Pub Date : 2023-05-18DOI: 10.1109/eIT57321.2023.10187314
Anton Skurdal, Youness Arjoune, Niroop Sugunaraj, Shree Ram Abayankar Balaji, Sreejith V. Nair, Prakash Ranganathan, Burton Johnson
Soybean Cyst Nematode (SCN) is a serious pathogen in soybean production and contributes to annual economic losses of more than $1.5 billion (1996–2016) in the U.S. SCN is a microscopic thread-like nematode that burrows into the roots of soybean plants and typically cannot be identified above ground. The paper investigates multitude of variables such as NDVI from multi-spectral images, egg counts, and micro-nutrient composition (e.g., pH, nitrogen, phosphorus, potassium) across two SCN-prone field plots in Casselton/Prosper, North Dakota. The preliminary results indicate that NDVI is a good metric to track for SCN density population during planting, growing, and harvesting periods along with other historical ground truth data. Also, a contour plot using Empirical Bayesian Kriging (EBK) was designed by integrating NDVI and egg count data for co-tracking distribution changes. Such access to ground truth data (i.e., aerial and soil properties) will enable the development and training of robust machine learning models for predicting SCN hotspots.
{"title":"Correlation of Egg counts, Micro-nutrients, and NDVI Distribution for Accurate Tracking of SCN Population Density Detection","authors":"Anton Skurdal, Youness Arjoune, Niroop Sugunaraj, Shree Ram Abayankar Balaji, Sreejith V. Nair, Prakash Ranganathan, Burton Johnson","doi":"10.1109/eIT57321.2023.10187314","DOIUrl":"https://doi.org/10.1109/eIT57321.2023.10187314","url":null,"abstract":"Soybean Cyst Nematode (SCN) is a serious pathogen in soybean production and contributes to annual economic losses of more than $1.5 billion (1996–2016) in the U.S. SCN is a microscopic thread-like nematode that burrows into the roots of soybean plants and typically cannot be identified above ground. The paper investigates multitude of variables such as NDVI from multi-spectral images, egg counts, and micro-nutrient composition (e.g., pH, nitrogen, phosphorus, potassium) across two SCN-prone field plots in Casselton/Prosper, North Dakota. The preliminary results indicate that NDVI is a good metric to track for SCN density population during planting, growing, and harvesting periods along with other historical ground truth data. Also, a contour plot using Empirical Bayesian Kriging (EBK) was designed by integrating NDVI and egg count data for co-tracking distribution changes. Such access to ground truth data (i.e., aerial and soil properties) will enable the development and training of robust machine learning models for predicting SCN hotspots.","PeriodicalId":113717,"journal":{"name":"2023 IEEE International Conference on Electro Information Technology (eIT)","volume":"12 3","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114025187","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-18DOI: 10.1109/eIT57321.2023.10187274
T. T. Khoei, Ghilas Aissou, K. Shamaileh, V. Devabhaktuni, N. Kaabouch
Unmanned Aerial Networks (UAVs) are prone to several cyber-atttacks, including Global Positioining Spoofing attacks. For this purpose, numerous studies have been conducted to detect, classify, and mitigate these attacks, using Artificial Intelligence technqiues; howver, most of these studies provided techniques with low detection, high misdetection, and high bias rates. To fill this gap, in this paper, we propose three supervised deep learning techniques, namely Deep Neural Network, U Neural Network, and Long Short Term Memory. These models are evaluated in terms of Accuracy, Detection Rate, Misdetection Rate, False Alarm Rate, Training Time per Sample, Prediction Time, and Memory Size. The simulation results indicated that the U Neural Network outperforms other models with accuracy of 98.80%, a probability of detection of 98.85%, a misdetection of 1.15%, a false alarm of 1.8%, a training time per sample of 0.22 seconds, a prediction time of 0.2 seconds, and a memory size of 199.87 MiB. In addition, these results depicted that the Long Short Term Memory model provides the lowest performance among other models for detecting these attacks on UAVs.
{"title":"Supervised Deep Learning Models for Detecting GPS Spoofing Attacks on Unmanned Aerial Vehicles","authors":"T. T. Khoei, Ghilas Aissou, K. Shamaileh, V. Devabhaktuni, N. Kaabouch","doi":"10.1109/eIT57321.2023.10187274","DOIUrl":"https://doi.org/10.1109/eIT57321.2023.10187274","url":null,"abstract":"Unmanned Aerial Networks (UAVs) are prone to several cyber-atttacks, including Global Positioining Spoofing attacks. For this purpose, numerous studies have been conducted to detect, classify, and mitigate these attacks, using Artificial Intelligence technqiues; howver, most of these studies provided techniques with low detection, high misdetection, and high bias rates. To fill this gap, in this paper, we propose three supervised deep learning techniques, namely Deep Neural Network, U Neural Network, and Long Short Term Memory. These models are evaluated in terms of Accuracy, Detection Rate, Misdetection Rate, False Alarm Rate, Training Time per Sample, Prediction Time, and Memory Size. The simulation results indicated that the U Neural Network outperforms other models with accuracy of 98.80%, a probability of detection of 98.85%, a misdetection of 1.15%, a false alarm of 1.8%, a training time per sample of 0.22 seconds, a prediction time of 0.2 seconds, and a memory size of 199.87 MiB. In addition, these results depicted that the Long Short Term Memory model provides the lowest performance among other models for detecting these attacks on UAVs.","PeriodicalId":113717,"journal":{"name":"2023 IEEE International Conference on Electro Information Technology (eIT)","volume":"171 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114158218","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-18DOI: 10.1109/eIT57321.2023.10187308
Rajrup Mitra, Amrita Das, Jack Heichel, S. Dorafshan, N. Kaabouch
Steel is used in different infrastructural constructions. The durability and serviceability of steel made it more suitable than other construction materials. However, exposure to weather elements can cause defects in steel structures. Early detection and treatment of structural defects can prevent the structure from becoming more damaged and more expensive to repair. Corrosion resistance and fatigue strength in any steel structure can be influenced by defects such as patches, scratches, and coating erosion. Current methods to detect steel defects are based on manual visual inspection. Autonomous UAS imaging-based defect detection methods have shown promising results in terms of accuracy and time. This paper compares the performance of two deep learning models, InceptionResnetV2 and ResNet152V2, for detecting steel defects. These models were trained in transfer learning mode and tested on two different datasets, the Severstal dataset present on Kaggle and a dataset generated by the authors of this paper. The results show that ResNet152V2 outperforms InceptionResnetV2 with an average accuracy of 95% and a misdetection rate of 5%. Overall, both the models, ResNet152V2 and InceptionResNetV2, showed an improvement of 12.59% and 9.59%, respectively, compared to MobileNet used in a previous study, when all were trained and tested on the Severstal dataset.
{"title":"A Comparative Analysis of Two Deep Learning Neural Networks for Defect Detection in Steel Structures Using UAS","authors":"Rajrup Mitra, Amrita Das, Jack Heichel, S. Dorafshan, N. Kaabouch","doi":"10.1109/eIT57321.2023.10187308","DOIUrl":"https://doi.org/10.1109/eIT57321.2023.10187308","url":null,"abstract":"Steel is used in different infrastructural constructions. The durability and serviceability of steel made it more suitable than other construction materials. However, exposure to weather elements can cause defects in steel structures. Early detection and treatment of structural defects can prevent the structure from becoming more damaged and more expensive to repair. Corrosion resistance and fatigue strength in any steel structure can be influenced by defects such as patches, scratches, and coating erosion. Current methods to detect steel defects are based on manual visual inspection. Autonomous UAS imaging-based defect detection methods have shown promising results in terms of accuracy and time. This paper compares the performance of two deep learning models, InceptionResnetV2 and ResNet152V2, for detecting steel defects. These models were trained in transfer learning mode and tested on two different datasets, the Severstal dataset present on Kaggle and a dataset generated by the authors of this paper. The results show that ResNet152V2 outperforms InceptionResnetV2 with an average accuracy of 95% and a misdetection rate of 5%. Overall, both the models, ResNet152V2 and InceptionResNetV2, showed an improvement of 12.59% and 9.59%, respectively, compared to MobileNet used in a previous study, when all were trained and tested on the Severstal dataset.","PeriodicalId":113717,"journal":{"name":"2023 IEEE International Conference on Electro Information Technology (eIT)","volume":"34 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117266169","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-18DOI: 10.1109/eIT57321.2023.10187344
Mohammad Nayfeh, Joshua Price, M. Alkhatib, K. Shamaileh, N. Kaabouch, Vijay K. Devabhakuni
In this paper, a three-class machine learning (ML) model is implemented on an unmanned aerial vehicle (UAV) with a Raspberry Pi processor for classifying two global positioning system (GPS) spoofing attacks (i.e., static, dynamic) in real-time. First, several models are developed and tested utilizing a dataset collected in a previous work. This dataset conveys GPS-specific features, including location information. Models evaluations are carried out using the detection rate, F-score, false alarm rate, and misdetection rate, which all showed an acceptable performance. Then, the optimum model is loaded to the processor and tested for real-time detection and classification. Location-dependent applications, such as fixed-route public transportations are expected to benefit from the methodology presented herein as the longitude, latitude, and altitude features are characterized in the developed model.
{"title":"A Real-time Machine Learning-based GPS Spoofing Solution for Location-dependent UAV Applications","authors":"Mohammad Nayfeh, Joshua Price, M. Alkhatib, K. Shamaileh, N. Kaabouch, Vijay K. Devabhakuni","doi":"10.1109/eIT57321.2023.10187344","DOIUrl":"https://doi.org/10.1109/eIT57321.2023.10187344","url":null,"abstract":"In this paper, a three-class machine learning (ML) model is implemented on an unmanned aerial vehicle (UAV) with a Raspberry Pi processor for classifying two global positioning system (GPS) spoofing attacks (i.e., static, dynamic) in real-time. First, several models are developed and tested utilizing a dataset collected in a previous work. This dataset conveys GPS-specific features, including location information. Models evaluations are carried out using the detection rate, F-score, false alarm rate, and misdetection rate, which all showed an acceptable performance. Then, the optimum model is loaded to the processor and tested for real-time detection and classification. Location-dependent applications, such as fixed-route public transportations are expected to benefit from the methodology presented herein as the longitude, latitude, and altitude features are characterized in the developed model.","PeriodicalId":113717,"journal":{"name":"2023 IEEE International Conference on Electro Information Technology (eIT)","volume":"44 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125788414","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-18DOI: 10.1109/eIT57321.2023.10187328
M. Rehman, I. Nizami, Ali Ahsan, K. Chong
A copy-move image forgery is the most common type of image tampering. It can be done by copying a part of an image and paste on another part of the same image. Therefore, it can be one of the challenging tasks to find that forgery. This paper suggested a different approach to detect the copy move image forgery by the natural scene statistic features. These features are extracted from both original and forged images of MICC-F2000 dataset. Natural scene statistics are the statistical properties of any natural image captured by any camera, so an attempt of forging an image makes these properties un-natural. By this method, an original and forged images can be easily classified by state-of-the-art machine learning models trained on these features. The performance of this method is quantitatively assessed using the famous evaluation metrics i-e accuracy, TPR, FPR, TNR, Recall and F1-score. A comparison with other state-of-the-art techniques has shown that the proposed technique has shown better results in comparison with the other techniques.
{"title":"Machine Learning Based Image Forgery Detection Using Natural Scene Statistics","authors":"M. Rehman, I. Nizami, Ali Ahsan, K. Chong","doi":"10.1109/eIT57321.2023.10187328","DOIUrl":"https://doi.org/10.1109/eIT57321.2023.10187328","url":null,"abstract":"A copy-move image forgery is the most common type of image tampering. It can be done by copying a part of an image and paste on another part of the same image. Therefore, it can be one of the challenging tasks to find that forgery. This paper suggested a different approach to detect the copy move image forgery by the natural scene statistic features. These features are extracted from both original and forged images of MICC-F2000 dataset. Natural scene statistics are the statistical properties of any natural image captured by any camera, so an attempt of forging an image makes these properties un-natural. By this method, an original and forged images can be easily classified by state-of-the-art machine learning models trained on these features. The performance of this method is quantitatively assessed using the famous evaluation metrics i-e accuracy, TPR, FPR, TNR, Recall and F1-score. A comparison with other state-of-the-art techniques has shown that the proposed technique has shown better results in comparison with the other techniques.","PeriodicalId":113717,"journal":{"name":"2023 IEEE International Conference on Electro Information Technology (eIT)","volume":"51 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131499276","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-18DOI: 10.1109/eIT57321.2023.10187367
Cole Fulton, Lara Taha, T. Emami
This paper presents project-based Learning on solar energy for undergraduate junior-level Electrical Engineering students. In this project, students conduct experiments to analyze and plot the impact of distance, incidence angle, and source brightness and shading on the given small solar panel. Second, they conduct experiments to investigate the effect of temperature on the solar panel. Third, students conduct experiments to analyze solar panel characteristics in series and parallel connections. Fourth, they design and build a circuit with different loads to explore solar panel characteristics by measuring a load's voltage, current, and power. Finally, they compare the electric powers produced by artificial and natural light sources in part of the measurement. The data collection utilizes an Arduino Uno, and an Adafruit DC sensor to ensure human errors are at a minimum.
{"title":"A Project-Based Learning on Solar Energy from Different Light Sources","authors":"Cole Fulton, Lara Taha, T. Emami","doi":"10.1109/eIT57321.2023.10187367","DOIUrl":"https://doi.org/10.1109/eIT57321.2023.10187367","url":null,"abstract":"This paper presents project-based Learning on solar energy for undergraduate junior-level Electrical Engineering students. In this project, students conduct experiments to analyze and plot the impact of distance, incidence angle, and source brightness and shading on the given small solar panel. Second, they conduct experiments to investigate the effect of temperature on the solar panel. Third, students conduct experiments to analyze solar panel characteristics in series and parallel connections. Fourth, they design and build a circuit with different loads to explore solar panel characteristics by measuring a load's voltage, current, and power. Finally, they compare the electric powers produced by artificial and natural light sources in part of the measurement. The data collection utilizes an Arduino Uno, and an Adafruit DC sensor to ensure human errors are at a minimum.","PeriodicalId":113717,"journal":{"name":"2023 IEEE International Conference on Electro Information Technology (eIT)","volume":"402 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121981577","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-18DOI: 10.1109/eIT57321.2023.10187233
Thinh Phan, R. Green
To monitor male and female bird nest attendance, the traditional methods are physical markings for identification. This paper presents two methods-Principal Component Analysis (PCA) combined with K Nearest Neighbor (KNN) and Cross-Correlation classification-that can identify individual birds based on the sounds of their wing flaps without the need for physically marking the birds. The study conducted on three male Zebra Finch birds resulted in identification accuracy ranging from 70% to 100%. To distinguish between individual birds, the conventional invasive technique involves capturing, marking, releasing, and recapturing. However, this approach has various limitations and drawbacks. As an alternative solution, researchers have resorted to using bird vocalizations for identification purposes. This research shows that birds can also be uniquely identified from the sounds produced by their wing flaps.
{"title":"Using Wing Flap Sounds to Distinguish Individual Birds","authors":"Thinh Phan, R. Green","doi":"10.1109/eIT57321.2023.10187233","DOIUrl":"https://doi.org/10.1109/eIT57321.2023.10187233","url":null,"abstract":"To monitor male and female bird nest attendance, the traditional methods are physical markings for identification. This paper presents two methods-Principal Component Analysis (PCA) combined with K Nearest Neighbor (KNN) and Cross-Correlation classification-that can identify individual birds based on the sounds of their wing flaps without the need for physically marking the birds. The study conducted on three male Zebra Finch birds resulted in identification accuracy ranging from 70% to 100%. To distinguish between individual birds, the conventional invasive technique involves capturing, marking, releasing, and recapturing. However, this approach has various limitations and drawbacks. As an alternative solution, researchers have resorted to using bird vocalizations for identification purposes. This research shows that birds can also be uniquely identified from the sounds produced by their wing flaps.","PeriodicalId":113717,"journal":{"name":"2023 IEEE International Conference on Electro Information Technology (eIT)","volume":"113 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121101797","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-18DOI: 10.1109/eIT57321.2023.10187385
M. Chowdhury, Nafiz Rifat, Shadman Latif, M. Ahsan, Md Saifur Rahman, Rahul Gomes
The field of Natural Language Processing has observed significant advancements in the development of sophisticated conversational Artificial Intelligence systems. ChatGPT is one such state-of-the-art conversational system that has attracted considerable interest and adoption. It enables developers to create highly interactive and engaging conversational applications using deep neural networks to produce human-like responses to user inputs. Such capabilities have made it popular in the threat actors' world. However, threat actors can abuse this chatbot to generate attack vectors as part of an operation. ChatGPT can be abused to produce practical and realistic communications that can be used in phishing attacks. These communications help the attack vectors distribution, i.e., prompt users to download and set up malware or disclose confidential information. ChatGPT has security measures to prevent malicious queries from generating attack vectors. However, the threat actors can circumvent such security controls through deception. This abusive use of ChatGPT makes the supply chain management of attack vectors effective and efficient. In this study, we presented evidence from various sources, showing how ChatGPT is abused to help the threat actors to improve each step of the attack vectors' supply chain management.
{"title":"ChatGPT: The Curious Case of Attack Vectors' Supply Chain Management Improvement","authors":"M. Chowdhury, Nafiz Rifat, Shadman Latif, M. Ahsan, Md Saifur Rahman, Rahul Gomes","doi":"10.1109/eIT57321.2023.10187385","DOIUrl":"https://doi.org/10.1109/eIT57321.2023.10187385","url":null,"abstract":"The field of Natural Language Processing has observed significant advancements in the development of sophisticated conversational Artificial Intelligence systems. ChatGPT is one such state-of-the-art conversational system that has attracted considerable interest and adoption. It enables developers to create highly interactive and engaging conversational applications using deep neural networks to produce human-like responses to user inputs. Such capabilities have made it popular in the threat actors' world. However, threat actors can abuse this chatbot to generate attack vectors as part of an operation. ChatGPT can be abused to produce practical and realistic communications that can be used in phishing attacks. These communications help the attack vectors distribution, i.e., prompt users to download and set up malware or disclose confidential information. ChatGPT has security measures to prevent malicious queries from generating attack vectors. However, the threat actors can circumvent such security controls through deception. This abusive use of ChatGPT makes the supply chain management of attack vectors effective and efficient. In this study, we presented evidence from various sources, showing how ChatGPT is abused to help the threat actors to improve each step of the attack vectors' supply chain management.","PeriodicalId":113717,"journal":{"name":"2023 IEEE International Conference on Electro Information Technology (eIT)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130829098","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-18DOI: 10.1109/eIT57321.2023.10187232
Steven B. Poore, Cristinel Ababei
In this paper, we investigate the use of machine learning (ML) models to estimate or predict concentrations of pollutants based on measured concentrations of other pollutants. Such models could be used in air quality index (AQI) detection systems to decrease the number of physical sensors in order to reduce overall and maintenance costs. Five different long-short term memory (LSTM) models were explored in the preliminary investigation. The most accurate model was then selected for further refinement via simple hyperparameter search. The final refined model was trained and tested on four different air quality datasets from four different countries. Simulation results indicate that prediction of pollutant concentrations based solely on measured concentrations of other pollutants is not accurate enough to warrant total sensor replacement with ML models. However, when the same ML models are provided as input past measurements of the predicted pollutant rather than previously predicted values, the prediction accuracy is excellent. We conclude that while ML models are not yet accurate enough to completely replace physical sensors, such models could be very helpful to provide predictions in situations of sensor failure and thus to guarantee continuous sensor fusion processes.
{"title":"Can Machine Learning Models be Used to Predict Pollutants based on Measured Other Pollutants?","authors":"Steven B. Poore, Cristinel Ababei","doi":"10.1109/eIT57321.2023.10187232","DOIUrl":"https://doi.org/10.1109/eIT57321.2023.10187232","url":null,"abstract":"In this paper, we investigate the use of machine learning (ML) models to estimate or predict concentrations of pollutants based on measured concentrations of other pollutants. Such models could be used in air quality index (AQI) detection systems to decrease the number of physical sensors in order to reduce overall and maintenance costs. Five different long-short term memory (LSTM) models were explored in the preliminary investigation. The most accurate model was then selected for further refinement via simple hyperparameter search. The final refined model was trained and tested on four different air quality datasets from four different countries. Simulation results indicate that prediction of pollutant concentrations based solely on measured concentrations of other pollutants is not accurate enough to warrant total sensor replacement with ML models. However, when the same ML models are provided as input past measurements of the predicted pollutant rather than previously predicted values, the prediction accuracy is excellent. We conclude that while ML models are not yet accurate enough to completely replace physical sensors, such models could be very helpful to provide predictions in situations of sensor failure and thus to guarantee continuous sensor fusion processes.","PeriodicalId":113717,"journal":{"name":"2023 IEEE International Conference on Electro Information Technology (eIT)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132910309","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-18DOI: 10.1109/eIT57321.2023.10187226
A. Sethi, S. Damani, Arshia K. Sethi, Anjali Rajagopal, K. Gopalakrishnan, A. Cherukuri, S. P. Arunachalam
More than 11% of Americans are affected by diseases related to the gastrointestinal (GI) tract. GI endoscopy is an established imaging modality for diagnostic and therapeutic procedures. Large volumes of images and videos generated during this procedure, makes image interpretation cumbersome and varies among physicians. Artificial intelligence (AI) assisted Computer-Aided Diagnosis (CAD) system for digital GI endoscopy is gaining attention that can disrupt GI practice. Several studies have reported the application of computer vision and machine learning algorithms in GI endoscopy. Endoscopic images of varying anatomic features of the Gi tract, challenges their accurate classification. Therefore, a need exists in accurately classifying different GI endoscopic images for upstream processing in the diagnostic platform for digital GI endoscopy. The purpose of this work was to develop a deep learning model using convolutional neural network (CNN) and wavelet decomposed CNN for improved accuracy using publically available GI endoscopic images from Kvasir dataset with 8 different image groups namely Z-line, Pylorus, Cecum, Esophagitis, Polyps, Ulcerative Colitis, Dyed and Lifted Polyps & Dyed Resection Margins. Wavelet decomposition along with CNN architecture allows utilization of spectral information which is mostly lost in conventional CNNs that can enhance model performance. The models were trained with 80% images and 20% were used for testing and accuracy was compared. 10% improvement in accuracy for multi-class classification was observed with wavelet CNN model compared to conventional CNN. The results indicate the potential of image decomposition methods for enhancing digital GI endoscopic procedures.
{"title":"Gastrointestinal Endoscopic Image Classification using a Novel Wavelet Decomposition Based Deep Learning Algorithm","authors":"A. Sethi, S. Damani, Arshia K. Sethi, Anjali Rajagopal, K. Gopalakrishnan, A. Cherukuri, S. P. Arunachalam","doi":"10.1109/eIT57321.2023.10187226","DOIUrl":"https://doi.org/10.1109/eIT57321.2023.10187226","url":null,"abstract":"More than 11% of Americans are affected by diseases related to the gastrointestinal (GI) tract. GI endoscopy is an established imaging modality for diagnostic and therapeutic procedures. Large volumes of images and videos generated during this procedure, makes image interpretation cumbersome and varies among physicians. Artificial intelligence (AI) assisted Computer-Aided Diagnosis (CAD) system for digital GI endoscopy is gaining attention that can disrupt GI practice. Several studies have reported the application of computer vision and machine learning algorithms in GI endoscopy. Endoscopic images of varying anatomic features of the Gi tract, challenges their accurate classification. Therefore, a need exists in accurately classifying different GI endoscopic images for upstream processing in the diagnostic platform for digital GI endoscopy. The purpose of this work was to develop a deep learning model using convolutional neural network (CNN) and wavelet decomposed CNN for improved accuracy using publically available GI endoscopic images from Kvasir dataset with 8 different image groups namely Z-line, Pylorus, Cecum, Esophagitis, Polyps, Ulcerative Colitis, Dyed and Lifted Polyps & Dyed Resection Margins. Wavelet decomposition along with CNN architecture allows utilization of spectral information which is mostly lost in conventional CNNs that can enhance model performance. The models were trained with 80% images and 20% were used for testing and accuracy was compared. 10% improvement in accuracy for multi-class classification was observed with wavelet CNN model compared to conventional CNN. The results indicate the potential of image decomposition methods for enhancing digital GI endoscopic procedures.","PeriodicalId":113717,"journal":{"name":"2023 IEEE International Conference on Electro Information Technology (eIT)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127910529","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：