Anzeige
Erschienen in:
2019 | OriginalPaper | Buchkapitel
Deep CNN Based System for Detection and Evaluation of RoIs in Flooded Areas
verfasst von : Dan Popescu, Loretta Ichim, George Cioroiu
Erschienen in: Neural Information Processing
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
Abstract
The flood detection and damage evaluation in precision agriculture are of great interest nowadays. The paper proposes the simultaneously detection and evaluation of small regions affected by non-severe flooding and, also, small regions with vegetation in county areas. The images are taken by a UAV team in a photogrammetry mission. As novelty, the paper proposes and compares four cheap, real time, and accurate methods based on convolutional neural networks (Full LeNet, Half LeNet, Pixel YOLO, and Decision YOLO) to segment these regions of interest. These methods are compared with classical methods of region segmentation based on extracting and comparing features from images. The real masks are manually created by operators using information of the color components R, G, B, H. A set of 2000 images were used for the learning phase and another set of 1400 image were used for method validation. The method presents the advantages of accuracy and time processing (especially in the testing phase).