Tuesday, August 22, 2006

BirdSpot - 4th place @ WESC 2006

Project name: BirdSpot
Contest: WESC 2006
Award: 4th place
Alpha Team: Alin-Iulian Lazar, Andrei Gheorghe, Mihai Ciureanu, Radu Nedelcut
Mentor: Nicolae Tapus
(raport)

In the news:




Abstract:


The Danube Delta located in Eastern Romania is one of the best preserved natural habitats in the world and is on the list of the UNESCO World Heritage Sites and Biosphere Reserves [1]. Besides a wide variety of plants and fish, the Delta is also home to over 300 different bird species, some of them being very rare or unique such as the cormorant or the white pelican [2]. Studying these birds without too much anthropic interference is thus an important element in conserving their habitat and sustaining growth. Furthermore, as millions of birds come to the Delta every year to lay their eggs, several cases of avian influenza have been detected in autumn 2005, the virus being brought by some species coming from Asia [3][4]. Identifying these species as soon as they arrive and monitoring the high risk areas has become vital in order to prevent the virus from spreading and thus affecting the health of both wildlife and human beings or having harsh economic consequences.
The BirdSpot system automatically detects, identifies and classifies bird species based on visual information in order to determine the evolution of the population density of distinct species in a designated area. The system is built upon non-obtrusive, energy-efficient and affordable wireless devices that can be placed in remote natural habitats to gather visual data. BirdSpot then uses wireless long range communication equipment to aggregate data from these remote devices to a processing server. Bird detection and classification are accomplished by means of intelligent image processing and adaptive machine learning algorithms. As the images are analyzed, the output results consisting of the identified bird species and their locations are integrated into a database and are accessible via a user-friendly web interface. Furthermore, these results can be used in order to create specific reports. For instance, in case the arrival of a migrating virus-susceptible species is detected, a special alert will be issued to the interested authorities.
BirdSpot improves on existing imaging equipment used by bird-watchers or scientists: BirdSpot devices are designed to be autonomous and operate unattended in remote areas. Moreover, the key feature of the system is that it allows automatic, real-time classification of birds into species starting from relevant captured images. Tracking animals and birds in particular currently involves tagging individual specimens using rings or cumbersome electronic transmitters [5]. What BirdSpot focuses on however is not following the habits of individual birds but monitoring and synthesizing the general evolution of the density of species over a certain territory.
Thus, as it receives input from a number of remote devices deployed in an area, the BirdSpot system is able to compute and offer a general picture of the distribution of bird species. This can greatly facilitate and streamline the work of ornithologists. Current methods for assessing bird populations require the participation of hundreds of volunteers who must follow tedious procedures [6]. BirdSpot, on the other hand, can provide similar information with minimal supervision and is able to process statistical and geographic results automatically. The human contribution is reduced to collaborative assistance in training an automated species classifier.
The system is also designed with the numerous hobbyists and enthusiasts in mind as BirdSpot can be a fun way to take part in an exciting and educative birdwatching experience. However the main features of BirdSpot primarily qualify it as a powerful scientific tool and in the context of the recent birdflu threat as a potential means of protecting lives.

No comments:

Post a Comment