JOIV : International Journal on Informatics Visualization
Not a member yet
786 research outputs found
Sort by
Multilayer Perceptron Model with Feature Extraction for Potassium Deficiency Identification of Cocoa Plants
Get PDFThe development of Multilayer Perceptron (MLP) models for networked learning systems heavily relies on the specific application case study and the accurate parameterization aligned with the chosen computer vision feature extraction models. This study proposes an MLP model for identifying potassium deficiency in cocoa plants. The feature extraction methodology employs object feature extraction that commonly used in computer vision, including Local Binary Pattern (LBP), Gray Level Co-Occurrence Matrix (GLCM), and Hue Saturation Value (HSV) models. These computer vision techniques aid in analyzing leaf characteristics classified into two categories: normal conditions and leaves identified with potassium deficiency. The dataset used in this research comprises two conditions: with a white background and without any specific background. The study evaluates various feature extraction techniques based on MLP parameters, incorporating network learning rates and optimizing solvers. Employing the ROC analysis method throughout the data collection, algorithm development, validation, and analysis phases reveals that the most effective classification performance, reaching up to 93.33% accuracy on the background dataset and 90.00% on the non-background dataset, is achieved using HSV-based color feature extraction with MLP parameters set at an initial learning rate of 10-3 and employing the Adam optimization solver. These outcomes underscore the suitability of HSV color feature extraction for identifying potassium deficiency in cocoa plant leaves. However, optimizing parameters remains crucial to maximize its application in real-time identification systems. Future research should refine these parameters to enhance the model's robustness and efficacy across broader agricultural contexts
Overview of Software Re-Engineering Concepts, Models and Approaches
Get PDF Legacy systems face issues such as integrating new technology, fulfilling new requirements in the ever-changing environment, and meeting new user expectations. Due to the old complex system structure and technology, modification is hardly applied. Therefore, re-engineering is needed to change the system to meet new requirements and adapt to new technology. Software re-engineering generally refers to creating a new system from the existing one. Software re-engineering is divided into three (3) main phases: reverse engineering alteration and forward engineering. Reverse engineering examines, analyzes, and understands the legacy system in deriving the abstract representation of a legacy system; then, through necessary alterations such as restructuring, recording, and a series of forward engineering processes, a new system is built. This paper introduces the concepts of software re-engineering, including the challenges, benefits, and motivation for re-engineering. In addition, beginning with the traditional model of software re-engineering, this paper provides an overview of other models that provide different processes of software re-engineering. Each model has its unique set of processes for performing software re-engineering. Furthermore, re-engineering approaches show various ways of performing software re-engineering. Software re-engineering is a complex process that requires knowledge, tools, and techniques from different areas such as software design, programming, testing, et cetera. Therefore, monitoring the re-engineering process to meet the expectations is necessary
Application of Digital Teaching Materials Based on Flipped Learning Model in Civics Education in Elementary School
Get PDFThis research aims to improve students' understanding of Pancasila Student Profile Values through the implementation of Flipped Learning by combining it with digital teaching materials according to the characteristics of students in elementary schools. In addition, this research also aims to create practical digital teaching materials for elementary school students in Padang City on Civic Education learning. This research is a development research using the 4D development model (Define, Design, Develop and Disseminate). This study involved a sample of elementary school students in Padang City who measured the practicality of the developed teaching materials assessed through a structured evaluation process. The results showed a high practicality score of 96%, which categorized the digital teaching materials as very practical for use in the classroom. In addition, researchers also measured the impact of the implementation of these teaching materials on student learning outcomes by obtaining significant results; 87% of students achieved scores above the threshold of completeness, with an average score of 88. The findings suggest that the integration of Flipped Learning with digital teaching materials not only facilitates a deeper understanding of Pancasila values but also positively affects students' overall performance. The implications of this study highlight the potential for further research to explore the long-term effects of digital teaching materials and Flipped Learning on different subjects and levels of education. Future research could also investigate the scalability of these materials in different educational contexts and their effectiveness in fostering critical thinking and civic engagement among students
Optimizing Quadrotor Stability: RBF Neural Network Control with Performance Bound for Center of Gravity Uncertainty
Get PDFThe Radial Basis Function (RBF) neural network has been widely applied for approximating nonlinear systems and improving control robustness, particularly in uncertain conditions such as dynamic shifts in the quadrotor’s Center of Gravity (COG). However, initial weight estimation errors can degrade transient responses, reducing tracking performance. This study proposes a novel RBF-based control scheme integrated with a performance-bound mechanism to enhance quadrotor stability under COG uncertainty. The performance bound ensures that the quadrotor’s motion remains within a defined region around the reference trajectory, thereby minimizing steady-state and transient errors. The RBF network is trained online to estimate the system’s dynamic changes, and the controller is designed using a Lyapunov-like function to ensure stability. Simulation results show that the proposed controller achieves better tracking accuracy and significantly lower energy usage, with total force and moment values reduced compared to the standard RBF controller. Specifically, the proposed controller uses 3010.7 N of force and 2.2427 Nm of moment, while the standard controller requires 3150.2 N and 15.197 Nm. These results confirm that the proposed method provides improved performance and energy efficiency. This research highlights the potential of integrating performance bounds in neural network control for robust quadrotor navigation. Future work includes real-world experiments to validate performance under varying COG perturbations
Development Extraction of Regional Features of Pleural Cavity Objects in Pneumothorax Lung X-ray Images by Dilation and Erosion Morphology
Get PDFImage processing is a solution in the development of chest X-ray technology, starting from the image segmentation process as a preprocessing stage to separate the image object from the original background. Spontaneous pneumothorax (SP) is a type of air collection in the pleural cavity that develops without trauma. The diagnosis of pneumothorax has a sensitivity of approximately 25 to 75% using an anteroposterior chest x-ray, which still provides a dubious picture of pneumothorax. However, the development of the Region Feature algorithm with a new algorithm, namely RM Multy, has improved the accuracy. The RM Multy algorithm can calculate the area of the object, allowing it to produce the area of infiltration in the right lung, left lung, and the lung as a whole. The Region Feature results of the Pneumothorax obtained with the detected image area as many as 19 areas, for the pixel size of each area are 145, 355, 110, 31, 31, 52, 30, 36, 54, 122, 58, 23, 476, 77, 192, 24, 168, 263, 41 and 44. So the total pixels for 19 areas is 2301. The area converted to mm2 is 2301 x 0.04 mm2 = 92.04 mm2. Classification results on lungs with Pneumothorax and Normal by detection process with RM Multy using the CNN algorithm with an accuracy of 96.43%. This accuracy confirms the success of the system, which has been processed using a new algorithm. Therefore, further development is needed to improve detection accuracy in pneumothorax cases with smaller area sizes
Advanced Instance Segmentation of Aeroponics Tissue Culture-Based Seeds Potatoes Based on Improved YOLOv8l-small
Get PDFTo improve agricultural production, this study develops an advanced instance segmentation system for aeroponic tissue culture-based potato seedlings. We present an IoT system that integrates multiple sensors for humidity, temperature, pH, and turbidity to enable real-time monitoring. Additionally, we adapt the YOLOv8l-small computer vision model, an optimized version of YOLOv8, designed explicitly for efficient potato leaf disease detection and segmentation, even in resource-constrained IoT environments. YOLOv8 is a significant advancement in the YOLO series, for instance, segmentation, combining better accuracy, efficiency, and flexibility. YOLOv8 outperforms previous methods in generating precise segmentation masks while maintaining real-time performance. These innovations make YOLOv8 a robust choice for a variety of computer vision tasks, including instance segmentation, in both research and practical applications. When tested on a custom dataset of potato leaf pictures, the suggested model produced mask mAP50 of 0.842 and mAP50-95 of 0.566, with a model size of 36.1 MB and an inference duration of 9.3 ms. These outcomes are similar to those of the original YOLOv8l model, which had a slower inference time of 11.0 ms and a much larger model size of 92.3 MB, albeit at the expense of a somewhat higher mAP50 of 0.843. The study concludes that the proposed model provides similar accuracy with greater computational efficiency, making it ideal for IoT-based agricultural systems. Future research will explore additional aspects, while practical experiments aim to reduce labor costs
Development of a Predictive Model for Citrus Shipments and Prices, and Analysis of Influencing Factors
Get PDFGiven the significance of the citrus industry, which accounts for more than half of Jeju Island's agricultural revenue (KRW 950.8 billion, 55.92% of farming income), this study aims to develop prediction models for open-field and greenhouse-grown citrus shipment volumes and prices. While previous research has explored crop production forecasting, there is a notable absence of comprehensive studies integrating deep learning approaches with environmental factors for Jeju citrus prediction, particularly in addressing the complex interplay between weather patterns and market dynamics. To bridge this gap, this study analyzed various domestic and international factors, including weather information, public holidays, and imported fruit data, which were utilized as independent variables in the model design. Deep learning-based models, specifically LSTM for capturing long-term dependencies, Seq2Seq for handling variable-length sequences, and Attention mechanisms for focusing on relevant temporal patterns, were employed to perform the predictions. Their accuracy and stability were thoroughly evaluated against traditional machine learning benchmarks. The findings revealed that citrus shipment volumes and prices are significantly influenced by temporal factors (average temperature, shipment timing) and market dynamics (transaction volume, competing fruit prices), with the Seq2Seq model achieving the highest prediction accuracy. Furthermore, by adjusting the window sizes in various time series models, we were able to simulate different scenarios, providing stakeholders with a robust tool for market planning and decision-making. The findings of this research are expected to contribute to the efficient operation of the citrus market and the maximization of benefits for related stakeholders
Web-Based Deep Learning Approach to Identifying AI-Generated Anime Illustration
Get PDFAs technology advances rapidly in artificial intelligence, the dominance of generative artificial intelligence (AI) images becomes increasingly evident in art, design, and the creative industry. However, the generative AI has processed numerous images from the Internet, including copyrighted content, trademarks, and artists' illustrations, which pose legal risks. Consequently, the manual tasks involved in managing and classifying these images have become more complex and time-consuming. Therefore, this research proposes the application of deep learning techniques, specifically Convolutional Neural Network (CNN), to automate the process of classifying AI-generated illustrations. The research was conducted by the Cross-Industry Standard Process for Data Mining (CRISP-DM) method. Initially, the study began with a literature review to describe the state-of-the-art in image detection. Then, a dataset of illustrations was collected from the Pixiv website using web scraping techniques. After data cleaning, separation, and augmentation, three pre-trained models were created and compared on 1200 training data and evaluated against 400 testing and 400 validation data. From the evaluation, the model using MobileNet V3 Large architecture achieved an impressive 94% accuracy, outperforming MobileNet V2 and Inception V3 architectures, respectively by 3% and 5%. Thus, the implementation of CNN holds the promise of providing an efficient solution for identifying and classifying various types of AI anime illustrations, benefiting consumers and artists practically. Future research could consider incorporating additional data categories and variations to further enhance the model's ability to distinguish between AI-generated and human-made illustrations
Intelligent Monitoring System Framework for Peatland Management in IoT-Integrated Precision Agriculture
Get PDFPeatlands have excellent air retention capabilities and are crucial for environmental health. They act as natural sponges, absorbing and releasing air, which helps maintain soil moisture levels vital for crops. However, peatlands are highly sensitive ecosystems often threatened by unsustainable agricultural practices. When managed sustainably, peatlands scattered across the globe can be utilized for various farming activities. Managing peatlands for food crops presents an alternative to agriculture in peatland areas, enhancing economic growth in rural regions. This research aims to introduce a framework that integrates IoT into the intelligent monitoring of peatland management for precision agriculture. The primary challenge is implementing effective monitoring and management strategies for sensitive peatlands within precision agriculture. The main principle of precision agriculture is data-driven decision-making, supported by modern agricultural management that employs technology and data analysis to optimize farming practices. The proposed system framework can be utilized to identify the best types of food crops for making new decisions while ensuring high yields at the agricultural level. Precision agriculture principles are then applied to enhance the accuracy of monitoring peatland management, focusing on suitable land potential and food crops planted in areas with the highest potential. The results indicate that prioritizing peatlands for food crops reduces inappropriate decisions in selecting food crops. Furthermore, the efficiency of agricultural management can be improved with lower management costs. This framework provides a practical and user-friendly basis for informing all stakeholders on automating Peatland agriculture for food crops using precision agriculture systems integrated with IoT. Management practices that apply information technology aim to optimize crop inputs based on temporal and spatial variability. The cost-effectiveness from this perspective creates transition opportunities for communities, positioning our framework as a solution for designing Peatland management with intelligent monitoring
Detection of Keratitis in the Cornea by Developing an Active Contour Method Based on Contrast Features
Get PDFDigital Image Processing (DIP) is a scientific discipline that uses computer image processing techniques. The object of this research is keratitis on the cornea. The image of keratitis is obtained using a slit lamp at Padang Aye Center (PAC) Hospital, based on the results of the diagnosis, namely by looking at the development of the infiltrate or also called hypopyon, measuring the ulcer borders horizontally and vertically to evaluate improvement or response to the treatment given. The clinical results cannot determine the extent and circumference of the keratitis layer area that responds to treatment in the corneal area. The images used were 206 slit lamp images of keratitis. This research provides knowledge in the form of contrast values in the Active Contour method, resulting in an update called Active Contour Contrast Adjustment (ACCA) in correctly segmenting keratitis objects and providing measurements of the area and perimeter of the keratitis area. Overall. The research results from 206 slit lamp images, 195 slit lamp images of keratitis could detect keratitis correctly, and eleven slit lamp images of keratitis could not be detected, resulting in an accuracy of 94.66%. Meanwhile, the standard Active Contour accuracy was not detected at all or 100% undetected. Based on 11 images not detected using the (ACCA) method from 206 images, an accuracy of 5.33% was obtained. So, the results obtained are outstanding and can be used as a reference for medical personnel