MALICIOUS UAVS CLASSIFICATION USING VARIOUS CNN ARCHITECTURES FEATURES AND MACHINE LEARNING ALGORITHMS

International Journal of 3D Printing Technologies and Digital Industry, Journal Year: 2023, Volume and Issue: 7(2), P. 277 - 285

Published: Aug. 26, 2023

Aircraft are used in many fields such as engineering, logistics, transportation and disaster management. With the development of drones, aerial vehicles have become more widely for entertainment purposes. However, addition to its useful applications, malicious use is also becoming widespread. It has a necessity eliminate this problem, especially since it poses significant danger other aircraft. In order identify aircraft solve problem quickly, study, five different were classified based on images. five-class dataset containing aeroplane, bird, drone, helicopter UAV (Unnamed Aerial Vehicle) images was used. Three CNN (Convolutional Neural Network) models employed extract features. Image features extracted with SqueezeNet, VGG16, VGG19 Artificial Network (ANN), Support Vector Machine (SVM) Logistic Regression (LR) machine learning methods. As result experiments, most accuracyful result, 92%, obtained from classification SqueezeNet model ANN. The proposed study will be integrated into various systems field aviation detect UAVs take necessary precautions.

Language: Английский

---

The Effectiveness of Deep Learning Methods on Groundnut Disease Detection

Proceedings of the International Conference on Advanced Technologies, Journal Year: 2023, Volume and Issue: unknown

Published: Aug. 19, 2023

Early detection of plant diseases in the agricultural sector is considered an important goal to increase productivity and minimize damage. This study deals with use deep learning methods realize automatic leaf peanut plants explicability model heatmap visualizations formed during diseases. In study, a dataset containing 3058 images 5 classes enriched diseased healthy samples leaves was used. The explainability property has also been studied understand why models detect particular disease. decision processes models, which are usually described as "magic box", were visualized method this study. By highlighting pixels that effective detecting visualization, decision-making process tried be made understandable. results show have high performance diseases, obtained by visualization reliable tool for specialists producers. Thanks visual explanations provided model, level confidence increased provided. constitutes step towards increasing efficiency applications providing more efficient approach disease management investigating impact field plants.

Language: Английский

---

Two-stage HOG/SVM for license plate detection and recognition

Indonesian Journal of Electrical Engineering and Computer Science, Journal Year: 2024, Volume and Issue: 34(1), P. 210 - 210

Published: Feb. 29, 2024

Automatic license plate recognition (ALPR) is one of the technologies used in intelligent transport systems (ITS) to read vehicle plates automatically. The extracted information has various potential applications, including but not limited an electronic payment gateway, a system for paying parking fees, road surveillance, and managing traffic flow. In this paper, we propose efficient method detect identify Algerian (LP). This consists two-stage algorithm that combines histogram oriented gradients (HOG) with support vector machine (SVM) classifier. purpose first stage HOG/SVM detection LP, while digits accomplished by second HOG/SVM. As contribution, dataset standard LP available elsewhere built (DZLP dataset), proposal very pre-processing step digit recognition. Experimental results show proposed approach yields high average rates, which 97.5% 99.46%, respectively.

Language: Английский

---

APPLICATION OF FUZZY METRICS IN CLUSTERING PROBLEMS OF AGRICULTURAL CROP VARIETIES

Ekonomika poljoprivrede, Journal Year: 2024, Volume and Issue: 71(1), P. 121 - 134

Published: March 31, 2024

The problem of image-based detection the variety beans, using artificial intelligence, is currently dealt with by scientists various profiles. idea this paper to show possibility applying different types distances, primarily those that are fuzzy metrics, in clustering models order improve existing and obtain more accurate results. presents method variable neighborhood search, which uses both standard t-metrics dual s-metrics characterized appropriate parameters. By varying parameters metric as well metaheuristic used, we have shown how it possible obtained results were compared ones from literature. criterion function used a metric, proven paper.

Language: Английский

---

A comparative study of optimization algorithms for feature selection on ML-based classification of agricultural data

Cluster Computing, Journal Year: 2023, Volume and Issue: 27(3), P. 3341 - 3362

Published: Oct. 3, 2023

Language: Английский

---