Recent Advances in Computer Science and Communications - Volume 18, Issue 4, 2025

Floods are among the deadliest natural calamities, devastating ecosystems and human lives worldwide. In India, Bihar is a state grappling with economic hardships and faces severe agricultural devastation due to recurring floods, destroying crops and natural resources, which significantly impacts local farmers. This research addresses the critical need to deeply understand the flood dynamics of selected study areas.

This research presents a case study that focuses on leveraging Remote Sensing tools and Machine Learning techniques for comprehensive flood mapping and damage analysis in Gopalganj District, Bihar, India, using remote sensing data. More specifically, this research presents three major objectives: (i) Flood damage mapping and change analysis before and after the flood using the Sentinel-2 satellite dataset, (ii) Evaluation of the impact of integrating spectral indices on the accuracy of classification, (iii) Identification of most robust predictor spectral indices for the classification.

The Sentinel-2 satellite dataset encompasses 13 bands with resolutions of 10m, 20m, and 60m. Here, four spectral bands (NIR, Red, Green, and Blue) with the finest resolution of 10m have been selected for this study. These bands are integrated with four spectral indices, namely Normalized Difference Water Index (NDWI), MNDWI (Modified NDWI), Normalized Difference Vegetation Index (NDVI), and Soil Adjusted Vegetation Index (SAVI). Two ML classifiers, namely Support Vector Machine (SVM) and Random Forest (RF) have been employed for pixel-based supervised classification.

Results have shown that RF outperformed and worked well in extracting water bodies and flood-damaged areas effectively. The results demonstrated that RF obtained (Overall Accuracy (OA)= 89.54% and kappa value (ka) = 0.872) and SVM reported (OA= 87.69%, ka= 0.849) for pre-crisis data, whereas, for post-crisis, RF reported (OA=91.54%, ka = 0.897), SVM reported (OA= 89.77%, ka= 0.875).

It was reported that the integration of spectral indices improved the OA by +3.41% and +2.86% using RF and SVM, respectively. The results of this study demonstrated that the waterbody area increased from 12.72 to 88.23 km^2, as shown by the RF classifier. The variable importance computation results indicated that MNDWI is the most important predictor variable, followed by NDWI. This study recommends the use of these two predictor variables for flood mapping.

Slicing technology focuses on data association behavior and is suitable for variable dependency analysis in tight defect contexts.

By constructing a control flow variable dependency graph(CFVDG), performing slicing and semantic analysis and understanding, finally applied to detect defection.

Compared with the slicing methods based on SDG and DFG, the consumption of time and space has decreased.

The experimental results show that this method can improve the precision of defect detection and reduce the false alarm rate, while reasonably and effectively shortening the time for graph construction and slicing, and reducing the consumption of edge storage space.

In order to realize seamless indoor and outdoor positioning, the positioning results of multiple positioning methods are taken into consideration, and a seamless indoor and outdoor positioning method that ignores the differences in indoor and outdoor environments is required now.

The implementation of Pseudo satellite/GNSS combined indoor and outdoor fusion positioning for seamless indoor and outdoor environment positioning.

An adaptive federated filter is needed for this environment, which can dynamically adjust the information allocation parameters and measurement noise of the sub-filters in the federated filters based on positioning data. It adopts multi-sensor fusion filter to design a seamless indoor and outdoor positioning method. Different positioning data is fused through federated filtering, ultimately seamless indoor and outdoor positioning is realized.

This algorithm achieves a fixed ambiguity pseudo satellite/GNSS accuracy of better than 0.15 meters in low-density buildings where there are more than 7 GNSS satellites. When there are fewer than 4 GNSS satellites and the positions is severely obstructed, GNSS alone cannot realize the position, but with the support of pseudo satellites, the accuracy of position can be better than 0.3 m. Even without GNSS and only 4 pseudo satellites, the accuracy of position can still be better than 0.5 m.

The relevant experimental results indicate that the method proposed can be used for practical applications of indoor and outdoor fusion positioning.

With the development of intelligent ship technology, computer vision technology has been widely utilized in the field of maritime monitoring. This is of great significance in ensuring the safety of navigation and improving the efficiency of shipping. However, complex and changing sea conditions and arbitrary traveling ships pose more accurate and faster requirements for the target detection algorithm used in the intelligent ship systems.

The primary objective of this paper is to propose an optimized version of the ship lightweight target detection algorithm based on YOLOv5s architecture. This enhancement involves the innovative fusion of the Shufflenetv2 network and the NAM attention mechanism, collectively termed as SN-YOLOv5s. This integration seeks to elevate the algorithm’s performance in detecting ship targets, offering improved accuracy and efficiency.

Firstly, the Shufflenetv2 network and NAM attention mechanism are used to replace the backbone network, significantly reducing the number of network parameters and improves the model detection accuracy. Secondly, in the process of converting the feature map to a fixed-size feature vector, SimSPPF is used to replace the fast pyramid pooling SPPF module, ensuring the efficiency and minimizing information loss. Lastly, EIOU is utilized to replace the bounding box regression loss function CIOU to make the model converge faster and with higher accuracy.

Test results on the SeaShips dataset show that compared to the original YOLOv5s network, the average accuracy of target detection using the SN-YOLOv5s network is improved by 4.7%, the amount of computation is reduced by 40%, the amount of parameters is reduced by 20.6%, and the volume of model weights is decreased by 15.4%.

The experimental results fully demonstrate that the algorithm can significantly reduce the running cost of the model and improve the detection accuracy of the model, thus effectively guaranteeing the efficiency and quality of ship target detection.

Due to its ferocity, enormous metastatic potential, and variability, cancer is responsible for a disproportionately high number of deaths. Cancers of the lung and colon are two of the most common forms of the disease in both sexes worldwide. The excellence of treatment and the endurance rate for cancer patients can be greatly improved with early and precise diagnosis.

We suggest a computationally efficient and highly accurate strategy for the rapid and precise diagnosis of lung and colon cancers as a substitute for the standard approaches now in use. The training and validation procedures in this work made use of an enormous dataset consisting of lung and colon histopathology pictures. There are 25,000 Histopathological Images (HIs) in the dataset, split evenly among 5 categories (mostly lung and colon tissues). Before training it on the dataset, a pretrained neural network (AlexNet) had its four layers fine-tuned.

The study enhances malignancy detection in lung and colon histopathology images by applying transfer learning with class-selective image processing. Instead of enhancing the entire dataset, a targeted contrast enrichment was applied to images from the underperforming class, improving the model's accuracy from 92.3% to 99.2% while reducing computational overhead.

This approach stands out by emphasizing class-specific enhancements, leading to significant performance gains. The results meet or exceed established CAD metrics for breast cancer histological images, demonstrating the method's efficiency and effectiveness.