Wang, Yuan-Kai
    • 王元凱 Towards Embedded Computer Vision p. 本簡報授權聲明  此簡報內容採用 Creative Commons 「姓名標示 - 非商 業性台灣 3.0 版」授權條款  歡迎非商業目的的重製、散布或修改本簡報的內容,但 請標明: (1)原作者姓名:王元凱; (2)圖標示:  簡報中所取用的部份圖形創作乃截取自網際網路,僅供 演講者於自由軟體推廣演講時主張合理使用,請讀者不 得對其再行取用,除非您本身自忖亦符合主張合理使用 之情狀,且自負相關法律責任。Fu Jen University Department of Electrical Engineering Wang, Yuan-Kai