Implementación de un sistema para el reconocimiento y clasificación de la contaminación superficial de la laguna de Colta mediante algoritmos de inteligencia artificial
The Laguna de Colta, being a natural attraction and one of the places with a large influx of tourists, takes on an important characteristic as regards the visual aspect. However, the Lagoon can be compromised by surface contamination (polluting solid waste), which in addition to affecting not only t...
সংরক্ষণ করুন:
| প্রধান লেখক: | |
|---|---|
| বিন্যাস: | bachelorThesis |
| ভাষা: | spa |
| প্রকাশিত: |
2020
|
| বিষয়গুলি: | |
| অনলাইন ব্যবহার করুন: | http://dspace.unach.edu.ec/handle/51000/7210 |
| ট্যাগগুলো: |
ট্যাগ যুক্ত করুন
কোনো ট্যাগ নেই, প্রথমজন হিসাবে ট্যাগ করুন!
|
ট্যাগ যুক্ত করুন | সংক্ষিপ্ত: | The Laguna de Colta, being a natural attraction and one of the places with a large influx of tourists, takes on an important characteristic as regards the visual aspect. However, the Lagoon can be compromised by surface contamination (polluting solid waste), which in addition to affecting not only the Lagoon in question, but also the entire place near the shores of the Lagoon, for this reason it has been proposed to carry out this research project. In this project the development of an algorithm in the field of artificial intelligence of neural networks is exposed, where Deep Learning is applied for the recognition and classification of polluting objects present on the surface of the Laguna de Colta, from images video obtained through an FPV camera that is mobilized by an unmanned aerial vehicle (UAV). Specifically, the system proposes to follow a robust method to detect contaminating objects (plastic bottles, metal containers and reeds) based on CNN (Convolutional Neural Network) and compiling a database of specific images and with Transfer Learning techniques, it will work in near real-time performance and satisfactory accuracy. To overcome the challenges of building an accurate contaminating object detection model, the core architecture of the pre-trained AlexNet network was transformed to a faster region-based convolutional neural network (Faster R-CNN) in the MATLAB environment. Consequently, the system will allow the automatic management of the concentration of surface contamination and provide acceptable results in terms of precision and execution time. Finally, two Graphical Interfaces (Matlab, Unity) have been created where the results and the position of the detected objects thrown by the system were visualized in real time. |
|---|