Endra Pitowarno ©2007
Inside the Robotic Vision Dr. Ir. Endra Pitowarno, M.Eng
PENS-ITS Seminar “New Concept Robotics: Robot Vision” 22 Februari 2007
Universitas Gunadarma - Jakarta
Endra Pitowarno ©2007
Vision System • Vision (penglihatan): referensi terbaik adalah sistem mata manusia: how the brain controls the eyes and organizes the images. • Capturing/generating image: “Melihat & Menangkap Image” > adalah tugas utama kamera/mata. • Low-level Image Processing: menelisik fitur dan membedakan gambar obyek (yg dicermati) dengan latar belakang (yg tidak dicermati) > proses iconic
1
Endra Pitowarno ©2007
Vision System • Intermediate-level Image Processing: Proses berbasis algorithma/metoda pada image yg telah ditelisik fitur-fiturnya > proses iconic to symbolic • High-level Image Processing: Penerapan untuk keperluan apa image yg telah terdeskripsi ini (identifikasi/pembandingan, pengambilan keputusan berbasis image, diskritisasi/konversi data image ke satuan parameter siap hitung, dll.)
Endra Pitowarno ©2007
Current Issues
Speech Recognition
Robotic Vision
Stereo Vision
Visual Servoing Visual-based Tracking (gesture motion tracking, etc.)
2
Endra Pitowarno ©2007
Image Processing Applications • Industrial Inspection: Printed Circuit Board (PCB) inspection, VLSI circuit mask checking, etc. • Robot Vision: Autonomous mobile robot system, coordinating/swarm robot, humanrobot interaction (AIBO, ASIMO, etc) • Interactive Biomedical Equip. System • Etc.
Endra Pitowarno ©2007
Image Processing levels application
scene
Low-level Image (Iconic) Processing
Intermediate-level Image (Iconic to Symbolic) Processing
High-level Image Processing
Domain knowledge, thresholds, parameters, methods, etc.
3
Endra Pitowarno ©2007
(High Speed) Image Processing (Vision Sensor) for Robotics application
scene
Low-level Image (Iconic) Processing
Intermediate-level Image (Iconic to Symbolic) Processing
High-level Image Processing
Domain knowledge, thresholds, parameters, methods, etc.
Idealnya: < 0.1 detik
Endra Pitowarno ©2007
Low-level Image Processing • Mengubah sebuah image menjadi “image yg lain” • Frame selection (image quality): size/resize (memory based) • Filtering/converting: color mode conversion (color or B/W & mode), color balancing, brightness, contrast, hue, sharp, blur, etc. • Edge Detection:
• Semua langkah ditempuh dalam soft programming • Etc.
4
Endra Pitowarno ©2007
Contoh 1: Low-level Image Processing
Endra Pitowarno ©2007
Contoh 2: Low-level Image Processing
5
Endra Pitowarno ©2007
Contoh 3: Low-level Image Processing
Endra Pitowarno ©2007
Contoh 4: Low-level Image Processing
6
Endra Pitowarno ©2007
Intermediate-level Image Processing • Mendeskripsikan data image menjadi symbol yg kemudian dapat digunakan dalam high-level image processing sebagai komponen dasar untuk menyelesaikan tugas tertentu. • Proses menggunakan berbagai metoda (matematik): perspektif, trigonometri, 2D-3D system, etc. • Contoh: penggunaan algorithma A* atau Dijkstra untuk path planning berbasis image dari robot environment • Semua langkah ditempuh dalam soft programming
Endra Pitowarno ©2007
High-level Image Processing • Pemrosesan output simbol dari proses intermediate menjadi “suatu keputusan”: contoh, image marka jalan diproses menjadi nilai sensor posisi robot terhadap jalan. • Output proses high-level image processing siap diumpankan ke (high-level) control pada sistem robot. • Semua langkah ditempuh dalam soft programming
7
Endra Pitowarno ©2007
Contoh Kasus: Camera-based Line Tracer Perspective scene Jalur PUTIH di lantai GELAP
Keterangan : PA1 dan PA0 adalah motor Kiri & Kanan
Motor Kiri
PA1
kamera
Motor Kanan
PA 0
Camera-based Line Tracer Robot
Endra Pitowarno ©2007
Contoh: Line Identification (Low-level process)
8
Endra Pitowarno ©2007
Contoh: Line Identification (Low-level process) 400
ref. frame
0 400
Endra Pitowarno ©2007
Contoh: Line Identification (Low-level process) 400
0 400
9
Endra Pitowarno ©2007
Contoh: Line Identification (intermediate-level process) 210 190
400 rescale
1
1
1
0
0
1
1
1
1
1
1
0
10
Tabulasi & Deskripsi Aplikasi
Endra Pitowarno ©2007
Contoh: Line Identification (high-level process) 1
0
0
1
1
0
0
0
0
0
0
0
10 Klasifikasi Nilai Sensor
Aksi ke Motor Kiri-Kanan
D9
D8
D7
D6
D5
D4
D3
D2
D1
D0
Implementasi Fuzzy Rule
1
0
0
0
0
0
0
0
0
0
Banting stir ke kiri
1
1
0
0
0
0
0
0
0
0
Belok kiri banyak
0
1
1
0
0
0
0
0
0
0
Belok kiri agak banyak
0
0
1
1
0
0
0
0
0
0
Belok kiri sedikit
0
0
0
1
1
0
0
0
0
0
Lurus
0
0
0
0
1
1
0
0
0
0
Lurus
0
0
0
0
0
1
1
0
0
0
Belok kanan sedikit Belok kanan agak banyak
0
0
0
0
0
0
1
1
0
0
0
0
0
0
0
0
0
1
1
0
Belok kanan banyak
0
0
0
0
0
0
0
0
1
1
Banting stir ke kanan
10
Endra Pitowarno ©2007
Contoh: High Speed Vision System for Robotic • Color Tracking: Object • Obstacle Avoidance • Road mapping • etc.
Endra Pitowarno ©2007
END
11
