Statistical modeling of wavelet subbands has frequently been used for image recognition and retrieval. However, traditional wavelets are unsuitable use with images containing distributed discontinuities, such as edges. Shearlets a newly developed extension that better suited to characterization. Here, we propose novel texture classification retrieval methods model adjacent shearlet subband dependences using linear regression. For classification, two energy features represent each in order...

In this paper, an adaptive improved ant colony algorithm based on population information entropy(AIACSE) is proposed to improve the optimization ability of algorithm. The diversity in iterative process described by entropy. non-uniform distribution initial pheromone constructed reduce blindness search at starting phase. diffusion model used enhance exploration and collaboration capacity between ants. parameter adjusting strategy novel updating mechanism evolutionary characteristics are...

This paper surveys the state-of-the-art in wireless capsule endoscopy terms of commercially available products and prototypes currently under development research labs worldwide. Challenges facing us designing manufacturing active endoscopes are outlined, together with potential methods to tackle them.

The key to successful positioning of autonomous mobile robots in complicated indoor environments lies the strong anti-interference system and accurate measurements from sensors. Inertial navigation systems (INS) are widely used for because they not susceptible external interferences work properly, but errors may be accumulated over time. Thus ultra wideband (UWB) is usually adopted compensate due its high ranging precision. Unfortunately, UWB easily affected by multipath effects...

Because single local or global characteristics can only depict the classification information of an object unilaterally partially, that may result in low recognition accuracy; this paper we propose improved SURF and modified Zernike moments descriptor (ISMZMD) for recognition. Firstly, extracted seven descriptors objects. Secondly, effectively fused two features together with different weight factors based on their contribution to identification. Thirdly, computed Euclidean distance decide...

Path planning of mobile robots in complex environments is the most challenging research. A hybrid approach combining enhanced ant colony system with local optimization algorithm based on path geometric features, called EACSPGO, has been presented this study for robot planning. Firstly, simplified model pheromone diffusion, initialization strategy unequal allocation, and adaptive update mechanism have simultaneously introduced to enhance classical algorithm, thus providing a significant...

The local binary pattern (LBP) model is a simple and effective method of texture classification, but it sensitive to rotational noisy images. Although many variants LBP are proposed by scholars, there still several urgent problems, such as poor noise rotation immunity. In this paper, we propose robust descriptor, jumping refined (JRLP) for classification. particular, first extract difference count (JLDCP) consisting second-order diagonal represent the information in domain. To capture detail...

Local binary pattern (LBP) is a frequently‐used texture descriptor. Lots of LBP‐variants have been proposed to improve its performance representing textures. However, most them ignore the global and neighbour‐difference information an image texture. In this study, authors propose structural difference histogram representation by fusing segmented structure (SSP), refined LBP (RLBP) (NDP) for classification. Particularly, structure, which contains contour texture, first constructed compute SSP...

The representation-based learning methods, such as sparse classification and low-rank representation, show effective robust for image clustering classification. However, these methods essentially belong to the transductive they cannot deal with new samples. Meanwhile, original high-dimensional data contains a large amount of redundant information. If are directly performed, it will not only degrade performance algorithm but also lead sharp increase in computation. Therefore, novel preserving...

Linear discriminant analysis (LDA) is a linear feature extraction approach, and it has received much attention. On the basis of LDA, researchers have done lot research work on it, many variant versions LDA were proposed. However, inherent problem cannot be solved very well by methods. The major disadvantages classical are as follows. First, sensitive to outliers noises. Second, only global structure preserved, while local information ignored. In this paper, we present new orthogonal sparse...

HU moments aren't invariant for scaling in the discrete state, so they are improved paper. The consistent with region, boundary and situation. Therefore, applied to three-dimensional object recognition. Firstly calculated. Then similarity measure is computed between objects be recognized standard one. Finally experiments simulated by MATLAB, experimental results demonstrate that translation, rotating of objects, recognition rate relatively high proposed algorithm has some practical value. So...

Abstract Considering the defect of object recognition with single global feature or local feature, in this letter authors propose a shape interest points descriptor (SIPD) for recognition. Particularly, extract features by improved HU moments and then Speeded up Robust Feature (SURF). Object is carried out similarity measure. Because influence SURF different, two different measures are fused effectively using weight factors. Experimental results show that author’ proposed method effective...

Multi-view feature learning aims at improving the performances of tasks, by fusing various kinds features (views), such as heterogeneous and/or homogeneous features. Current leading multi-view approaches usually learn in each view separately while not uncovering shared information from multiple views. In this paper, we propose a framework, which can simultaneously separate subspace for and all views, respectively; specifically, preserve particular within view, meanwhile, capture correlation...

Nonnegative matrix factorization, as a classical part-based representation method, has been widely used in pattern recognition, data mining and other fields. However, the traditional nonnegative factorization directly factoring decomposes original data, often contains lot of redundancy noise, which seriously affect subsequent processing data. In this work, we propose an adaptive graph regularization discriminant (AGDNMF) for image clustering. The AGDNMF algorithm makes full use local...