Asma Abdulelah Abdulrahman
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Asma Abdulelah Abdulrahman has a total of 21 co-authors. Here are the top co-authors by number of shared publications:
- Fouad Shaker Tahir
University of Technology
15 shared publications - Yurii Khlaponin
State University of Trade and Economics
11 shared publications - Hassan Mohamed Muhi-Aldeen
Aliraqia University
10 shared publications - Jabbar Abed Eleiwy
University of Technology
4 shared publications - Mayada Taki Wazi
Department of ElectreMechanical Engineering, University of Technology, Iraq
3 shared publications Fouad Shaker Tahir
University of Technology
2 shared publications
Latest Publications
- Open Access
- 30 July 2026
- Published by OU Scientific Route in EUREKA: Physics and Engineering
- No. 4,p. 206-215
Development of a hybrid convolutional neural network assisted wavelet architecture for robust inorganic waste classification
The increasing volume of waste globally poses a significant challenge in reducing environmental pollution. Recycling waste or converting it into energy is crucial for achieving the Sustainable Development Goals. This research addresses this issue through artificial intelligence. The proposed solution involves a hybrid system utilizing the Discrete Legendre Microwave Transform (DLEGWT) derived from Legendre Polynomials (LEGP) for the initial processing of waste images using a novel filter. Multi-resolution analysis (MRA) is performed to analyze the image into approximate parameters representing the LL and detailed parameters representing the three quadrants: HL, LH, and HH. Noise is then removed, and the image is compressed to accommodate the new filter. The main image quality parameters are mean squared error (MSE), peak signal-to-noise ratio (PSNR), number of bits per pixel, and compression ratio (CR). After the initial processing stage, the classification stage begins using deep learning. A Google Net, consisting of 144 layers generated by convolution, is trained using the newly proposed Discrete Legendre Waveform Transform Filter (CNNDLEGWT), taking image size into account, to classify waste into four groups: plastic, bottles, spoons, and cups. The accuracy of each resulting image is 94.12%, compared to the accuracy of the neural network. This work contributes to solving the problem of waste being divided into organic and inorganic waste, but this work specifically uses the proposed technique to separate inorganic waste, namely plastic, glass, paper, and metal, with good accuracy. The promising results obtained confirm and demonstrate the efficiency of the proposed system.
- 30 July 2026
- Published by OU Scientific Route in EUREKA: Physics and Engineering
- No. 4,p. 206-215
Open AccessDevelopment of a hybrid convolutional neural network assisted wavelet architecture for robust inorganic waste classification
The increasing volume of waste globally poses a significant challenge in reducing environmental pollution. Recycling waste or converting it into energy is crucial for achieving the Sustainable Development Goals. This research addresses this issue through artificial intelligence. The proposed solution involves a hybrid system utilizing the Discrete Legendre Microwave Transform (DLEGWT) derived from Legendre Polynomials (LEGP) for the initial processing of waste images using a novel filter. Multi-resolution analysis (MRA) is performed to analyze the image into approximate parameters representing the LL and detailed parameters representing the three quadrants: HL, LH, and HH. Noise is then removed, and the image is compressed to accommodate the new filter. The main image quality parameters are mean squared error (MSE), peak signal-to-noise ratio (PSNR), number of bits per pixel, and compression ratio (CR). After the initial processing stage, the classification stage begins using deep learning. A Google Net, consisting of 144 layers generated by convolution, is trained using the newly proposed Discrete Legendre Waveform Transform Filter (CNNDLEGWT), taking image size into account, to classify waste into four groups: plastic, bottles, spoons, and cups. The accuracy of each resulting image is 94.12%, compared to the accuracy of the neural network. This work contributes to solving the problem of waste being divided into organic and inorganic waste, but this work specifically uses the proposed technique to separate inorganic waste, namely plastic, glass, paper, and metal, with good accuracy. The promising results obtained confirm and demonstrate the efficiency of the proposed system.
- Open Access
- 11 May 2026
- Published by OU Scientific Route in EUREKA: Physics and Engineering
- No. 3,p. 201-213
Developing a deep learning approach for detecting rice crop diseases using discrete wavelet transform
The research topic is to derive new wavelets Discrete Boubaker Wavelet Transform (DBWT) from the proposed polynomials Boubaker polynomials (BP) for constructing convolutional neural networks to be used in identifying rice plant diseases by processing images of healthy and diseased rice leaves to classify these images and determine the type of disease. The problem addressed in this work is the delay in detecting diseases affecting the aforementioned plants. To address this problem, the proposed theory is used to detect the disease early and with high accuracy, enabling rapid treatment before the disease worsens. After the input image is passed through the image analysis and enhancement stage using discrete wavelet transforms to produce a good image based on the frequencies that characterize the waves, the input image is compressed to reduce the spatial space occupied by the image data, to validate the convolution process and improve the results in terms of accuracy. Therefore, the results obtained have contributed significantly to plant disease identification. After creating the discrete Boubaker wavelet transform (DBWT) for processing agricultural crop images, the image data was analyzed and clarified after denoising. The most important results were recorded: maximum signal-to-noise ratio (PSNR), mean square error (MSE), bet per pixel ratio (BPP), and compression ratio (CR). To achieve image quality, the best results were obtained. The new filter was used in the convolution process to form the convolutional neural network. Layers were then generated for the deep learning phase, resulting in a fully connected layer for identifying and classifying agricultural crop diseases with high accuracy, enabling early treatment
- 11 May 2026
- Published by OU Scientific Route in EUREKA: Physics and Engineering
- No. 3,p. 201-213
Open AccessDeveloping a deep learning approach for detecting rice crop diseases using discrete wavelet transform
The research topic is to derive new wavelets Discrete Boubaker Wavelet Transform (DBWT) from the proposed polynomials Boubaker polynomials (BP) for constructing convolutional neural networks to be used in identifying rice plant diseases by processing images of healthy and diseased rice leaves to classify these images and determine the type of disease. The problem addressed in this work is the delay in detecting diseases affecting the aforementioned plants. To address this problem, the proposed theory is used to detect the disease early and with high accuracy, enabling rapid treatment before the disease worsens. After the input image is passed through the image analysis and enhancement stage using discrete wavelet transforms to produce a good image based on the frequencies that characterize the waves, the input image is compressed to reduce the spatial space occupied by the image data, to validate the convolution process and improve the results in terms of accuracy. Therefore, the results obtained have contributed significantly to plant disease identification. After creating the discrete Boubaker wavelet transform (DBWT) for processing agricultural crop images, the image data was analyzed and clarified after denoising. The most important results were recorded: maximum signal-to-noise ratio (PSNR), mean square error (MSE), bet per pixel ratio (BPP), and compression ratio (CR). To achieve image quality, the best results were obtained. The new filter was used in the convolution process to form the convolutional neural network. Layers were then generated for the deep learning phase, resulting in a fully connected layer for identifying and classifying agricultural crop diseases with high accuracy, enabling early treatment
- Open Access
- 10 February 2026
- Published by OU Scientific Route in EUREKA: Physics and Engineering
- No. 2,p. 200-209
The development of a new technique for detecting bone cancer quickly using a specific clustered discrete Boubaker wavelet transform (CDBWT) method
The aim of this work is to detect bone cancer using a new bone image processing technology technique in bone image processing to aid in rapid and early detection and treatment of bone cancer. The drawbacks of current methods for detecting bone cancer include slow speed and limited accuracy. This research addresses these drawbacks with a new image processing technique to speed up detection and increase accuracy by addressing noise in the input images and applying compression to reduce the space occupied by large datasets in the images, leading to improved image quality and faster diagnosis. The new technique, the clustered discrete wavelet Boubaker transform (CDBWT), has shown significant improvements, with accuracy increasing from 96.13% to 99.90%, average accuracy from 95.70% to 97.95%, and intersection-over-union (IoU) accuracy from 96.56% to 99.81%. The resulting image quality results, including peak signal-to-noise ratio (PSNR), mean square error (MSE), bits per pixel (BPP), and compression ratio (CR), are ideal. The accuracy achieved with the new technique after combining it with a k-means clustering process after preprocessing, which involves a multi-resolution analysis (MRA) process that transforms images into high and low frequencies, helped improve the image quality to be free of noise and compression, thus improving the accuracy of bone cancer identification in the input images. Due to the effectiveness of the new technique, high accuracy was achieved by obtaining image quality parameters such as PSNR, MSE, BPP, and CR obtained in the pre-processing with MRA phase. This was achieved by reducing noise and compression to provide clearer images, improving the detection process for bone cancer. The new technique was combined with k-means clustering, making it an effective technique for rapid and early detection of bone cancer
- 10 February 2026
- Published by OU Scientific Route in EUREKA: Physics and Engineering
- No. 2,p. 200-209
Open AccessThe development of a new technique for detecting bone cancer quickly using a specific clustered discrete Boubaker wavelet transform (CDBWT) method
The aim of this work is to detect bone cancer using a new bone image processing technology technique in bone image processing to aid in rapid and early detection and treatment of bone cancer. The drawbacks of current methods for detecting bone cancer include slow speed and limited accuracy. This research addresses these drawbacks with a new image processing technique to speed up detection and increase accuracy by addressing noise in the input images and applying compression to reduce the space occupied by large datasets in the images, leading to improved image quality and faster diagnosis. The new technique, the clustered discrete wavelet Boubaker transform (CDBWT), has shown significant improvements, with accuracy increasing from 96.13% to 99.90%, average accuracy from 95.70% to 97.95%, and intersection-over-union (IoU) accuracy from 96.56% to 99.81%. The resulting image quality results, including peak signal-to-noise ratio (PSNR), mean square error (MSE), bits per pixel (BPP), and compression ratio (CR), are ideal. The accuracy achieved with the new technique after combining it with a k-means clustering process after preprocessing, which involves a multi-resolution analysis (MRA) process that transforms images into high and low frequencies, helped improve the image quality to be free of noise and compression, thus improving the accuracy of bone cancer identification in the input images. Due to the effectiveness of the new technique, high accuracy was achieved by obtaining image quality parameters such as PSNR, MSE, BPP, and CR obtained in the pre-processing with MRA phase. This was achieved by reducing noise and compression to provide clearer images, improving the detection process for bone cancer. The new technique was combined with k-means clustering, making it an effective technique for rapid and early detection of bone cancer
- Open Access
- 30 January 2026
- Published by OU Scientific Route in EUREKA: Physics and Engineering
- No. 1,p. 206-216
Development of data envelopment analysis model for measuring healthcare efficiency using fuzzy randomization
This study presents a data envelopment analysis (DEA) for measuring productivity based on the SBM model, integrated with the DEMATEL-ANP hybrid method. It was used to evaluate several hospitals by calculating the number of hospital visits, a crucial metric for the proposed model, which proved its effectiveness. A random SBM model is proposed in this study to manage uncertain conditions. This study proposes a random SBM system to evaluate performance using the DEMATEL and ANP methods to identify critical criteria. Given the large number of key criteria identified for performance evaluation, a process based on these two methods is employed to determine the most important criteria. The primary objective is to identify the most effective criteria in the hospital sector, building upon previous studies that identified 16 key criteria affecting hospital efficiency. The DEMATEL/ANP method identified the number of patient visits as the most important criterion, with an initial weight of 0.237 and an average efficiency of 0.847 ± 0.156. A comparison using the deterministic SBM model revealed a bias of 7.1%. Hospital 3 achieved an efficiency of 0.642, while Hospital 8 achieved an efficiency of 0.981. After applying the uncertainty model, four hospitals demonstrated complete efficiency. The problems were solved by achieving efficiency ratios of 34% and preventing Type I errors in 27% of decision-making performance. The integration of DEMATEL-ANP further reduced this to achieve 94% efficiency. On the practical side, the results regarding resource efficiency rely on reliability indicators, improvement priorities based on standards and evidence, and the measurement of operational risks through probabilistic confidence intervals. For policymakers, evaluating healthcare system performance takes into account the degree of fluctuation in demand. The proposed system demonstrates several quantitative advantages: 15% higher discriminatory power, 23% better correlation with post-assessment clinical outcomes (r = 0.712, p < 0.01), and a reduction in negatives from 18% to 4%. Additionally, it offers 81% accuracy in uncertainty predictions, compared to 58% accuracy for deterministic models
- 30 January 2026
- Published by OU Scientific Route in EUREKA: Physics and Engineering
- No. 1,p. 206-216
Open AccessDevelopment of data envelopment analysis model for measuring healthcare efficiency using fuzzy randomization
This study presents a data envelopment analysis (DEA) for measuring productivity based on the SBM model, integrated with the DEMATEL-ANP hybrid method. It was used to evaluate several hospitals by calculating the number of hospital visits, a crucial metric for the proposed model, which proved its effectiveness. A random SBM model is proposed in this study to manage uncertain conditions. This study proposes a random SBM system to evaluate performance using the DEMATEL and ANP methods to identify critical criteria. Given the large number of key criteria identified for performance evaluation, a process based on these two methods is employed to determine the most important criteria. The primary objective is to identify the most effective criteria in the hospital sector, building upon previous studies that identified 16 key criteria affecting hospital efficiency. The DEMATEL/ANP method identified the number of patient visits as the most important criterion, with an initial weight of 0.237 and an average efficiency of 0.847 ± 0.156. A comparison using the deterministic SBM model revealed a bias of 7.1%. Hospital 3 achieved an efficiency of 0.642, while Hospital 8 achieved an efficiency of 0.981. After applying the uncertainty model, four hospitals demonstrated complete efficiency. The problems were solved by achieving efficiency ratios of 34% and preventing Type I errors in 27% of decision-making performance. The integration of DEMATEL-ANP further reduced this to achieve 94% efficiency. On the practical side, the results regarding resource efficiency rely on reliability indicators, improvement priorities based on standards and evidence, and the measurement of operational risks through probabilistic confidence intervals. For policymakers, evaluating healthcare system performance takes into account the degree of fluctuation in demand. The proposed system demonstrates several quantitative advantages: 15% higher discriminatory power, 23% better correlation with post-assessment clinical outcomes (r = 0.712, p < 0.01), and a reduction in negatives from 18% to 4%. Additionally, it offers 81% accuracy in uncertainty predictions, compared to 58% accuracy for deterministic models
- Open Access
- 29 September 2025
- Published by OU Scientific Route in EUREKA: Physics and Engineering
- No. 6,p. 181-194
Development of algorithm for early detection of tobacco diseases in Iraq with wavelet and deep learning technology
The subject of the research is early and accurate detection of tobacco diseases, which is a very important economic matter for achieving rapid treatment, preventing economic damage to farmers due to crop loss. Manual examination is one of the traditional methods for detecting tobacco diseases, which leads to loss of time and is subject to errors in the results, which leads to delay in treating the problem. An effective study was conducted to detect tobacco plant diseases using a neural network concerned with collecting image data represented by classifying diseased leaves from healthy leaves in Iraq, which aims to achieve high accuracy for early detection of these diseases. This work aims to propose an advanced and effective technique to detect known tobacco diseases in Iraq by constructing new discrete waves discrete Legendre wavelet transform (DLEGWT) derived from Legendre polynomials using mathematical methods and deriving a new filter to analyze the image and remove noise and collect image data such as statistics in noise removal and image compression to reduce the space it occupies. The data from these images and the new convolutional neural network discrete wavelet (NCNNDW) are trained with the new filter to complete the convolution process and create the primary and hidden layers and then the fully connected layers after determining the tobacco leaf image data so that the infected leaves are identified in different regions of Iraq with high accuracy after completing the process of classifying and detecting tobacco diseases. This is done by creating an advanced algorithm in the MATLAB program to complete the classification process for tobacco plant diseases and obtaining an accuracy of 88.24% in 41 seconds
- 29 September 2025
- Published by OU Scientific Route in EUREKA: Physics and Engineering
- No. 6,p. 181-194
Open AccessDevelopment of algorithm for early detection of tobacco diseases in Iraq with wavelet and deep learning technology
The subject of the research is early and accurate detection of tobacco diseases, which is a very important economic matter for achieving rapid treatment, preventing economic damage to farmers due to crop loss. Manual examination is one of the traditional methods for detecting tobacco diseases, which leads to loss of time and is subject to errors in the results, which leads to delay in treating the problem. An effective study was conducted to detect tobacco plant diseases using a neural network concerned with collecting image data represented by classifying diseased leaves from healthy leaves in Iraq, which aims to achieve high accuracy for early detection of these diseases. This work aims to propose an advanced and effective technique to detect known tobacco diseases in Iraq by constructing new discrete waves discrete Legendre wavelet transform (DLEGWT) derived from Legendre polynomials using mathematical methods and deriving a new filter to analyze the image and remove noise and collect image data such as statistics in noise removal and image compression to reduce the space it occupies. The data from these images and the new convolutional neural network discrete wavelet (NCNNDW) are trained with the new filter to complete the convolution process and create the primary and hidden layers and then the fully connected layers after determining the tobacco leaf image data so that the infected leaves are identified in different regions of Iraq with high accuracy after completing the process of classifying and detecting tobacco diseases. This is done by creating an advanced algorithm in the MATLAB program to complete the classification process for tobacco plant diseases and obtaining an accuracy of 88.24% in 41 seconds
- Open Access
- 23 September 2025
- Published by OU Scientific Route in EUREKA: Physics and Engineering
- No. 5,p. 171-180
A novel approach and implementation of discrete wavelets transform (DWT) for clustering and compression of kidney disease images
Kidney disease posed a major problem in the field of medicine in terms of the inability to limit the location of the disease in the diseased kidney. To solve this problem, the clustering technique was applied to the image of the diseased kidney to find the value of k in the theory of Clustering With K-means using the MATLAB program. To improve the results in this work, it was derived A new technique is to create a new filter derived from the first Chebyshev polynomials using high-tech mathematical methods to obtain into New Discrete First Chebyshev Wavelets Transform (NDFCWT). The large data contained in the kidney sample, and to reduce the areas exploited by this data, the image was compressed, and after the completion of the compression process, the image quality standards are recorded, which are peak signal to noise ratio (PSNR), mean square error (MSE), compression ratio (CR) and bit-per-pixel (BPP). The study comprehensively evaluates its performance through extensive computational experiments across multiple datasets, including high-dimensional, noisy, and complex data structures. The results are compared before and after using the proposed technique, MSE before used new method 5.131. However, after used new method which resulted in improved results in obtaining value K-means is 0.950. The proposed approach, demonstrates significant improvements in clustering accuracy, computational efficiency, and noise resistance compared to existing wavelet transform techniques. Establishment of the mathematical formulation of the new discrete first Chebyshev wavelet transform (NDFCWT) with the ability to process digital image signals, The NDFCWT achieves superior performance through different calculations compared to existing wavelets, detailed insights into the algorithmic mechanics are the scientific and mathematical foundations of the proposed method
- 23 September 2025
- Published by OU Scientific Route in EUREKA: Physics and Engineering
- No. 5,p. 171-180
Open AccessA novel approach and implementation of discrete wavelets transform (DWT) for clustering and compression of kidney disease images
Kidney disease posed a major problem in the field of medicine in terms of the inability to limit the location of the disease in the diseased kidney. To solve this problem, the clustering technique was applied to the image of the diseased kidney to find the value of k in the theory of Clustering With K-means using the MATLAB program. To improve the results in this work, it was derived A new technique is to create a new filter derived from the first Chebyshev polynomials using high-tech mathematical methods to obtain into New Discrete First Chebyshev Wavelets Transform (NDFCWT). The large data contained in the kidney sample, and to reduce the areas exploited by this data, the image was compressed, and after the completion of the compression process, the image quality standards are recorded, which are peak signal to noise ratio (PSNR), mean square error (MSE), compression ratio (CR) and bit-per-pixel (BPP). The study comprehensively evaluates its performance through extensive computational experiments across multiple datasets, including high-dimensional, noisy, and complex data structures. The results are compared before and after using the proposed technique, MSE before used new method 5.131. However, after used new method which resulted in improved results in obtaining value K-means is 0.950. The proposed approach, demonstrates significant improvements in clustering accuracy, computational efficiency, and noise resistance compared to existing wavelet transform techniques. Establishment of the mathematical formulation of the new discrete first Chebyshev wavelet transform (NDFCWT) with the ability to process digital image signals, The NDFCWT achieves superior performance through different calculations compared to existing wavelets, detailed insights into the algorithmic mechanics are the scientific and mathematical foundations of the proposed method
- Open Access
- 28 March 2025
- Published by OU Scientific Route in EUREKA: Physics and Engineering
- No. 2,p. 211-220
Deep learning algorithm for detecting skin diseases with new DWT technology
Timely diagnosis of skin diseases and the importance of their treatment with the discrete wavelet transform technology with deep learning through the classification process to pass through two stages, which are the initial processing with discrete wavelets to start the process of Multi resolution analyses (MRA) to analyze the input image to low pass filter and high pass filter the first represents the approach coefficients low low and the second represents the detail coefficients to be high low low high and high high to perform the two processes of denoising the input image and compressing it. One of the skin diseases is skin cancer, which is considered the most dangerous, but if it is not diagnosed early, it can spread to the rest of the body, leading to death. Therefore, the most dangerous type of cancer is skin cancer, and the death rate due to this cancer is 75 %. In this study, early detection and diagnosis of skin cancer was performed with a new and advanced technique, which is discrete Chebyshev wavelet transform (DCHWT) derived from Chebyshev polynomials (CHP) with mathematical methods to obtain new wavelets to create a new filter from these wavelets. The new filter was used to improve the image of affected skin. With the help of this filter, the affected part of the skin is distinguished using the MATLAB program with high accuracy, through deep learning and training on a new convolutional neural network (CNN) with a proposed filter to distinguish the affected part of the skin, which can be called Convolution neural network discrete Chebyshev wavelet transform (CNNDCHWT) in deep learning, where the accuracy is calculated and the percentage was obtained at 98.58 % in a time of 34 seconds, which proves the efficiency of the proposed theory
- 28 March 2025
- Published by OU Scientific Route in EUREKA: Physics and Engineering
- No. 2,p. 211-220
Open AccessDeep learning algorithm for detecting skin diseases with new DWT technology
Timely diagnosis of skin diseases and the importance of their treatment with the discrete wavelet transform technology with deep learning through the classification process to pass through two stages, which are the initial processing with discrete wavelets to start the process of Multi resolution analyses (MRA) to analyze the input image to low pass filter and high pass filter the first represents the approach coefficients low low and the second represents the detail coefficients to be high low low high and high high to perform the two processes of denoising the input image and compressing it. One of the skin diseases is skin cancer, which is considered the most dangerous, but if it is not diagnosed early, it can spread to the rest of the body, leading to death. Therefore, the most dangerous type of cancer is skin cancer, and the death rate due to this cancer is 75 %. In this study, early detection and diagnosis of skin cancer was performed with a new and advanced technique, which is discrete Chebyshev wavelet transform (DCHWT) derived from Chebyshev polynomials (CHP) with mathematical methods to obtain new wavelets to create a new filter from these wavelets. The new filter was used to improve the image of affected skin. With the help of this filter, the affected part of the skin is distinguished using the MATLAB program with high accuracy, through deep learning and training on a new convolutional neural network (CNN) with a proposed filter to distinguish the affected part of the skin, which can be called Convolution neural network discrete Chebyshev wavelet transform (CNNDCHWT) in deep learning, where the accuracy is calculated and the percentage was obtained at 98.58 % in a time of 34 seconds, which proves the efficiency of the proposed theory
- Open Access
- 31 January 2025
- Published by OU Scientific Route in EUREKA: Physics and Engineering
- No. 1,p. 152-162
Enhancement of watermarking based on deep neural network for images
The spread of the Internet of Things applications and the shared massive amount of data leads to the absence of data security and missing product protection, such as copyrights, pharmaceutical products, and online medical result fraud. To address the security gap, it is possible to propose a model based on developing a new filter derived from Laguerre polynomials based on essential mathematical operations by using the parent function to construct a new discrete Laguerre wavelets transform (DLWT) in image analysis with the watermark, as well as reaching an ideal algorithm through the readings from the deep learning convolutional neural network construction. New wavelets are presented, in the resultant network. The technique of separating the host image from the watermark is compared before and after separation is known as the convolution neural network wavelet transform (CNNWT). Peak signal-to-noise ratio (PSNR) and natural correlation coefficient (NCC) values are the most crucial metrics from these results that demonstrate the effectiveness of the suggested algorithm CNNWT with the QR code and the MATLAB is used in this approach. In this work, the watermark is immersed in the host image with the new discrete wavelets and the convolutional neural network, then the attacks are directed to complete the logistical process of recourse the results of PSNR and the values of the Natural Correlation Coefficient (NCC), reaching NC equal to one and worthy PSNR values. The results obtained demonstrate the effectiveness of the suggested theory in this study the quality of the image and the watermark are preserved, which proves the efficiency of the proposed theory in addition, the accuracy is reached 98.64 % with a time of 23 seconds in an epoch.
- 31 January 2025
- Published by OU Scientific Route in EUREKA: Physics and Engineering
- No. 1,p. 152-162
Open AccessEnhancement of watermarking based on deep neural network for images
The spread of the Internet of Things applications and the shared massive amount of data leads to the absence of data security and missing product protection, such as copyrights, pharmaceutical products, and online medical result fraud. To address the security gap, it is possible to propose a model based on developing a new filter derived from Laguerre polynomials based on essential mathematical operations by using the parent function to construct a new discrete Laguerre wavelets transform (DLWT) in image analysis with the watermark, as well as reaching an ideal algorithm through the readings from the deep learning convolutional neural network construction. New wavelets are presented, in the resultant network. The technique of separating the host image from the watermark is compared before and after separation is known as the convolution neural network wavelet transform (CNNWT). Peak signal-to-noise ratio (PSNR) and natural correlation coefficient (NCC) values are the most crucial metrics from these results that demonstrate the effectiveness of the suggested algorithm CNNWT with the QR code and the MATLAB is used in this approach. In this work, the watermark is immersed in the host image with the new discrete wavelets and the convolutional neural network, then the attacks are directed to complete the logistical process of recourse the results of PSNR and the values of the Natural Correlation Coefficient (NCC), reaching NC equal to one and worthy PSNR values. The results obtained demonstrate the effectiveness of the suggested theory in this study the quality of the image and the watermark are preserved, which proves the efficiency of the proposed theory in addition, the accuracy is reached 98.64 % with a time of 23 seconds in an epoch.
- Open Access
- 19 July 2024
- Published by OU Scientific Route in EUREKA: Physics and Engineering
- No. 4,p. 184-192
Classification of food items in trash through images using a hybrid deep learning system
The issue that arouses concern is the issue of waste in general, especially waste in food. Deep learning has made its way into the aforementioned field. Pre-processing is done using a new filter derived from discrete wavelets that were derived from Legendre polynomials (LEGP) to achieve Multi Resolution Analyses (MRA), in which the image parameters are separated into proximity parameters and detail parameters, where the first is concentrated in LL, while the second is in HL, LH, HH to improve the input image of the waste in terms of noise removal and image compression. To begin the deep learning process, configure the New Convolutional Neural Network (NCNN) to configure 144 layer with the image size (224x224x3) with Google Net so that the waste resulting from leftover food is sorted. In this work, four foods were sorted: tomatoes, potatoes, bread, and rice, and the result was reached for accuracy for each upcoming image with the accuracy of the neural arrangement of rice (99.9337) to potato (100 %) to tomato (99.8959), and finally for bread (83.5492), with the accuracy of the neural network 94.12 % in 1 min 2 sec. The accuracy reached in this work with Discrete Legendre Wavelets Transform (DLEGWT) and training the new network for google net was 94.12 % in 1 minute and 2 seconds with the new filter (DLEGWT). A program was created that was translated into a new algorithm. It was shown that the new filter worked in two stages, which is the initial processing stage, included analyses image with removed the noise from the image, after which the image is compressed to reduce the space used by the image data. As for the second stage, the deep learning stage takes place with the new filter, so that a new convolutional neural network is created after Dividing the image into three channels RGB to complete the convolution process for each channel, so that the hidden layers are formed until it reaches full connectivity, so that the classification process is performed with triple accuracy. The proposed algorithm has proven its efficiency
- 19 July 2024
- Published by OU Scientific Route in EUREKA: Physics and Engineering
- No. 4,p. 184-192
Open AccessClassification of food items in trash through images using a hybrid deep learning system
The issue that arouses concern is the issue of waste in general, especially waste in food. Deep learning has made its way into the aforementioned field. Pre-processing is done using a new filter derived from discrete wavelets that were derived from Legendre polynomials (LEGP) to achieve Multi Resolution Analyses (MRA), in which the image parameters are separated into proximity parameters and detail parameters, where the first is concentrated in LL, while the second is in HL, LH, HH to improve the input image of the waste in terms of noise removal and image compression. To begin the deep learning process, configure the New Convolutional Neural Network (NCNN) to configure 144 layer with the image size (224x224x3) with Google Net so that the waste resulting from leftover food is sorted. In this work, four foods were sorted: tomatoes, potatoes, bread, and rice, and the result was reached for accuracy for each upcoming image with the accuracy of the neural arrangement of rice (99.9337) to potato (100 %) to tomato (99.8959), and finally for bread (83.5492), with the accuracy of the neural network 94.12 % in 1 min 2 sec. The accuracy reached in this work with Discrete Legendre Wavelets Transform (DLEGWT) and training the new network for google net was 94.12 % in 1 minute and 2 seconds with the new filter (DLEGWT). A program was created that was translated into a new algorithm. It was shown that the new filter worked in two stages, which is the initial processing stage, included analyses image with removed the noise from the image, after which the image is compressed to reduce the space used by the image data. As for the second stage, the deep learning stage takes place with the new filter, so that a new convolutional neural network is created after Dividing the image into three channels RGB to complete the convolution process for each channel, so that the hidden layers are formed until it reaches full connectivity, so that the classification process is performed with triple accuracy. The proposed algorithm has proven its efficiency
- Open Access
- 29 June 2024
- Published by Kyiv National University of Construction and Architecture in Pidvodni tehnologii
- Vol. 1 (14) , 13-17
Developing the regression network generated from the classification network with deep learning
The field of image processing is very important in deep learning. One of the important applications in deep learning is the classification process after storing images in a data set, which requires training a deep neural network and a convolutional neural network (CNN) in MATLAB, because of the large amount of data carried by the color image, which leads to the need for deep learning technology, which leads to Discovering the number technology to create a matrix that includes the numbers from 0 to 9 in the 10 * 10 matrix, which leads to expanding the image store of the data set. In this work, a fast algorithm was created to create a large number matrix to store images, and the results that were achieved proved the efficiency of the algorithm.
- 29 June 2024
- Published by Kyiv National University of Construction and Architecture in Pidvodni tehnologii
- Vol. 1 (14) , 13-17
Open AccessDeveloping the regression network generated from the classification network with deep learning
The field of image processing is very important in deep learning. One of the important applications in deep learning is the classification process after storing images in a data set, which requires training a deep neural network and a convolutional neural network (CNN) in MATLAB, because of the large amount of data carried by the color image, which leads to the need for deep learning technology, which leads to Discovering the number technology to create a matrix that includes the numbers from 0 to 9 in the 10 * 10 matrix, which leads to expanding the image store of the data set. In this work, a fast algorithm was created to create a large number matrix to store images, and the results that were achieved proved the efficiency of the algorithm.