The application of visual technology to mine robots has become a hot topic in the development of coal mine automatic production. Key techniques of robot control are the feature recognition of sampled videos and the pe...
详细信息
The application of visual technology to mine robots has become a hot topic in the development of coal mine automatic production. Key techniques of robot control are the feature recognition of sampled videos and the perception of complex surroundings. However, it is difficult for features in underground images with dark hue and low target discrimination to be recognized and extracted, especially for reasons of the nonuniform illumination and heavy dust concentration in mines. Hence, an edge detection algorithm based on the Retinex theory and wavelet multiscale product is proposed in this paper for low-light-level mine image feature extraction, which employs a modified multiscale Retinex method to deal with the low frequency subplot after the wavelet decomposition, an improved fuzzy enhancement approach to handle high frequency components, and finally a revised multiscale product edge detection algorithm to obtain the ultima edge image. Compared with a variety of algorithms by detecting edges of both normal illuminated and underground images, experimental results show that with characteristics of high real-time performance and detection accuracy, the proposed algorithm can exactly meet the needs of surrounding environment perception for mine robots, which applies well to image edge detection in low illumination mines. (C) 2016 Optical Society of America
As massive open online courses (MOOCs) and online intelligent tutoring systems(ITS) have become increasingly widespread, the number of learners enrolled in online courses has shown explosive growth. However, these lea...
详细信息
Pedestrian segmentation in infrared images is a difficult problem for the defects of low SNR and inhomogeneous luminance distribution. In this paper, we propose a method which aims to obtain the accurate pedestrian se...
详细信息
ISBN:
(纸本)9781467399616
Pedestrian segmentation in infrared images is a difficult problem for the defects of low SNR and inhomogeneous luminance distribution. In this paper, we propose a method which aims to obtain the accurate pedestrian segmentation through a background prior and boundary weight-based saliency. Background likelihood is firstly calculated as background prior to get an abstract representation for infrared pedestrian. Then, by considering the object-center prior, the object-biased Gaussian model is applied to derive the probability density estimation for pedestrians. Finally, the above two results are integrated with the boundary weight to obtain the final saliency map for infrared image, based on which pedestrians can be easily segmented. Experimental results on real infrared images captured by intelligent transportation systems demonstrate the effectiveness of the proposed approach against the state-of-the-art algorithms.
Palmprint based identification has attracted much attention in the past decades. In some real-life applications, portable personal authentication systems with high accuracy and speed efficiency are required. This pape...
详细信息
Palmprint based identification has attracted much attention in the past decades. In some real-life applications, portable personal authentication systems with high accuracy and speed efficiency are required. This paper presents an embedded palmprint recognition solution based on the multispectral image modality. We first develop an effective recognition algorithm by using partial least squares regression, then a FPGA prototype is implemented and optimized through high-level synthesis technique. The evaluation experiments demonstrate that the proposed system can achieve a higher recognition rate at a lower running cost comparing to the reference implementations.
Surveillance is very essential for the safety of power substation. The detection of whether wearing safety helmets or not for perambulatory workers is the key component of overall intelligent surveillance system in po...
详细信息
ISBN:
(纸本)9781538604915
Surveillance is very essential for the safety of power substation. The detection of whether wearing safety helmets or not for perambulatory workers is the key component of overall intelligent surveillance system in power substation. In this paper, a novel and practical safety helmet detection framework based on computer vision, machine learning and imageprocessing is proposed. In order to ascertain motion objects in power substation, the ViBe background modelling algorithm is employed. Moreover, based on the result of motion objects segmentation, real-time human classification framework C4 is applied to locate pedestrian in power substation accurately and quickly. Finally, according to the result of pedestrian detection, the safety helmet wearing detection is implemented using the head location, the color space transformation and the color feature discrimination. Extensive compelling experimental results in power substation illustrate the efficiency and effectiveness of the proposed framework.
Interdisciplinary coastal observations over a two-week period in the northern Gulf of Mexico reveal a complex and dynamic bottom boundary layer (BBL) that is characterized by both biological and suspended sediment (bi...
详细信息
ISBN:
(纸本)9781510608733;9781510608740
Interdisciplinary coastal observations over a two-week period in the northern Gulf of Mexico reveal a complex and dynamic bottom boundary layer (BBL) that is characterized by both biological and suspended sediment (biogeo-) optical signals. Much of the BBL optical variance is concealed from remote sensing by the opacity of the nearly omnipresent surface river plume, however, the BBL physical dynamics and resulting optical excitation are indeed responding to surface wind stress forcing and surface gravity wave-induced turbulence. Here we present a series of numerical modeling efforts and approaches aimed towards resolving and simulating these observed biogeo- physical and -optical processes. First, we examine results from the Tactical Ocean Data System (TODS), which combines daily satellite imagery with numerical circulation model results to render a three-dimensional estimate of the optical field and then execute a reduced-order complexity advection-diffusion-reaction model to render hourly forecasts. Whereas the TODS system has the advantage of effectively assimilating both glider data and satellite images, the 3D generation algorithms still have difficulty in the northern Gulf's complex 3-layered system (surface plume, geostrophic interior, BBL). Second, we present results from the Coupled Ocean-Atmosphere Prediction (COAMPS) system that has been modified to include interactive surface-gravity wave simulations. Results from this complex numerical modeling system suggest that Stokes drift current (SDC) has a potentially major role in determining the physical and kinematic characteristics of the BBL, and will substantially impact model-based estimates of sediment resuspension and transport.
Compressive image recovery is a challenging problem that requires fast and accurate algorithms. Recently, neural networks have been applied to this problem with promising results. By exploiting massively parallel GPU ...
ISBN:
(纸本)9781510860964
Compressive image recovery is a challenging problem that requires fast and accurate algorithms. Recently, neural networks have been applied to this problem with promising results. By exploiting massively parallel GPU processing architectures and oodles of training data, they can run orders of magnitude faster than existing techniques. However, these methods are largely unprincipled black boxes that are difficult to train and often-times specific to a single measurement *** was recently demonstrated that iterative sparse-signal-recovery algorithms can be "unrolled" to form interpretable deep networks. Taking inspiration from this work, we develop a novel neural network architecture that mimics the behavior of the denoising-based approximate message passing (D-AMP) algorithm. We call this new network Learned D-AMP (LDAMP).The LDAMP network is easy to train, can be applied to a variety of different measurement matrices, and comes with a state-evolution heuristic that accurately predicts its performance. Most importantly, it outperforms the state-of-the-art BM3D-AMP and NLR-CS algorithms in terms of both accuracy and run time. At high resolutions, and when used with sensing matrices that have fast implementations, LDAMP runs over 50 x faster than BM3D-AMP and hundreds of times faster than NLR-CS.
image segmentation is an integral part of critical imageprocessing applications. Segmentation involves removal of a region of interest from the background. Recent researches in segmentation incorporate clustering alg...
详细信息
ISBN:
(纸本)9781509010660
image segmentation is an integral part of critical imageprocessing applications. Segmentation involves removal of a region of interest from the background. Recent researches in segmentation incorporate clustering algorithms for separation or removal of regions of interest. Prominent segmentation algorithms include K - means which segment the region from the background and further median filtering could be utilized to remove the unwanted regions in the segmented image. This research paper utilizes an adaptive wavelet neural network model with training or learning process optimized by the particle swarm optimization algorithm. The proposed algorithm has been tested and experimental results indicate a high precision of segmentation when compared with the conventional techniques.
Additive noise removal from a given image is an important task in digital imageprocessing for which denoising algorithms are used. The goal of any denoising algorithm is to attenuate the noise properly and to preserv...
详细信息
ISBN:
(纸本)9781509032105
Additive noise removal from a given image is an important task in digital imageprocessing for which denoising algorithms are used. The goal of any denoising algorithm is to attenuate the noise properly and to preserve the useful content of an image. Although various denoising algorithms have been proposed to remove noise but there is still scope of improvement. The main focus of this paper is, first, analyze the basic denoising approaches and to compare them, second, to study post-stage filtering technique using method noise and reweight schemes. In this case study, we observe through our experiments that the post-filtering techniques have more potential to attenuate the noise properly, which is left by the initially applied denoising approach. The denoising performance of all considered methods is compared using two parameters: PSNR and MSSIM.
The paper is focused on demonstration of image inpainting technique using the F-transform theory. Side by side with many algorithms for the image reconstruction we developed a new method of patch-based filling of an u...
详细信息
ISBN:
(纸本)9783319405964;9783319405957
The paper is focused on demonstration of image inpainting technique using the F-transform theory. Side by side with many algorithms for the image reconstruction we developed a new method of patch-based filling of an unknown (damaged) image area. The unknown area is proposed to be recursively filled by those known patches that have non-empty overlaps with the unknown area and are the closest ones among others from a database. We propose to use the closeness measure on the basis of the F-1-transform.
暂无评论