The addition of some professional experience is critical for maximizing the return on education. Internships are a great way to gain professional experience and can serve as a springboard for long-term professional success. Towards the end of my first semester of my master, I was looking for an internship opportunity to meet the requirements set forth by EM CDE for completion of the course and to expand my work experience. With the goal of finishing, it by the end of the second semester and before moving on to my specialization track in France, I looked for several organizations that work in the geospatial domain.
During my search, I learned about many organizations working in the geospatial domain over Europe. However, because of the pandemic, I wanted to stay in Austria and thus I applied for an internship at Spatial Service, which was well-known in Europe for its contributions in the field of geoinformatics. When applying for the internship I knew Spatial Service as a spin-off from the University of Salzburg that develops and implement geoinformation products and service. I was encouraged to forward my request after attending a presentation from Peter Zeil, Gina Schwendemann and Ahmad Alobaidi about their contribution to humanitarian action during a conference session in the Geo-Humanitarian Action course.
Spatial Services GmbH is a service and consulting company located within Salzburg University, with a tagline of "Geoinformatics Meets Technology: We are spatial experts". They aim to develop and implements geoinformation products and services aiming at resource-efficient and sustainable development in support of climate change adaptation, stable livelihoods, secure energy supply as well as smart city and mobility planning. They focus on Earth Observation data exploitation, sensor system and integration of spatial indicators. As a spin-off from the university, the company represents a link to research results from the Interfaculty Department for Geoinformatics, Z_GIS and incorporated their developments for creating improved user-oriented products.
When interview by Hubert Schöndorfer and Ahmad Alobaidi, I was given an opportunity to express my field of interest and skills in GIS-related applications and programming language. Also, I expressed my willingness to work on any project that would help me improve my understanding of EO and GIS. From March 1, 2021, to July 24, 2021, I was hired as a paid part-time intern for Spatial Service GmbH for a humanitarian related project. As required, I completed the total working hours that correspond to the normal time equivalent to 8 weeks of full-time employment.
Spatial Services GmbH handled many projects for Médecins Sans Frontières (MSF) and other humanitarian organizations. Some of these projects involved extraction of dwellings in Refugee & Internally Displaced Persons (IDP) camps and updating of some of the previous projects due to abrupt changes in the size of the refugee or IDP camps. I was given the opportunity to participate in several tasks regarding of these projects.
Image Tasking and Dwelling Extraction in refugee campsWhen a service request is activated, it's critical to provide an up-to-date database of image availability. The creation of the dwelling maps necessitated the use of very high-resolution satellite imagery (Airbus, Pleaides, SPOT, etc.) and, depending on the map's purpose, satellite mission tasking. This involved a periodic data search, gaining access to data from various sources, and selecting the most appropriate images based on the area of interest. In Airbus, we were guided through the image tasking procedure for a specific area, followed by perusing the product. The step of image tasking was then followed by performing basic image pre-processing steps.
Once the data was ready, using a deep learning algorithm built on ArcGIS Pro, semi-automatic dwelling extraction was used to extract the dwellings from the imagery, though fine tuning and manual editing were done in ArcGIS Pro to fully achieve the result.
Following the task of dwellings extractions, ArcGIS pro was used to generate dwelling density maps for MSF tasks involving refugee camps. These maps can help decision-makers make more informed decisions based on the requirements uncovered by the analysis.
Implementing Machine Learning for Dwelling Extraction in eCognitionDespite the fact that the deep learning model was now implemented in the initial dwelling extract, I was also guided through the working procedures and dwelling extraction using machine learning algorithms based on ecognition. We used this library to perform an experiment in ecognition developer for dwelling extraction. I was able to gain insight into the rules set used in this process while working on this task.
Working with Deep Learning Model for Dwelling Extraction in python environment of ArcGIS pro.Mask R-CNN Deep learning approach was used in the extraction of dwellings. The process involved general dwelling extraction without any classification. Here, we were provided with high resolution commercial data, and training samples was generated from few manually extracted dwelling. These samples were exported to a format compatible with a deep learning model for training. To get familiar with the task and methods involved this task was done with both ArcGIS pro toolset for Deep learning and using ArcGIS pro notebook by setting up the environment as per required for the task.
Finally, a simple tool was created in ArcGIS pro using model builder for filtering out the deep learning results. This tool was used in dissolving the multipart to single part to avoid the duplicity of the dataset.
Spatial Services have a good work atmosphere and encouraging team members. During my stay at the workplace, I was introduced to office activities and was able to complete the tasks that were allocated to me while doing my best to meet the expectations. I was provided with the opportunity to improve my knowledge in Machine Learning for dwelling extraction that not just helped me to understand the terminologies involved in the process but also helped in gaining ability to solve the error that appeared during the process through research.
In addition to the office work, these tasks guided in performing research on dwelling extraction using deep learning algorithm built in eCognition environment. As a result, my teammate Tanya Singh and I came up with the idea of developing a GUI in ecognition using deep learning model in dwelling extraction as our integrated project, based on our knowledge of dwelling extraction through ecognition using machine learning and through ArcGIS pro using deep learning.
With the completion of my internship at Spatial Services, I am excited and confident in my ability to work in the field of Geo-Humanitarian utilizing my knowledge on spatial science. I was able to discover and learn more on spatial use for humanitarian. As I worked as a part-time intern while also attending classes, I developed a sense of time management by distributing my time in such a way that I could properly complete tasks assigned to me both in the office and in class. Also, I am delighted to congratulate Spatial Service on winning the "2021 EARSC EO Partnership Award for humanitarian downstream service." I'm happy to be a small part of this group.