Posts

Showing posts from 2011

Outliers filtering – ROI extraction

Image
q removing the outlier features q extract the region of interest by computing a bounding-box rectangle around the inlier features.

Visual appearance model of an unknown object based on the data collected with an RGB camera observing the robot’s hand while moving the object

T u t o r i a l

If you are interested in obtaining more informations about in hand object modeling please have a look over the next tutorial. You can find the source code and also a briefly description about how to run the program . http://www.ros.org/wiki/model_completion

OpenGL robot server visualization

Image

System Overview

Image

Region of interest extraction

Image

Kinect style

Image
Over the last decade, the availability of public image repositories and recognition benchmarks has enabled rapid progress in visual object category and instance detection. Today we are witnessing the birth of a new generation of sensing technologies capable of providing high quality synchronized videos of both color and depth, the RGB-D (Kinect style) camera. With its advanced sensing capabilities and the potential for mass adoption, this technology represents an opportunity to dramatically increase robotic object recognition, manipulation, navigation, and interaction capabilities. Recently, sensors combining RGB images with depth measurements (RGB-D sensors) have come to prominence due to their gaming applications and in particular due to the release of the Xbox 360 Kinect (Microsoft, 2010). Such sensors are now both very affordable (around $150) and readily available, making them ideal for personal robotics applications.

Testing the detection and recognition of objects

Image
Recognizing and manipulating objects is an important task for mobile robots  performing useful services in everyday environments. While existing techniques  for object recognition related to manipulation provide very good results even for  noisy and incomplete data, they are typically trained using data generated in an offline  process.Specifically, we develop an approach to building a 3D surface model of an unknown object based on data collected by a kinect sensor observing the robot’s hand moving the object.

PR2 in action !

Image

Enought theory, time to get the hands dirty!

Image
http://www.youtube.com/watch?v=Hjwj0YN2z5w

About vocabulary tree

Image
A vocabulary tree is a hierarchical-structured tree that can efficiently integrate quantization and classification . Object recognition is one of the core problems in computer vision, and it is a very extensively investigated topic. Due to appearance variabilities caused for example by non-rigidity, background clutter, differences in viewpoint, orientation, scale or lighting conditions, it is a hard problem.  One of the important challenges is to construct methods that scale well with the size of the database, and can select one out of a large number of objects in acceptable time. For each cluster a vocabulary tree is build for searching the most similar view . Vocabulary tree is usually used to quantize SURF into more compact features

About OpenCV

Image
OpenCV (Open Source Computer Vision) is a library of programming functions for real time computer vision. SIFT Image matching is a fundamental aspect of many problems in computer vision, including object or scene recognition, solving for 3D structure from multiple images, stereo correspondence, and motion tracking.

My project : Objects Modelling

The ability to recognize objects and to localize them precisely is essential in all service robotic applications. One of the main challenges for service robots during operation lies in the handling of unavoidable uncertainties which originate from model sensor inaccuracies are characteristic for realistic application scenarios. ODUfinder is a perception system for autonomous service robots acting in human living environments. The perception system enables robots to detect and recognize large sets of textured objects of daily use .We need a robot object detection and recognition  system that can recognize thousands of objects by learning and using vocabulary three of SIFT descriptors . The usage of robots to aid humans is becoming more and more widespread, typically in the industry, but increasingly also in public services and in some cases, home applications too. Robots are starting to be more ...

Initial thoughts

The story began on February 01,2011 when I arrived in Munich. It was a rainy day and someone was waiting us at the airport. The first lesson: how to arrive at the Konigsplatz from Studentenstadt and vice versa. The best solution: the subway. I met new people from the lab  (Cotesys Central Robotics Laboratory)  who helped me to acquaint with my new home for the next months. I was very excited about all that happened to me. Two months have passed since I arrived here and I am still surprised how quickly the time flies ...