Building an oil spill segmentation model is very challenging because of the limited available information on oil spill accidents. Therefore, this paper proposes a custom data generator based on Segmentation Network (Seg-Net) model implemented in Conditional Generative Adversarial Network (CGAN). The proposed model is trained for oil spill segmentation using 50 Sentinal-1 Synthetic Aperture Radar (SAR) images. The proposed model employs a modified Seg-Net as a generator to produce high-quality oil spills’ images and a Patch-GAN as discriminator. This architecture results in a significant improvement of the final oil segmentation results, in comparison with Seg-Net model, while using relatively small training dataset. For performance assessment, the paper presents the oil spills segmentation results of four suggested models using Sentinel-1 SAR images. The presented models are U-Net, Seg-Net, CGAN, and a Seg-Net-based CGAN the performance assessment reveals that the proposed model produces oil spill segmentation images with an average accuracy of 99.04%, Intersection over Union (IoU) index of 96.59%, and a precision of 85.24%. In addition, the training time required for the proposed model is 3 h 20 min per 50 epochs, while it is nearly 10 h and 55 min for training a CGAN model.
Measurements of Global Positioning Satellite System receivers are affected by systematic offsets related to group and phase delays of the signal generation and processing chain. One of the important factors affecting the ionosphere Total Electron Content estimation accuracy is the hardware differential code biases inherited in both Global Positioning System satellites and receivers. The resulting code and phase biases depend on the transmission frequency and the employed signal modulation. An efficient algorithm using the geometry conditions between satellite and tracking receivers is proposed to determine the receiver differential code biases using Egyptian permanent reference stations. This method does not require a traditional single-layer ionosphere model and can be used for estimating differential code biases of receivers in a regional network. This paper estimates receiver differential code biases for nine receivers located within Egyptian network. The results showed that the estimated mean value of the receiver differential code biases varied from −28 ns (nanosecond) to 39 ns. It is clear from the results that differential code biases values for Egyptian sites do not vary much with latitude and longitude, except at Aswan and Abu Simpel. Differential code biases values increase gradually with increasing height.
29 January 2025 in Reports on Geodesy and Geoinformatics
Assessing the reliability of using open-source software packages for post-processing of the Global Navigation Satellite System (GNSS) is essential since GNSS modernization has the potential to assist satellite navigation users. The purpose of this study is to evaluate the accuracy of using two MATLAB-based programs, raPPPid and PPPH for precise point positioning in Europe. Data from 12 MGEX stations over two days were used, with one day dedicated to each of the 6 stations and a 30-second observation interval. The data were post processed by PPPH and raPPPid programs and a comparison was made to evaluate the results accuracy produced by each software and the ones acquired from MGEX stations. Convergence time was also estimated. By comparing the root mean square error (RMSE) values for North, East and Up directions estimated by PPPH and raPPPid, it was found that raPPPid gives more accurate results where the RMSE in N direction estimated by raPPPid varied from 0.5 cm to 1.9 cm; however, RMSE in N direction estimated by PPPH varied from 0.7 cm to 2.8 cm. RMSE in E direction estimated by raPPPid varied from 0.4 cm to 3.3 cm, but RMSE in E direction estimated by PPPH varied from 0.5 cm to 3.7 cm. RMSE in Up direction estimated by raPPPid varied from 0.8 cm to 5.2 cm, while RMSE in Up direction estimated by PPPH varied from 0.9 cm to 5.5 cm. 3D Positioning error was also estimated by both software and it was found that the 3D positioning error estimated by raPPPid varied from 0.2 cm to 2.2 cm, whereas the 3D positioning error estimated by PPPH varied from 0.9 cm to 4.1 cm. Finally, the average convergence time achieved by raPPPid was 16.5 minutes, while the average convergence time achieved by PPPH was 32 minutes.