查看更多>>摘要：The cost and delivery time have become the biggest challenges in small volume perishable products logistics under E-commerce. In order to better solve this problem, an integrated multi-task operation framework is proposed from the system optimization perspective by analyzing the characteristics of perishable products. As the current scheduling models seldom consider both the benefits of individual tasks and overall multi-task simultaneously, a novel multi-objective model aiming at maximizing the mean fractional remaining shelf-life, minimizing the total logistics cost and remaining fractional shelf-life deviation of all tasks is proposed. In view of variable length path selection and resources capability conflict in the process of solving, a Non-dominated Sorting Genetic Algorithm (NSGA-Ⅱ) with adaptive variable length chromosome cross and double mutation mechanisms is designed. Finally, the validity of the model is verified by comparing different models based on a corresponding case.
查看更多>>摘要：This paper presents an optimal method for constructing bicubic Coons surfaces based upon twist vectors at rectangular grid points. These twists are determined by using a new optimal criterion derived from Euler-Lagrange PDE. By applying Euler-Lagrange PDE on every patch and make it independent of surface parameters, a new optimization problem is derived. A linear system with block tridiagonal coefficient matrix is established. The coefficient matrix is the Kronecker product of two identical tridiagonal matrices and its non-singularity is proved. At last, we analyse some numerical and graphic examples in order to show the efficiency of the presented method.
查看更多>>摘要：By the interference of light, shadow, oil pollution and noise, pavement crack segmentation is prone to appear error segmentation. In order to accurately extract the crack edge which conforms to the human eye characteristics, pavement crack detection algorithm based on visual model is proposed in this paper. This algorithm uses opening and closing operations based on multi-scale contour structuring elements to remove noise, and grade image is obtained by the smallest size contour structuring. An image can be divided into low dark area, middle area and high bright area through Weber's law according to bionics principle and background brightness. The edge thresholds for the different areas are calculated according to the different equations respectively. The proposed crack edge detector is tested on popular images having different image properties and also compared with popular edge detectors from the literature. Experimental results show that the proposed crack edge detector exhibit s much better performance than the one of competing operators, and may efficiently be used for the detection of crack in digital pavement images.
查看更多>>摘要：In this paper, we address fast fingerprint recognition and related technical problems arising in various of embedded systems with critical information security considerations nowadays. Our method adopts a truly parallel processing neuron chip to accelerate the recognition speed. We propose an effective algorithm by combining Eight Directions Gabor Filters to filter the real-life finger images and Principal Component Analysis (PCA) method to reduce the dimensions of feature vectors. Experiments show that the hardware based fingerprint system has better performance than these conventional systems in regards to accuracy and speed.
Win KongMohamad Hanif Md SaadMohd Asyraf ZulkifleyM A Hannan...
6693-6700页
查看更多>>摘要：This paper describes a method for real-time detection of human fall for video surveillance application. The proposed algorithm utilizes gait information in judging the fall incident. Gait refers to the pattern of human walking. Therefore, a fall is defined whenever there is a variation from the normal gait parameters. The detection algorithm is divided into three stages: human gait modeling, feature extraction and fall classification. Point Distribution Model (PDM) is employed to fit a skeleton model from a training set of data on the extracted human body contour. Then, the gait features are derived from the skeleton. The detection performance relies on the threshold values, which differ according to gait pattern. The results demonstrate that the gait analysis was able to detect fall incident accurately.
查看更多>>摘要：Hepatic CT (computed tomography) image is the main technique for diagnosing diseases. When the contrast agent is injected into the patients body with high-pressure syringe, blood vessels can be shown due to the density difference. However, in the pediatric diagnosis, high doses are not allowed. So great noise will appear in the CT image. Therefore, there is a strong demand for noise reduction in clinical diagnosis. We consider using TV model to remove the noise of original CT images. However, the edges of blood vessels in the liver are blur in the process of Hepatic CT images denoising which is based on two-dimensional TV model. As a result, it is not good for later extraction of blood vessels and Hepatic reconstruction. In this paper, we adopt three-dimensional TV model which fully takes the interface between the upper and lower Hepatic CT images into consideration and effectively maintains the edges of the liver and blood vessels and removes large noise. In order to simplify the minimization problem, the Split Bregman algorithm is proposed and then we use it for CT image denoising.
查看更多>>摘要：Sound source localization system based on microphone array has practical applications in the field of video conferencing, human-computer interaction, etc. In most of sound source localization system, microphone arrays are unblocked; sound wave can propagate to the microphones directly. In this paper we consider a kind of blocked microphone array-head-worn microphone array, and utilize additional spatial information generated by block effect of head-worn helmet to estimate the direction angle of source. In this article, we propose an angle estimation algorithm based on time delay compensation of diffraction path, which can estimate the time difference of arrival between head-worn microphones according to the diffraction effect of helmet on low frequency sound signals. Experiments show that for the head-worn microphone array, the sound source localization algorithm we proposed based on the diffraction path of low-frequency signal can accurately estimate the direction angle of sound source.
查看更多>>摘要：This paper introduces a Modified Eigenvector Projection approach for Subspace estimation (MEPS), which is robust and has high convergence rate. First, the KR signal subspace is exploited to estimate the covariance matrix. Then, interference plus noise subspace is constructed. And the desired signal steering vector is projected to this subspace. MEPS can estimate the sample covariance matrix more precisely without knowing the number of the signal sources. Simulation results show that MEPS can make convergence in limited snapshots and maintain a high Signal to Interference Plus Noise ratio (SINR) in a large range of input Signal to Interference Rate (SNR). Besides, without knowing the number of the signal sources, this method can estimate the sample covariance matrix more precisely.
查看更多>>摘要：Shape is one of most prominent characters of image. Shape-based image retrieval is an extremely important part of content-based image retrieval, extracting and describing shape features and then realizing retrieval of images by the similarities among shape features. For shape-based image retrieval, the extraction and description of shape features of images are significant research content, and in the paper, we made further research, expatiated on current methods of describing shape features, and improved extraction methods of characteristic parameters, giving corresponding algorithms. On that basis, we extracted a set of eigenvalue, forming eigenvector. After normalizing them, we could conduct shape recognition through BP neural network.
查看更多>>摘要：In the light of slow-speed computation and poor performance in multi-scale detection of traditional HOG feature extraction algorithm, we put forward an improved pedestrian detection algorithm based on multi-scale HOG feature and integral image. On the basis of traditional HOG algorithm, firstly we introduce integral image and multi-scale HOG feature to reduce the feature dimension and avoid some repetitive computation. Next, we use INRIA data-set to train the SVM classifier and test it based on two different algorithms. It is concluded that the improved algorithm we have put forward enable the detection efficiency improvement with better recall, computation speed and robustness.
NETL
NSTL