Search

Deep neural networks have demonstrated outstanding performances in agriculture production. Agriculture production is one of the most important sectors because it has a direct impact on the economy and social life of any society. Plant disease identification is a big challenge for agriculture production, for which we need a fast and accurate technique to identify plant disease. With the recent advancement in deep learning, we can develop a robust and accurate system. This research investigated the use of deep learning for accurate and fast tomato plant disease identification. In this research, we have used individual and merged datasets of tomato plants with 10 diseases (including healthy plants). The main aim of this work is to check the accuracy of the existing convolutional neural network models such as Visual Geometry Group, Residual Net, and DenseNet on tomato plant disease detection and then design a custom deep neural network model to give the best accuracy in case of the tomato plant. We have trained and tested our models with datasets containing over 18,000 and 25,000 images with 10 classes. We achieved over 99% accuracy with our custom model. This high accuracy was achieved with less training time and lower computational cost compared to other CNNs. This research demonstrates the potential of deep learning for efficient and accurate tomato plant disease detection, which can benefit farmers and contribute to improved agricultural production. The custom model’s efficient performance makes it promising for practical implementation in real-world agricultural settings.

Background: microwave imaging (MWI) has emerged as a promising modality for breast cancer screening, offering cost-effective, rapid, safe and comfortable exams. However, the practical application of MWI for tumor detection and localization is hampered by its inherent low resolution and low detection capability. Methods: this study aims to generate an accurate tumor probability map directly from the scattering matrix. This direct conversion makes the probability map independent of specific image formation techniques and thus potentially complementary to any image formation technique. An approach based on a convolutional neural network (CNN) is used to convert the scattering matrix into a tumor probability map. The proposed deep learning model is trained using a large realistic numerical dataset of two-dimensional (2D) breast slices. The performance of the model is assessed through visual inspection and quantitative measures to assess the predictive quality at various levels of detail. Results: the results demonstrate a remarkably high accuracy (0.9995) in classifying profiles as healthy or diseased, and exhibit the model’s ability to accurately locate the core of a single tumor (within 0.9 cm for most cases). Conclusion: overall, this research demonstrates that an approach based on neural networks (NN) for direct conversion from scattering matrices to tumor probability maps holds promise in advancing state-of-the-art tumor detection algorithms in the MWI domain.

Multilook approaches have been applied in synthetic aperture radar (SAR) tomography (TomoSAR), for improving the density and regularity of persistent scatterers reconstructed from multipass SAR images in both rural and urban regions. Multilook operations assume that all scatterers in a given neighborhood are similar in height, thereby providing additional data for recovering the position and reflectivity of a single scatterer, so that a higher signal-to-noise ratio can be achieved. This is equivalent to assuming that scatterers belonging to a local neighborhood of range–azimuth cells are located on horizontal planes. The present article generalizes this approach by adopting the so-called local plane (LP) model for TomoSAR imaging in urban areas, accounting for local variations in the height of scatterers that are not negligible. Furthermore, an LP-generalized likelihood ratio test (LP-GLRT) algorithm is developed to implement the previous idea. Compared with the multilook generalized likelihood ratio test algorithm, LP-GLRT shows better performance in the case of urban structures and terrains in experiments based on both simulated data and TerraSAR-X images.

In this paper, a deep learning technique for tumor detection in a microwave tomography framework is proposed. Providing an easy and effective imaging technique for breast cancer detection is one of the main focuses for biomedical researchers. Recently, microwave tomography gained a great attention due to its ability to reconstruct the electric properties maps of the inner breast tissues, exploiting nonionizing radiations. A major drawback of tomographic approaches is related to the inversion algorithms, since the problem at hand is nonlinear and ill-posed. In recent decades, numerous studies focused on image reconstruction techniques, in same cases exploiting deep learning. In this study, deep learning is exploited to provide information about the presence of tumors based on tomographic measures. The proposed approach has been tested with a simulated database showing interesting performances, in particular for scenarios where the tumor mass is particularly small. In these cases, conventional reconstruction techniques fail in identifying the presence of suspicious tissues, while our approach correctly identifies these profiles as potentially pathological. Therefore, the proposed method can be exploited for early diagnosis purposes, where the mass to be detected can be particularly small.

Ground-penetrating radar (GPR) technology for underground exploration consists of the transmission of an electromagnetic signal in the ground for sensing the presence of buried objects. While monostatic or bistatic configurations are usually adopted, a limited number of multistatic GPR systems have been proposed in the scientific literature. In this article, we investigate the recovery performance of a specific and unconventional contactless multistatic GPR system, designed at the Georgia Institute of Technology for the subsurface imaging of antitank and antipersonnel plastic mines. In particular, for the first time, tomographic approaches are tested against this experimental multistatic GPR system, while most GPR processing in the scientific literature processes multimonostatic experimental data sets. First, by mimicking the system at hand, an accurate theoretical as well as numerical analysis is performed in order to estimate the data information content and the performance achievable. Two different tomographic linear approaches are adopted, i.e., the linear sampling method and the Born approximation (BA) method, this latter enhanced by means of the compressive sensing (CS) theoretical framework. Then, the experimental data provided by the Georgia Institute of Technology are processed by means of a multifrequency CS- and BA-based method, thus generating very accurate 3D maps of the investigated underground scenario.

The topic of denoising magnetic resonance (MR) images is considered in this paper. More in detail, an enhanced Non-Local Means (NLM) filter using the Kolmogorov-Smirnov (KS) distance is proposed. The KS-NLM approach estimates the similarity between image patches by computing the KS distance. To overcome that NLM filters assign the same role to all pixels in patches, that is, not privileging the central one, we propose a new filter, namely the Anisotropic Weighted KS-NLM (Aw KS-NLM), which better deals with central pixels within the patches by, on one hand, including a suitable weighted strategy and, on the other, by performing a local anisotropy analysis. The Aw KS-NLM has been compared to other existing non-local Means (NLM) methodologies in both MRI simulated and real datasets. The results provide excellent noise reduction and image-detail preservation.

Hand gestures represent a natural way to express concepts and emotions which are peculiar to each culture. Several studies exploit biometric traits, such as fingerprint, iris or face for subject identification purposes. Within this paper, a novel ultrasound system for person identification that exploits hand gestures is presented. The system works as a sonar, measuring the ultrasonic pressure waves scattered by the subject’s hand, and analysing its Doppler information. Further, several transformations for obtaining time/frequency representations of the acquired signal are computed and a deep learning detector is implemented. The proposed system is cheap, reliable, contactless and can be easily integrated with other personal identification approaches allowing different security levels. The performances are evaluated via experimental tests carried out on a group of 25 volunteers. Results are encouraging, showing the promising potential of the system.

(1) Background: In this paper, an artificial neural network approach for effective and real-time quantitative microwave breast imaging is proposed. It proposes some numerical analyses for the optimization of the network architecture and the improvement of recovery performance and processing time in the microwave breast imaging framework, which represents a fundamental preliminary step for future diagnostic applications. (2) Methods: The methodological analysis of the proposed approach is based on two main aspects: firstly, the definition and generation of a proper database adopted for the training of the neural networks and, secondly, the design and analysis of different neural network architectures. (3) Results: The methodology was tested in noisy numerical scenarios with different values of SNR showing good robustness against noise. The results seem very promising in comparison with conventional nonlinear inverse scattering approaches from a qualitative as well as a quantitative point of view. (4) Conclusion: The use of quantitative microwave imaging and neural networks can represent a valid alternative to (or completion of) modern conventional medical imaging techniques since it is cheaper, safer, fast, and quantitative, thus suitable to assist medical decisions.

The advancement of new promising techniques in the field of biomedical imaging has always been paramount for the research community. Recently, ultrasound tomography has proved to be a good candidate for non-invasive and safe diagnostics. In particular, breast cancer imaging may benefit from this approach, as frequent screening and early diagnosis require decreased system size and costs compared to conventional imaging techniques. Furthermore, a major advantage of these approaches consists in the operator-independent feature, which is very desirable compared to conventional hand-held ultrasound imaging. In this framework, the authors present some imaging results on an experimental campaign acquired via an in-house ultrasound tomographic system designed and built at the University of Naples Parthenope. Imaging performance was evaluated via different tests, showing good potentiality in structural information retrieval. This study represents a first proof of concept in order to validate the system and to consider further realistic cases in near future applications.

This paper studies inverse synthetic aperture radar (ISAR) image matching and three-dimensional (3D) scattering imaging based on extracted dominant scatterers. In the condition of a long baseline between two radars, it is easy for obvious rotation, scale, distortion, and shift to occur between two-dimensional (2D) radar images. These problems lead to the difficulty of radar-image matching, which cannot be resolved by motion compensation and cross-correlation. What is more, due to the anisotropy, existing image-matching algorithms, such as scale invariant feature transform (SIFT), do not adapt to ISAR images very well. In addition, the angle between the target rotation axis and the radar line of sight (LOS) cannot be neglected. If so, the calibration result will be smaller than the real projection size. Furthermore, this angle cannot be estimated by monostatic radar. Therefore, instead of matching image by image, this paper proposes a novel ISAR imaging matching and 3D imaging based on extracted scatterers to deal with these issues. First, taking advantage of ISAR image sparsity, radar images are converted into scattering point sets. Then, a coarse scatterer matching based on the random sampling consistency algorithm (RANSAC) is performed. The scatterer height and accurate affine transformation parameters are estimated iteratively. Based on matched scatterers, information such as the angle and 3D image can be obtained. Finally, experiments based on the electromagnetic simulation software CADFEKO have been conducted to demonstrate the effectiveness of the proposed algorithm.

The problem of determining and understanding the nature of buried objects by means of nondestructive and noninvasive techniques represents an interesting issue for a great variety of applications. In this framework, the theory of electromagnetic inverse scattering problems can help in such an issue by starting from the measures of the scattered field collected on a surface. What will be presented in this communication is a two-dimensional (2-D) technique based on the so-called Born approximation (BA) combined with a compressive sensing (CS) approach, in order to improve reconstruction capabilities for a proper class of targets. The use of a multiview-multistatic configuration will be employed together with a multifrequency approach to overcome the limited amount of data due to the single-frequency technique. Therefore, after a first numerical analysis of the performance of the considered algorithm, some numerical examples for 2-D aspect-limited configurations will be presented. The scenario is composed of a simplified scene, which consists of two half-spaces, and with the probes located close to the interface between the two media. As proposed in the following, it is easy to observe that the use of CS for this kind of problems may improve reconstruction capabilities, confirming the validity of the presented approach.

Many pathologies can be identified by evaluating differences raised in the physical parameters of involved tissues. In a Magnetic Resonance Imaging (MRI) framework, spin-lattice T1 and spin-spin T2 relaxation time parameters play a major role in such an identification. In this manuscript, a theoretical study related to the evaluation of the achievable performances in the estimation of relaxation times in MRI is proposed. After a discussion about the considered acquisition model, an analysis on the ideal imaging acquisition parameters in the case of spin echo sequences, i.e., echo and repetition times, is conducted. In particular, the aim of the manuscript consists in providing an empirical rule for optimal imaging parameter identification with respect to the tissues under investigation. Theoretical results are validated on different datasets in order to show the effectiveness of the presented study and of the proposed methodology.

Speckle noise is presented as an inherent dilemma that affects the image processing field, and in particular synthetic aperture radar images. In order to mitigate the adverse effects caused by this phenomenon, several approaches have been introduced in the scientific community during the last three decades including spatial-based and non-local filtering approaches. However, these proposed techniques suffer from some limitations. In fact, it is very difficult to find an approach that is able, on the one hand, to perform well in terms of noise reduction and image detail preservation and, on the other hand, provide a filtering output solution without high computational complexity and within a short processing time. In this paper, we aim to evaluate the performance of a newly-developed despeckling algorithm, presented as an enhancement of the classical Wiener filter and properly designed to work with a Graphics Processing Unit (GPU). The algorithm is tested on both a simulated framework and real Sentinel-1 SAR data. The results, obtained in comparison with other filters, are interesting and promising. Indeed, the proposed method turns out to be a useful filtering instrument in the case of large images by performing the processing within a limited time and ensuring good speckle noise reduction with a considerable image detail preservation.

Magnetic Resonance (MR) imaging techniques are used to measure biophysical properties of tissues. As clinical diagnoses are mainly based on the evaluation of contrast in MR images, relaxation times assume a fundamental role providing a major source of contrast. Moreover, they can give useful information in cancer diagnostic. In this paper we present a statistical technique to estimate relaxation times exploiting complex-valued MR images. Working in the complex domain instead of the amplitude one allows us to consider the data bivariate Gaussian distributed, and thus to implement a simple Least Square (LS) estimator on the available complex data. The proposed estimator results to be simple, accurate and unbiased.

Today there is a growing interest in automotive sensor monitoring systems. One of the main challenges is to make them an effective and valuable aid in dangerous situations, improving transportation safety. The main limitation of visual aid systems is that they do not produce accurate results in critical visibility conditions, such as in presence of rain, fog or smoke. Radar systems can greatly help in overcoming such limitations. In particular, imaging radar is gaining interest in the framework of Driver Assistance Systems (DAS). In this manuscript, a new methodology able to reconstruct the 3D imaged scene and to detect the presence of multiple targets within each line of sight is proposed. The technique is based on the use of Compressive Sensing (CS) theory and produces the estimation of multiple targets for each line of sight, their range distance and their reflectivities. Moreover, a fast approach for 2D focus based on the FFT algorithm is proposed. After the description of the proposed methodology, different simulated case studies are reported in order to evaluate the performances of the proposed approach.