Deep Learning, refers to large set of neural network based algorithms, have emerged as promising machine learning tools in the general imaging and computer vision domains. Convolutional neural networks (CNNs), a speci...
详细信息
ISBN:
(纸本)9781510600201
Deep Learning, refers to large set of neural network based algorithms, have emerged as promising machine learning tools in the general imaging and computer vision domains. Convolutional neural networks (CNNs), a specific class of deep learning algorithms, have been extremely effective in object recognition and localization in natural images. A characteristic feature of CNNs, is the use of a locally connected multi layer topology that is inspired by the animal visual cortex (the most powerful vision system in existence). While CNNs, perform admirably in object identification and localization tasks, typically require training on extremely large datasets. Unfortunately, in medical image analysis, large datasets are either unavailable or are extremely expensive to obtain. Further, the primary tasks in medical imaging are organ identification and segmentation from 3D scans, which are different from the standard computer vision tasks of object recognition. Thus, in order to translate the advantages of deep learning to medical image analysis, there is a need to develop deep network topologies and training methodologies, that are geared towards medical imaging related tasks and can work in a setting where dataset sizes are relatively small. In this paper, we present a technique for stacked supervised training of deep feed forward neural networks for segmenting organs from medical scans. Each 'neural network layer' in the stack is trained to identify a sub region of the original image, that contains the organ of interest. By layering several such stacks together a very deep neural network is constructed. Such a network can be used to identify extremely small regions of interest in extremely large images, inspite of a lack of clear contrast in the signal or easily identifiable shape characteristics. What is even more intriguing is that the network stack achieves accurate segmentation even when it is trained on a single image with manually labelled ground truth. We validate th
Unmanned Aircraft systems (UAS) are being used commonly for video surveillance, providing valuable video data and reducing the risks associated with human operators. Thanks to its benefits, the UAS traffic is nearly d...
详细信息
ISBN:
(纸本)9781467393355
Unmanned Aircraft systems (UAS) are being used commonly for video surveillance, providing valuable video data and reducing the risks associated with human operators. Thanks to its benefits, the UAS traffic is nearly doubling every year. However, the risks associated with the UAS are also growing. According to the FAA, the volume of air traffic will grow steadily, doubling in the next 20 years. Paired with the exponential growth of the UAS traffic, the risk of collision is also growing as well as privacy concerns. An effective UAS detection and/or tracking method is critically needed for air traffic safety. This research is aimed at developing a system that can identify/detect a UAS, which will subsequently enable counter measures against UAS. The proposed system will identify a UAS through various methods including imageprocessing and mechanical tracking. Once a UAS is detected, a countermeasure can be employed along with the tracking system. In this research, we describe the design, algorithms, and implementation details of the system as well as some performance aspects. The proposed system will help keep the malicious or harmful UAS away from the restricted or residential areas.
随着电网系统对安全性的要求不断提高,机器人等自动化设备越来越多地应用到电力巡检中,人工手段对设备采集的图像进行缺陷检测存在效率低、检测结果不稳定的缺点。为此提出并开发了一种输电线路巡检图像智能诊断系统。在系统中构建了分层软件结构,基于Visual Studio 2010开发环境,开发了图像导入、数据库访问、文本输出等功能模块,使用了多种检测算法,且利用ADO技术以实现对数据库的访问和修改。软件测试的结果表明该系统具有工作效率高、错误率低、界面友好等优点,适用于输电线路巡检图像的检测工作。
DR images differ significantly in appearance from traditional screen film radiographs, largely due to the sophisticated imageprocessingalgorithms applied at the ‘For processing’ stage. However the IEC protocol (IE...
DR images differ significantly in appearance from traditional screen film radiographs, largely due to the sophisticated imageprocessingalgorithms applied at the ‘For processing’ stage. However the IEC protocol (IEC, 2003) [1] for measuring the pre-sampled MTF of DR systems is based on the ‘For-processing’ version of the final image, which tests only the hardware. Crucially any software effects on image quality are ignored. MTF Curves were calculated on a range of ‘For Presentation’ images and were compared with the corresponding IEC mandated MTF curves from ‘For processing’ images. ‘For Presentation’ MTF curves differ not only between clinical protocols, but also between manufacturers. In each case the high frequency response was greater in ‘For-Presentation’ images, indicating the effectiveness of the imageprocessingalgorithms. The absence of any guidance or standards on appropriate levels of imageprocessing for clinical protocols is noted. This is necessary as the imageprocessing applied significantly affects the system response and, therefore, the appearance of the ‘For Presentation’ images reported by radiologists. Further investigation is merited, perhaps at the European level. A new set of standards should be considered, which take into account the software effects on image quality; these could be investigated as part of a revised RP 162.
As the detection rate of headlights under complex lighting scenes in nighttime is relatively low, a novel algorithm of headlights detection based on atmospheric reflection-scattering model has been presented. The diff...
详细信息
The article deals with imageprocessingalgorithms in the analysis plane of an angle measuring two-axis autocollimator, which uses a reflector in the form of a quadrangular pyramid. This algorithm uses Hough transform...
The article deals with imageprocessingalgorithms in the analysis plane of an angle measuring two-axis autocollimator, which uses a reflector in the form of a quadrangular pyramid. This algorithm uses Hough transform and the method of weighted summation. The proposed algorithm can reduce the area of nonoperability and open up new possibilities for this class of devices.
暂无评论