In this paper, we propose a new approach for matching images observed in different camera views with complex cross-view transforms and apply it to person re-identification. It jointly partitions the image spaces of two into configurations according similarity transforms. The visual features an pair from are first locally aligned by being projected common feature space then matched softly assigned metrics which optimized. optimal recognizing identities those clustering They learned utilizing...

The performance of a generic pedestrian detector may drop significantly when it is applied to specific scene due the mismatch between source training set and samples from target scene. We propose new approach automatically transferring scene-specific in static video surveillance without manually labeling proposed transfer learning framework consists four steps. 1) Through exploring indegrees on visual affinity graph, are weighted match distribution samples. 2) It explores context cues select...

The performance of a generic pedestrian detector may drop significantly when it is applied to specific scene due mismatch between the source dataset used train and samples in target scene. In this paper, we investigate how automatically scene-specific starting with video surveillance without further manually labeling any under novel transfer learning framework. It tackles problem from three aspects. (1) With graphical representation through exploring indegrees samples, are properly...

ABSTRACT Rapid and accurate identification of plastic beverage bottles is great importance because can be encountered as physical evidence in cases involving assaults, thefts, homicides. In this experiment, 40 commercially available were collected experimental samples, their Raman spectral data collected. Initially, the samples classified into two categories polyethylene terephthalate (PET) (PE), 35 PET further clustered three by K‐means clustering. Savitzky–Golay algorithm smoothing,...

Person re-identification tackles the problem whether an observed person of interest reappears in a network cameras. The difficulty primarily originates from few samples per class but large amounts intra-class variations real scenarios: illumination, pose and viewpoint changes across So far, proposals literature have treated this either as matching focusing on feature representation or classification/ranking relying metric optimization. This paper presents new way called Common-Near-Neighbor...

Electroencephalogram (EEG) data are an effective indicator to evaluate driver fatigue, but it is usually disturbed by noise. The frequent head nodding, as well the time of day and total driving time, also have very close relationship with fatigue. All these factors should be taken into account for comprehensive fatigue evaluation. 50 drivers recruited take part in fatigue‐oriented experiment on simulator. Based EEG samples, EEG‐based has been established artificial neural network....

We develop a completion pipeline for fragmented and damaged skulls. The goal of this work is to convert scanned incomplete skull fragments complete model subsequent forensic or archeological tasks such as facial reconstruction. proposed assembly algorithms can also be used repair other objects with inherent symmetry. A two-step framework proposed: (1) rough by an ICP-like template matching algorithm integrated the slippage features spin-image descriptors; (2) refinement global optimization...

Automatic image annotation plays a significant role in understanding, retrieval, classification, and indexing. Today, it is becoming increasingly important order to annotate large-scale social media images from content-sharing websites networks. These are usually annotated by user-provided low-quality tags. The topic model considered as promising method describe these weak-labeling learning latent representations of training samples. recent methods based on models have two shortcomings....

We present a skull assembly and completion framework based on shape matching. In order to assemble fragmented skulls, we need compute rigid transformations from these fragments their assembled geometry. develop reliable pipeline where each fragment is matched transformed be aligned with the template. further complete several damaged regions, use template repair regions skull. The entire has been conducted real models demonstrated great robustness effectiveness.

With the popularity of mobile devices, people are now able to access services like website surfing, e-banking, e-shopping and voice-mail etc. anywhere anytime. Most require secure biometric based personal authentication. This paper aims improve accuracy authentication in environment by integrating multiple modal biometrics, i.e. face speech. A specially designed Mobile Biometry (MOBIO) database was used test performances. While facial biometrics achieved an average EER 27.3%, speech 33.4%....

This paper presents a subject centric group feature for person re-identification. Our approach is inspired by the observation that people often tend to walk alongside others or in group. We argue co-travelers' information, including geometry and visual cues, can reduce re-identification ambiguity lead better accuracy, compared approaches rely only on cues. introduce person-group capture both information of co-travelers around subject. compute dis-similarity between features solving an...

In this paper, a new face recognition method based on kernel discriminate local preserve projection(KDLPP) and Multi-feature fusion under smart environment is proposed . order to solve the small sample size problem, combined with theory QR decomposition, algorithm named projection algorithm. considered external features are useful in recognition, because hair highly variable feature of human ,so we DCT layer. The experiments AMI database indicate can enhance accuracy system effectively.

In this paper, we will present a novel face recognition algorithm using 3D geometric information of the face, which is acquired parallel stereo vision. Firstly, an effective method detecting facial feature, mainly depends on skin color in YCbCr space. We define feature points for and compute depth data each point vision principle. At same time, mesh generation based Ohtake’s clearing employed to generate structure face. Finally, use nD-PCA recognize data. A simulation was carried out...

Multiple-shot person re-identification, which is valuable for application in visual surveillance, tackles the problem of building correspondence between images same from different cameras. It challenging because large within-class variations due to changeable body appearance and environment small between-class differences arising possibly similar shape clothes style. A novel method named “Bi-level Relative Information Analysis” proposed this paper issue by treating it as a set-based ranking...

In this paper, a novel multi-pose face image recognizing method is presented. algorithm, the with can be recognized by comparing it eigenface which generated from 3D model database, view point of estimated image, estimation algorithm utilize facial geometrical feature to estimate head pose parameter applied model. Through method, keep same real captured face. Finally, PCA based employed extract exemplar database and input image. The cosine distance matching will used compare similarity...

Person re-identification (re-id) is one of the hottest research topics due to its great value in video analysis applications, such as indoor security and road surveillance. It has been verified beneficial for re-id joint global local features recent literature. However, most existing methods usually extract region or divide whole image into several parts without considering alignment different parts, which are not discriminating robust complex scenarios. In this paper, we propose a novel...

Re-identifying the same person in different images is a distinct challenge for visual surveillance systems. Building an accurate correspondence between highly variable requires suitable dissimilarity measure. To date, most existing measures have used adapted distance based on learned metric. Unfortunately, real-world human image data, which tends to show large intra-class variations and small inter-class differences, continues prevent these from achieving satisfactory re-identification...

In this paper, we introduce a spatial appearance group feature for human re-identification in the videos. The observation that motivates is co-traveler can serve as clue to reduce ambiguity of reidentification task. encodes appearances and structure, estimate probability two observed groups different cameras by combining score predicted model model, which are trained using person samples. We conduct experiments demonstrate improvements brought proposed feature. also compare our approach with...

This paper proposes a new method for improving face image resolution (i.e. super-resolution method). Our utilizes single high-resolution sample to infer the lost information in low-resolution test while all and are divided into patches as atoms by sliding windows sampling. The entire formed raw dictionary then compacted optimized K-SVD algorithm get trained dictionary. Correspondingly, we can obtain down sampling each atom any patch of be linearly represented with sparse coding method. With...