Search

The bag-of-words (BoW) has been widely regarded as the most successful algorithms for content-based image related tasks, such as large scale image retrieval, classification, and object categorization. Large visual words acquired by BoW quantization through large vocabulary or codebooks have been receiving much attention in the past years. However, not only construction of large vocabulary but also the quantization process impose a heavy burden in terms of time and memory complexities. In order to tackle this issue, we propose an efficient hierarchical BoW (HBoW) to achieve large visual words through quantization by a compact vocabulary instead of large vocabulary. Our vocabulary is very compact since it is only composed of two small dictionaries which is learned through segmental sparse decomposition of local features. To generate the BoW with large size, we first divide the local features into two half parts, and use the two small dictionaries to compute their sparse codes. Then, we map the two indices of the maximum elements of the two sparse codes to a large set of visual words based upon the fact that data with similar properties will share the same base weighted with the largest sparse coefficient. To further make similar patches have higher probability of select the same dictionary base to get similar BoW vectors, we propose a novel collaborative dictionary learning method by imposing the similarity regularization factor together with the row sparsity regularization across data instances during group sparse coding. Additionally, based on index combination of top-2 large sparse codes of local descriptors, we propose a soft BoW assignment method so that our proposed HBoW can tolerate different word selection for similar patches. By employing the inverted file structure built through our HBoW, K-nearest neighbors (KNN) can be efficiently retrieved. After incorporation of our fast KNN search into the SVM-KNN classification method, our HBoW can be used for efficient image classification and logo recognition. Experiments on serval well-known datasets show that our approach is effective for large scale image classification and retrieval.

The functional role of 5D4 antibody-reactive keratan sulfate (KS) in the pathogenesis of neurodegenerative diseases is unknown. We therefore studied the expression of 5D4-reactive KS in amyotrophic lateral sclerosis (ALS), a motor neuron-degenerative disease, with the use of SOD1 G93A ALS model mice and patients with ALS. Histochemical and immunoelectron microscopic characterizations showed that the 5D4-reactive KS is expressed in Mac2/galectin-3–positive activated or proliferating microglia of SOD1 G93A ALS model mice at disease end stage and that the KS is an O -linked glycan modified with sialic acid and fucose, which was thus far shown to exist in cartilage. Intriguingly, microglial KS was detected in the spinal cord and brainstem but not in the cerebral cortex of SOD1 G93A mice. We found that KSGal6ST, a galactose-6-sulfotransferase, is required for biosynthesis of the microglial 5D4-reactive KS by generating SOD1 G93A /KSGal6ST −/− mice. The requirement of GlcNAc6ST1 for this synthesis was corroborated by analyzing SOD1 G93A /GlcNAc6ST1 −/− mice. These results indicate that both galactose-6– and N acteylglucosamine-6–sulfated KS elicited in the spinal cord and brainstem are associated with the degeneration of spinal and bulbar lower motor neurons in ALS pathology and may play a role in disease progression via microglial activation and proliferation.

Dating the desert lake sediments using 210 Pb uns and 137 Cs methods is challenging because of the low atmospheric fluxes and the unexpected irregular variations in the specific activity profiles. Due to the large numbers of permanent lakes in the Badain Jaran Desert, they provide abundant sites for investigating the desert 210 Pb uns and 137 Cs geochronology. Research was performed to determine the reliable ages using this methodology and obtain a better understanding of the principal factors that affect the dating results. The top 20–30 cm of core SH-1 from Lake Shuanghaizi and of core YD-1 from Lake Yindeertu in Badain Jaran Desert were dated. The results support multiple conclusions. (1) Both cores have low surficial 210 Pb uns activities (SH-1, 24.5 Bq kg −1 ; YD-1, 59.7 Bq kg −1 ) compared with those of the cores from arid Northwest China, such as Lake Bosten (350 Bq kg −1 ) and Lake Wulungu (523.37 Bq kg −1 ) from Xinjiang. Although the 210 Pb record in SH-1 has multiple irregular variations with depth, the exponential decrease in the YD-1 activity profiles appears fairly regular. (2) The 210 Pb uns activities are highly correlated with the sample salt concentration (SSC). (3) Sediment dates for core SH-1 derived from the constant rate of 210 Pb supply (CRS) model indicate that it spans a period of 65 years (1942 AD–2007 AD) with an average sedimentation rate of 0.21 cm a −1 (0.16 g cm −2 a −1 ). Sediment dates for core YD-1 derived from the corrected CRS model indicate that it spans a period of 121 years old (1889 AD–2010 AD) with an average sedimentation rate of 0.16 cm a −1 (0.12 g cm −2 a −1 ). In both cores, the 137 Cs time markers of 1963 appeared at different depths, which may result from the differing sedimentation rates at the sites of the cores. The sedimentation rates calculated from the CRS-determined age differences between the SH-1 and YD-1 cores appear to have resulted from the topographies containing various mega-dunes and regional aeolian activity patterns. We believe that the low 210 Pb flux and the unexpected fluctuations in the 210 Pb profiles are the principal factors affecting our dating results. The low 210 Pb uns activities were caused by infrequent rainfall and mega-dune barriers. The unexpected fluctuations in the 210 Pb uns profiles resulted from the salt crystallisation that occurred via deposition or post-sedimentation. The salt crystallisation has been closely associated with abrupt regional climate changes, including rainfall, temperature and serious drought events during the 1950s–1970s.

Badain Jaran Desert is China's the third largest desert. Research on the distribution of desert interdune lakes and the surrounding geomorphical features, for the discussion on the formation of desert lakes, recharging of lake water, as well as such issues as desertification prevention and control is of great significance. This paper uses these basic geographical data, such as ETMPP, 1: 50,000 topographic map, GDEM, Google Earth data and so on. Then we integrate geological surveys in the field, and employed methods of RS, GIS to extract the relevant information on the desert lakes. The results show that: The Badain Jaran desert lakes developed in a giant sedimentary basin. It has 122 lakes whose total area is 22.9kmPP. The palaeogeomorphology and palaeohydrology play a significant part in Badain Jaran desert’s formation of the shape of the Lake and the surrounding geomorphical features.

Generating Chinese sentences require three kinds of words, including object words (lexical item), attribute words and relational words. Modal words (modal auxiliaries) are nature words which express a speaker’s attitude. They help to generate modal sentences and express modal meanings. In terms of modal meanings, modal words can be divided into three categories, expressing necessity, expectation, and disjunction respectively. Different types of modal words can be discovered by data mining and their semantic strength can be calculated. In the use of modal words the relationship between speaker’s identity and power should be considered.