The remarkable growth of machine learning has shown that it can perform at an expert level in a number of difficult tasks, such as medical decision-making and imageprocessing. The goal of this research is to take adv...
详细信息
High-performance line-scan cameras are highly accurate and efficient for tunnel detection, however, the exponential growth of image data creates new challenges for real-time imageprocessing. The inference speed of ex...
详细信息
High-performance line-scan cameras are highly accurate and efficient for tunnel detection, however, the exponential growth of image data creates new challenges for real-time imageprocessing. The inference speed of existing tunnel detection algorithms is insufficient for massive amounts of image data, and traditional data structures limit the parallelism of the detection system, thereby reducing the speed of imageprocessing. In this study, a deep learning (DL) network is developed for real-time tunnel defect detection based on the you only look once (YOLO) family. The network includes innovative defect feature-extraction modules, optimization of the network architecture, and improvement of the decoupling heads to increase defect detection accuracy and efficiency. Furthermore, we developed critical technologies for real-time detection. First, our method employs a deployment method for DL networks based on TensorRT, which enables our network to be applied directly to C++ image acquisition programs away from the traditional DL frameworks. The proposed line-scan data structure enables the raw images to be fed into the neural network in batches, which, combined with the simultaneous technique of image acquisition, detection, and storage, significantly increases the efficiency of realtime detection. The paper demonstrates the advantages of the proposed model and hardware system through an experiment study. The test results indicate that the mean average precision and F1-score of our network were 86.07% and 84.53%, respectively, while it had 21.42 M parameters. Combined with the proposed storage strategy, a recall rate of more than 90% was achieved for defect detection. Moreover, batch imageprocessing.method has a significant advantage in terms of inference speed, with 243.9 FPS in PyTorch, 381.46 FPS in TensorRT, and 305.3 FPS in real-time detection, which is five times faster than existing defect-detection models. This study significantly increases the efficiency of tu
Traditional methods for detecting plant diseases and pests are time-consuming, labor-intensive, and require specialized skills and resources, making them insufficient to meet the demands of modern agricultural develop...
详细信息
Traditional methods for detecting plant diseases and pests are time-consuming, labor-intensive, and require specialized skills and resources, making them insufficient to meet the demands of modern agricultural development. To address these challenges, deep learning technologies have emerged as a promising solution for the accurate and timely identification of plant diseases and pests, thereby reducing crop losses and optimizing agricultural resource allocation. By leveraging its advantages in imageprocessing. deep learning technology has significantly enhanced the accuracy of plant disease and pest detection and identification. This review provides a comprehensive overview of recent advancements in applying deep learning algorithms to plant disease and pest detection. It begins by outlining the limitations of traditional methods in this domain, followed by a systematic discussion of the latest developments in applying various deep learning techniques-including image classification, object detection, semantic segmentation, and change detection-to plant disease and pest identification. Additionally, this study highlights the role of large-scale pre-trained models and transfer learning in improving detection accuracy and scalability across diverse crop types and environmental conditions. Key challenges, such as enhancing model generalization, addressing small lesion detection, and ensuring the availability of high-quality, diverse training datasets, are critically examined. Emerging opportunities for optimizing pest and disease monitoring through advanced algorithms are also emphasized. Deep learning technology, with its powerful capabilities in data processing.and pattern recognition, has become a pivotal tool for promoting sustainable agricultural practices, enhancing productivity, and advancing precision agriculture.
This paper presents an integrated image processor architecture designed for real-time interfacing and processing.of high-resolution thermal video obtained from an uncooled infrared focal plane array (IRFPA) utilizing ...
详细信息
ISBN:
(纸本)9798350344196
This paper presents an integrated image processor architecture designed for real-time interfacing and processing.of high-resolution thermal video obtained from an uncooled infrared focal plane array (IRFPA) utilizing a modern system-on-chip field-programmable gate array (SoC FPGA). Our processor provides a one-chip solution for incorporating non-uniformity correction (NUC) algorithms and contrast enhancement methods (CEM) to be performed seamlessly. We have employed NUC algorithms that utilize multiple coefficients to ensure robust image quality, free from ghosting effects and blurring. These algorithms include polynomial modeling-based thermal drift compensation (TDC), two-point correction (TPC), and run-time discrete flat field correction (FFC). To address the memory bottlenecks originating from the parallel execution of NUC algorithms in real-time, we designed accelerators and parallel caching modules for pixel-wise algorithms based on a multi-parameter polynomial expression. Furthermore, we designed a specialized accelerator architecture to minimize the interrupted time for run-time FFC. The implementation on the XC7Z020CLG400 SoC FPGA with the QuantumRed VR thermal module demonstrates that our imageprocessing.module achieves a throughput of 60 frames per second (FPS) when processing.14-bit 640x480 resolution infrared video acquired from an uncooled IRFPA.
Advancements in deep learning algorithms and high-performance GPU computation have spurred extensive research in imageprocessing. This study focuses on the synthesis of super-resolution images, a technique aimed at g...
详细信息
Advancements in deep learning algorithms and high-performance GPU computation have spurred extensive research in imageprocessing. This study focuses on the synthesis of super-resolution images, a technique aimed at generating high-resolution images from low-resolution images that contain various types of degradation and noise. In recent years, the emergence of convolutional neural networks and adversarial generative networks has made it possible to generate higher quality images. In this research, we propose two approaches to enhance the image quality based on a mainstream algorithm ESRGAN, a GAN-based generator and discriminator model architecture. The first approach applies two types of channel attention (SENet and ECA-Net) to the generator. SENet explicitly models the interdependence of convolutional features across channels, thereby enhancing the quality of the representation generated by the network. ECA-Net decomposes the channel attention of SENet to reduce model complexity while preserving performance through appropriate cross-channel interactions. The second approach applies LPIPS to the image evaluation methods of the discriminator. LPIPS serving as an image quality assessment metric, enhances perceptual evaluation by combining feature extraction from a pre-trained neural network using human evaluation data. To assess the effectiveness of our proposed methods, we employed four benchmark datasets for synthesizing super-resolution images. We used two image quality evaluation metrics: NIQE, which evaluates the naturalness of images, and LPIPS, which provides a human-like perceptual evaluation results. Experimental results demonstrate a significant enhancement in image naturalness and perceptual evaluation values compared to previous studies highlighting the effectiveness of the proposed methods.
With the rapid development of information technology, computer technology has been widely used in many industries;industry upgrading speed is getting faster and faster, market competition is becoming more and more fie...
详细信息
Cloud-based Healthcare 4.0 systems have research challenges with secure medical data processing. especially biomedical imageprocessing.with privacy protection. Medical records are generally text/numerical or multimed...
详细信息
Cloud-based Healthcare 4.0 systems have research challenges with secure medical data processing. especially biomedical imageprocessing.with privacy protection. Medical records are generally text/numerical or multimedia. Multimedia data includes X-ray scans, Computed Tomography (CT) scans, Magnetic Resonance Imaging (MRI) scans, etc. Transferring biomedical multimedia data to medical authorities raises various security concerns. This paper proposes a one-of-a-kind blockchain-based secure biomedical imageprocessing.system that maintains anonymity. The integrated Healthcare 4.0 assisted multimedia imageprocessing.architecture includes an edge layer, fog computing layer, cloud storage layer, and blockchain layer. The edge layer collects and sends periodic medical information from the patient to the higher layer. The multimedia data from the edge layer is securely preserved in blockchain-assisted cloud storage through fog nodes using lightweight cryptography. Medical users then safely search such data for medical treatment or monitoring. Lightweight cryptographic procedures are proposed by employing Elliptic Curve Cryptography (ECC) with Elliptic Curve Diffie-Hellman (ECDH) and Elliptic Curve digital Signature (ECDS) algorithm to secure biomedical imageprocessing.while maintaining privacy (ECDSA). The proposed technique is experimented with using publically available chest X-ray and CT images. The experimental results revealed that the proposed model shows higher computational efficiency (encryption and decryption time), Peak to Signal Noise Ratio (PSNR), and Meas Square Error (MSE).
image fusion is a long-established and well-known study area of digitalimageprocessing. The reason is its substantial approach in several practical applications in which multi-focus image fusion (MFIF) is one of the...
详细信息
image fusion is a long-established and well-known study area of digitalimageprocessing. The reason is its substantial approach in several practical applications in which multi-focus image fusion (MFIF) is one of the most essential and commonly employed applications in the current situation. However, low contrast, colour distortion, and different fusion losses are significant challenges that must be addressed while generating the composite image. So, this study proposes an experimental and comprehensive review with an idea/methodology to encounter these challenges in which pre-hand enhancement criteria are applied for both gray-scale and colour images before fusion for multi-focus images. First, a detailed analysis of many categories and their sub-categories is offered, providing the groundwork for this study. Following this, we explain the enhancement methodology, which utilises the histogram for gray-scale and colour balancing for colour images. Besides this, several non-reference objective evaluation fusion metrics are also presented. In addition, simulation using MATLAB software from the conventional approach to the recent deep learning approach is carried out with and without enhancement for twenty state-of-the-art MFIF algorithms to check the validity of the adopted criteria. The results obtained with the enhancement approach as a pre-processing.step show that almost all the outcomes are better or comparable with the original methodologies in both an objective and subjective manner. Moreover, the additional running time of just 0.2582 s at pre-processing.stage indicates little computational complexity. Lastly, we discuss the current status of research, ongoing difficulties, and future possibilities.
In this paper we deal with the issue of digital camera identification (DCI) based on images. This area matches the digital forensics (DF) research. This topic has attracted many researchers and number of algorithms fo...
详细信息
In this paper we deal with the issue of digital camera identification (DCI) based on images. This area matches the digital forensics (DF) research. This topic has attracted many researchers and number of algorithms for DCI have been proposed. However, majority of them focus only on camera identification with high accuracy without taking into account the speed of imageprocessing. In this paper we propose an effective algorithm for much faster camera identification than state-of-the-art algorithms. Experimental evaluation conducted on two large image datasets including almost 14.000 images confirms that the proposed algorithm achieves high classification accuracy of 97 [%] in much shorter time compared with state-of-the-art algorithms which obtained 92.0 - 96.0 [%]. We also perform a statistical analysis of obtained results which confirms their reliability.
Computer vision can be defined as a converter used by artificial systems to understand the information captured in digitalimages and videos sequences. Video sequences are digitalimages disposed at certain frames per...
详细信息
ISBN:
(纸本)9789819735556;9789819735563
Computer vision can be defined as a converter used by artificial systems to understand the information captured in digitalimages and videos sequences. Video sequences are digitalimages disposed at certain frames per second. To achieve excellent performance in the process of computer vision, it is necessary to compile a large amount of data which can be time-consuming and/or expensive. To alleviate such issues which consist in insufficient digitalimages to be processed during deep learning, different augmentation algorithms were developed, being classified in three types of categories: model-based, optimizing policy-based and model-free. Different approaches were carried out in the matter of images augmentation, but most did not consider a solution that would also reduce the dataset size. In this current approach, we evaluate the impact of mathematical morphology operators (MMO) on original digitalimages to achieve synthesized images using a model-based augmentation algorithm. Deep learning (DL) algorithms consisting in two trained deep convolutional neural networks (CNN) were involved in imageprocessing. Using one CNN trained on original images dataset and the second CNN trained on augmentation images dataset with MMO, relevant results will be provided and will allow valid comparison.
暂无评论