Ph. D. Project
Title:
Monitoring of migratory birds in the Annaba/Tarf wetland (Algeria) by deep-learning through wireless multimedia sensor networks
Dates:
2021/04/29 - 2024/12/16
Student:
Supervisor(s): 
Other supervisor(s):
Pr. DOGHMANE Noureddine (ndoghmane@univ-annaba.org) , Dr HARIZE Saliha (shrz.dj@gmail.com)
Description:
I. Description of the research problem
Birds respond to possible changes in the environment and are good indicators of biodiversity. If some bird
species begin to decline, this can reveal serious problems both in the food chain and in environmental
safety and can even affect public health. To this end, monitoring of fauna and flora in general and birds in
particular can provide a lot of valuable information on: the environment, land health and productivity,
...etc. The monitoring of these birds in their natural habitats is therefore of several interests:
- On the one hand, to have a reliable indicator to regularly assess the evolution of the number of
endangered birds in a region. Indeed, monitoring the size of bird populations is a common activity for
many ornithologists, thus making it possible to highlight possible environmental or anthropic changes that
negatively affect the species. Therefore, estimating the size of bird populations remains an essential
prerequisite for the conservation and management of bird species.
- On the other hand, birds can be considered as a tool for environmental and even public health
monitoring, for example by forecasting the possible spread of animal diseases such as avian influenza.
Indeed, the avian influenza virus (H5N1) in migratory birds, even low pathogenic, once transmitted to
domestic animals such as poultry, can evolve into highly pathogenic strains that thus constitute a
significant risk to public health and also lead to significant economic losses.

II. Description of the objectives of the thesis
Monitoring of birds in their natural habitat consists of regular observations and measurements that would
allow protective measures to be taken if adverse changes are observed. Habitat degradation, human
activity, environmental factors, climate change and forest fires would be the most significant threats to
bird species in general. The survival of our environmental system, both its fauna and flora, is a paramount
stake and a challenge to be taken up. But often this surveillance is carried out manually during campaigns
limited in time and space.
This thesis aims at an automatic continuous monitoring based on Wireless Multimedia Sensor Networks
(WMSN) of these birds in their natural habitat. The idea is to identify, recognize and count the number of a
bird species (considered endangered) based on their vocalizations (songs and cries) and/or their
photographs. Sensor networks prove to be ideal platforms for recording and processing such data due to
their characteristics' conformity to project requirements such as energy independence, average financial
cost, wide geographical coverage and environmental preservation. However, the proposed monitoring
method, based on the WSN, has to meet several major challenges specified below that we propose to
solve through this thesis.

III. Work plan of the thesis
1. Preliminary work: Position of the problem, State of the art, Delimitation of the geographical area
targeted by the study, Identification of the species of birds subject to the project, Description of the
biometric characteristics of the birds targeted by the study, ...etc.
2. Detection of birds by their vocalizations : Detect good bird specimens before taking photographs. This
can be done using a scalar sensor (microphone) associated with the sensor node. It must allow to
recognize the specimen thanks to its vocalization (songs/scries). Each bird specimen, in particular the two
migratory birds of this study like the Bald Duck (Oxyura leucocephala) and the Ferruginous Duck (Aythya
nyroca), have particular calls and songs that can be used in their detection and recognition. Nevertheless,
we will be confronted with several difficulties that we must overcome. We therefore recommend to work
in this axis and to adapt recognition techniques based on CNN Convolutional Neural Networks in particular
and deep-learning in general to bird sounds in order to be able to detect and recognize the targeted bird
species despite the different difficulties involved.
3. Detection of birds through their photographs: Implement and estimate models for the detection of
birds of endangered species, based on deep-learning using photographs acquired and transmitted by
WMSNs.
4. Optimization of communication resources in relation to the bird recognition method: Adapt the
communication architecture in terms of number of nodes, bandwidth. Reduce the power consumption of
the controller and the sensor node transmission unit. This to increase the lifetime of the sensor node. This
part will be evaluated using network simulation tools before experimentation.
5. Evaluation of the system in a realistic context: Evaluate the performance of the sensor nodes used in a
realistic context. This can be done by implementing a testbed and/or by using adequate platforms
(wireless network simulators for example).
Keywords:
Bird recognition, WMSN, IoT, Energy Consumption, image compression, Deep-learning
Department(s): 
Modeling and Control of Industrial Systems