Search

Script identification is an essential step in the text extraction pipeline for multi-lingual application. This paper presents an effective approach to identify scripts in scene text images. Due to the complicated background, various text styles, character similarity of different languages, script identification has not been solved yet. Under the general classification framework of script identification, we investigate two important components: feature extraction and classification layer. In the feature extraction, we utilize a hierarchical feature fusion block to extract the multi-scale features. Furthermore, we adopt an attention mechanism to obtain the local discriminative parts of feature maps. In the classification layer, we utilize a fully convolutional classifier to generate channel-level classifications which are then processed by a global pooling layer to improve classification efficiency. We evaluated the proposed approach on benchmark datasets of RRC-MLT2017, SIW-13, CVSI-2015 and MLe2e, and the experimental results show the effectiveness of each elaborate designed component. Finally, we achieve better performances than those competitive models, where the correct rates are 89.66%, 96.11%, 98.78% and 97.20% on PRC-MLT2017, SIW-13, CVSI-2015 and MLe2e, respectively. [ABSTRACT FROM AUTHOR]

Deep neural networks (DNN), such as convolutional neural networks (CNN) have been widely used for object recognition. However, they are usually unable to ensure the required intra-class compactness and inter-class separability in the kernel space. These are known to be important in pattern recognition for achieving both robustness and accuracy. In this paper, we propose to integrate a kernelized Min-Max objective in the DNN training in order to explicitly enforce both kernelized within-class compactness and between-class margin. The involved kernel space is implicitly mapped from the feature space associated with a certain upper layer of DNN by exploiting a kernel trick, while the Min-Max objective in this space is interpolated with the original DNN loss function and finally optimized in the training phase. With a very small additional computation cost, the proposed strategy can be easily integrated in different DNN models without changing any other part of the original model. The comparative recognition accuracy of the proposed method is evaluated with multiple DNN models (including shallow CNN, deep CNN and deep residual neural network models) on two benchmark datasets: CIFAR-10 and CIFAR-100. Extensive experimental results demonstrate that the integration of kernelized Min-Max objective in the training of DNN models can achieve better results compared to state-of-the-art models, without incurring additional model complexity. [ABSTRACT FROM AUTHOR]

• To our knowledge, it is the first work to study intensively mixup methods within the GAN framework for adversarial data augmentation in the long-tailed oracle character recognition. • We propose two novel mixup-based methods, including Repatch to generalize samples in the mixup generator and the TailMix mechanism to replenish samples in the tail classes for the adversarial generator. • We extensively validate the proposed model on three benchmark oracle datasets (Oracle-AYNU, OBC306 and Oracle-20K). Experimental results show that our model can yield the state-of-the-art performance as opposed to other existing oracle recognition methods. Deciphering oracle bone script is of great significance to the study of ancient Chinese culture as well as archaeology. Although recent studies on oracle character recognition have made substantial progress, they still suffer from the long-tailed data situation that results in a noticeable performance drop on the tail classes. To mitigate this issue, we propose a generative adversarial framework to augment oracle characters in the problematic classes. In this framework, the generator produces synthetic data through convex combinations of all the available samples in the corresponding classes, and is further optimized through adversarial learning with the classifier and simultaneously the discriminator. Meanwhile, we introduce Repatch to generalize samples in the generator. Since tail classes do not have sufficient data for convex combinations, we propose the TailMix mechanism to generate suitable tail class samples from other classes. Experimental results show that our proposed algorithm obtains remarkable performance in oracle character recognition and achieves new state-of-the-art average (total) accuracy with 86.03% (89.46%), 86.54% (93.86%), 95.22% (96.17%) on the three datasets Oracle-AYNU, OBC306 and Oracle-20K, respectively. [ABSTRACT FROM AUTHOR]

Abstract: This paper presents an effective approach for unsupervised language model adaptation (LMA) using multiple models in offline recognition of unconstrained handwritten Chinese texts. The domain of the document to recognize is variable and usually unknown a priori, so we use a two-pass recognition strategy with a pre-defined multi-domain language model set. We propose three methods to dynamically generate an adaptive language model to match the text output by first-pass recognition: model selection, model combination and model reconstruction. In model selection, we use the language model with minimum perplexity on the first-pass recognized text. By model combination, we learn the combination weights via minimizing the sum of squared error with both L2-norm and L1-norm regularization. For model reconstruction, we use a group of orthogonal bases to reconstruct a language model with the coefficients learned to match the document to recognize. Moreover, we reduce the storage size of multiple language models using two compression methods of split vector quantization (SVQ) and principal component analysis (PCA). Comprehensive experiments on two public Chinese handwriting databases CASIA-HWDB and HIT-MW show that the proposed unsupervised LMA approach improves the recognition performance impressively, particularly for ancient domain documents with the recognition accuracy improved by 7 percent. Meanwhile, the combination of the two compression methods largely reduces the storage size of language models with little loss of recognition accuracy. [Copyright &y& Elsevier]

Abstract: Creating document image datasets with ground-truths of regions, text lines and characters is a prerequisite for document analysis research. However, ground-truthing large datasets is not only laborious and time consuming but also prone to errors due to the difficulty of character segmentation and the large variability of character shape, size and position. This paper describes an effective recognition-based annotation approach for ground-truthing handwritten Chinese documents. Under the Bayesian framework, the alignment of text line images with text transcript, which is the crucial step of annotation, is formulated as an optimization problem by incorporating geometric context of characters and character recognition model. We evaluated the alignment performance on a Chinese handwriting database CASIA-HWDB, which contains nearly four million character samples of 7356 classes and 5091 pages of unconstrained handwritten texts. The experimental results demonstrate the superiority of recognition-based text line alignment and the benefit of integrating geometric context. On a test set of 1015 handwritten pages (10,449 text lines), the proposed approach achieved character level alignment accuracy 92.32% when involving under-segmentation errors and 99.04% when excluding under-segmentation errors. The tool based on the proposed approach has been practically used for labeling handwritten Chinese documents. [Copyright &y& Elsevier]

Seasonal changes in CO 2 exchange determine the strength and dynamics of the carbon budget in terrestrial ecosystems. Knowledge about seasonal and spatial variations in carbon fluxes (net ecosystem productivity (NEP), gross primary productivity (GPP), and ecosystem respiration (RE)) is thus essential to support accurate modeling of carbon flux dynamics. To the best of our knowledge, this study represents the first comprehensive evaluation of how carbon fluxes vary seasonally in different ecoregions in China. Using eddy covariance measurements from ChinaFLUX, we analyzed the spatial variations and climate controls of the phenological and physiological properties of NEP, GPP, and RE across China. Apart from the beginning and ending day of NEP (net carbon uptake), the phenological properties of the carbon fluxes varied significantly with latitude; however, there were no evident trends in the physiological properties of the carbon fluxes with latitude (apart from the annual values of GPP and RE, known as AGPP and ARE). The spatial variation in the physiological properties of GPP and RE was influenced by the precipitation patterns, while the spatial distribution of the phenological properties of GPP and RE was related to air temperature. Moreover, the spatial variation for the end of the growing season changed with the autumn mean soil temperatures. The factors that contributed to the spatial distribution in the annual NEP (ANEP) and annual GPP and RE (AGPP and ARE) were quite different. The mean daily GPP and RE were the main contributors to the spatial variations in AGPP and ARE, while the net carbon uptake period was the main contributor to the spatial variation in ANEP. In this study, we identified a series of ecological parameters and reference values for seasonal patterns that can be used to validate models that simulate changes in regional carbon fluxes. [ABSTRACT FROM AUTHOR]

Net ecosystem productivity (NEP), the difference between gross primary productivity (GPP) and ecosystem respiration (ER), is the basis of forest carbon sinks. Revealing NEP differences between naturally regenerating forests (NF) and planted forests (PF) can benefit for making carbon neutrality strategies. Based on 35 eddy covariance measurements in China, we analyzed NEP differences in values and spatial patterns between NF and PF. The results showed that NF had slightly lower NEP than PF, resulting from the high stand age (SA) and soil fertilizer, while their differences were not significant (p > 0.05). The increasing latitude decreased mean annual air temperature thus decreased GPP both in NF and PF. However, the higher SA and soil fertilizer in NF made most GPP release as ER thus induced no significant NEP spatial variation, while lower SA and soil fertilizer in PF made NEP spatially couple with GPP thus showed a decreasing latitudinal pattern. Therefore, stand characteristics determined the differences in NEP values but indirectly affected the differences in NEP spatial variations through altering GPP allocation. The decreasing latitudinal pattern of NEP in PF indicates a higher sequestration capacity in the PF of South China. Our results provide a basis for improving the forest carbon sequestration. [ABSTRACT FROM AUTHOR]

• We present a timely and comprehensive survey on robust adversarial training. • This survey offers the fundamentals of adversarial training, a unified theory that can be used to interpret various methods, and a comprehensive summarization of different methodologies. • This survey also addresses three important research focus in adversarial training: interpretability, robust generalization, and robustness evaluation, which can stimulate future inspirations as well as research outlook. Deep neural networks have achieved remarkable success in machine learning, computer vision, and pattern recognition in the last few decades. Recent studies, however, show that neural networks (both shallow and deep) may be easily fooled by certain imperceptibly perturbed input samples called adversarial examples. Such security vulnerability has resulted in a large body of research in recent years because real-world threats could be introduced due to the vast applications of neural networks. To address the robustness issue to adversarial examples particularly in pattern recognition, robust adversarial training has become one mainstream. Various ideas, methods, and applications have boomed in the field. Yet, a deep understanding of adversarial training including characteristics, interpretations, theories, and connections among different models has remained elusive. This paper presents a comprehensive survey trying to offer a systematic and structured investigation on robust adversarial training in pattern recognition. We start with fundamentals including definition, notations, and properties of adversarial examples. We then introduce a general theoretical framework with gradient regularization for defending against adversarial samples - robust adversarial training with visualizations and interpretations on why adversarial training can lead to model robustness. Connections will also be established between adversarial training and other traditional learning theories. After that, we summarize, review, and discuss various methodologies with defense/training algorithms in a structured way. Finally, we present analysis, outlook, and remarks on adversarial training. [ABSTRACT FROM AUTHOR]

This paper presents an effective approach for the offline recognition of unconstrained handwritten Chinese texts. Under the general integrated segmentation-and-recognition framework with character oversegmentation, we investigate three important issues: candidate path evaluation, path search, and parameter estimation. For path evaluation, we combine multiple contexts (character recognition scores, geometric and linguistic contexts) from the Bayesian decision view, and convert the classifier outputs to posterior probabilities via confidence transformation. In path search, we use a refined beam search algorithm to improve the search efficiency and, meanwhile, use a candidate character augmentation strategy to improve the recognition accuracy. The combining weights of the path evaluation function are optimized by supervised learning using a Maximum Character Accuracy criterion. We evaluated the recognition performance on a Chinese handwriting database CASIA-HWDB, which contains nearly four million character samples of 7,356 classes and 5,091 pages of unconstrained handwritten texts. The experimental results show that confidence transformation and combining multiple contexts improve the text line recognition performance significantly. On a test set of 1,015 handwritten pages, the proposed approach achieved character-level accurate rate of 90.75 percent and correct rate of 91.39 percent, which are superior by far to the best results reported in the literature. [ABSTRACT FROM PUBLISHER]

Whilst adversarial training has been shown as a promising wisdom to promote model robustness in computer vision and machine learning, adversarially trained models often suffer from poor robust generalization on unseen adversarial examples. Namely, there still remains a big gap between the performance on training and test adversarial examples. In this paper, we propose to tackle this issue from a new perspective of the inter-feature relationship. Specifically, we aim to generate adversarial examples which maximize the loss function while maintaining the inter-feature relationship of natural data as well as penalizing the correlation distance between natural features and adversarial counterparts. As a key contribution, we prove that training with such examples while penalizing the distance between correlations can help promote both the generalization on natural and adversarial examples theoretically. We empirically validate our method through extensive experiments over different vision datasets (CIFAR-10, CIFAR-100, and SVHN), against several competitive methods. Our method substantially outperforms the baseline adversarial training by a large margin, especially for PGD20 on CIFAR-10, CIFAR-100, and SVHN with around 20%, 15% and 29% improvements. [ABSTRACT FROM AUTHOR]

Analyzing the approaches that climatic factors affect the spatial variation of annual gross primary productivity (GPP yr ) would improve our understanding on its spatial pattern. Based on network eddy covariance measurements and published data in literature, we separated GPP yr into radiation use efficiency (RUE) and annual absorbed photosynthesis active radiation (APAR yr ), where APAR yr can be regarded as the product of the fraction of absorbed annual photosynthesis active radiation (FPAR yr ) and annual PAR (PAR yr ). Given that PAR yr affects the spatial variation of GPP yr directly through itself, we investigated factors affecting the spatial variations of RUE and FPAR yr , to reveal how climatic factors affect the spatial variation of GPP yr . Results suggest that the spatial variation of RUE was directly affected by annual mean air temperature (MAT) and annual mean CO 2 mass concentration ( ρ cyr ). The increasing MAT and ρ cyr directly enhanced RUE. The increasing annual precipitation (MAP) directly prompted FPAR yr . Therefore, MAT and ρ cyr affected the spatial variation of GPP yr through altering RUE while the effect of MAP was achieved through altering FPAR yr . Our study could also provide an alternative way for regional GPP yr assessment. [ABSTRACT FROM AUTHOR]

Water use efficiency (WUE) reflects the coupling of carbon and water cycles. Analyzing the spatial variability of WUE can improve our understanding on the interaction between carbon and water cycles at a large scale, which also provides a basis for improving the regional carbon budget assessment. Based on China's eddy covariance measurements, we examined the spatial variation of China's WUE and its affecting factors. WUE showed a decreasing trend with the increasing altitude, which was the result of ecosystem type distribution resulting from the climatic gradient. After fully considering the vertical variation of WUE, we found that not only mean annual air temperature (MAT), mean annual precipitation (MAP), and mean leaf area index (MLAI) but also mean annual total photosynthesis active radiation (MAR) affected the spatial variation of WUE. With the increasing MAT, MAP, and MLAI, WUE increased significantly but the increasing MAR decreased WUE. The spatial variation of WUE could be directly depicted by MLAI and altitude, the equation including which explained 65% of the spatial variation of WUE. The effects of MAT and MAP on the spatial variation of WUE may be achieved through altering MLAI, while the mechanism underlying the effect of MAR on the spatial variation of WUE was still unclear, which should be the subject of future investigations. This study reveals the vertical variation of WUE and provides a new approach to generate the spatial variation in WUE, which will benefit the regional carbon budget assessment. [ABSTRACT FROM AUTHOR]

Anomaly detection under the open-set scenario is a challenging task that requires learning discriminative features to detect anomalies that were even unseen during training. As a cheap yet effective approach, data augmentation has been widely used to create pseudo anomalies for better training of such models. Recent wisdom of augmentation methods focuses on generating random pseudo instances that may lead to a mixture of augmented instances with seen anomalies, or out of the typical range of anomalies. To address this issue, we propose a novel saliency-guided data augmentation method, SaliencyCut, to produce pseudo but more common anomalies that tend to stay in the plausible range of anomalies. Furthermore, we deploy a two-head learning strategy consisting of normal and anomaly learning heads to learn the anomaly score of each sample. Theoretical analyses show that this mechanism offers a more tractable and tighter lower bound of the data log-likelihood. We then design a novel patch-wise residual module in the anomaly learning head to extract and assess anomaly features from each sample, facilitating the learning of discriminative representations of anomaly instances. Extensive experiments conducted on six real-world anomaly detection datasets demonstrate the superiority of our method to competing methods under various settings. Codes are available at: https://github.com/yjnanan/SaliencyCut. [Display omitted] • We propose a saliency-guided data augmentation for creating pseudo anomalies, namely SaliencyCut. • We deploy a two-head learning mechanism based on the theoretical lower bound. • We further design a novel patch-wise residual module to gather anomaly knowledge. • Comprehensive experiments on six real application datasets validate the effectiveness of our method. [ABSTRACT FROM AUTHOR]

The segmentation of touching characters is still a challenging task, posing a bottleneck for offline Chinese handwriting recognition. In this paper, we propose an effective over-segmentation method with learning-based filtering using geometric features for single-touching Chinese handwriting. First, we detect candidate cuts by skeleton and contour analysis to guarantee a high recall rate of character separation. A filter is designed by supervised learning and used to prune implausible cuts to improve the precision. Since the segmentation rules and features are independent of the string length, the proposed method can deal with touching strings with more than two characters. The proposed method is evaluated on both the character segmentation task and the text line recognition task. The results on two large databases demonstrate the superiority of the proposed method in dealing with single-touching Chinese handwriting. [ABSTRACT FROM AUTHOR]

• An AGPP data was generated through building its yearly mapping schemes. • Chinese total AGPP was 7.39 ± 0.62 PgC yr−1 with a growth rate of 0.095 PgC yr−2. • Chinese AGPP trends spatially decreased from east to west. • Ecosystem responses dominate AGPP trends via structural properties in north regions. • Ecosystem responses dominate AGPP trends in west and south regions via functional properties. The annual gross primary productivity (AGPP) forms the foundation for terrestrial carbon sink. AGPP trends stem from the combined impacts of environmental factors and ecosystem responses to those factors. These responses include model structure modifications altering both the structural properties (represented by leaf area index) and the functional properties (represented by per leaf AGPP). Understanding the contributions of ecosystem responses to AGPP trends is crucial for accurately assessing spatiotemporal variations in AGPP, which aids in effective carbon management. Based on eddy covariance measurements of AGPP in China, we generated AGPP during 2000–2020 by constructing yearly AGPP mapping schemes from site measurements and analyzed AGPP trends. The roles of ecosystem responses and their pathways were further elucidated by employing analysis of variance. Results showed that the total AGPP in China was 7.39 ± 0.62 PgC yr−1, with an increasing trend of 0.095 PgC yr−2. This increase primarily sourced from the total AGPP in Subtropical Evergreen Needleleaf Forest. AGPP trends spatially decreased from east to west, with most regions showing positive values. Structural and functional properties contributed similarly to AGPP trends, with varying regional contributions. The structural properties played a dominant role in north regions while the functional properties were more important in south and west regions. Ecosystem responses similarly determined the trends of ecosystem properties, which decreased from west to east. Ecosystem responses thus dominated AGPP trends through the structural properties in north regions but through the functional properties in south and west regions. Therefore, Chinese AGPP exhibited a spatially varied increasing trend during 2000–2020. Ecosystem responses dominated AGPP trends through divergent properties across regions. Our results highlight the significanc of ecosystem responses in AGPP trends and their divergent pathways across regions. This also provides independent data support for quantifying regional carbon budgets. [Display omitted] [ABSTRACT FROM AUTHOR]

• cylinder primitive extraction is addressed as a clustering problem of possible cylinder candidates; • candidates are generated by cutting the scene through several random planes; • a novel cylinder similarity function based on dual quaternion algebra is defined; • the clustering process, based on Game Theory, inherently discards incompatible candidates; • the proposed method can deal with noisy, partially occluded and non-oriented point clouds. Generative Adversarial Networks (GAN) receive great attention recently due to its excellent performance in image generation, transformation, and super-resolution. However, less emphasis or study has been put on GAN for classification with super-resolution. Moreover, though GANs may fabricate images which perceptually looks realistic, they usually fabricate some fake details especially in character data; this would impose further difficulties when they are input for classification. In this paper, we propose a novel Generative Adversarial Classifier (GAC) for low-resolution handwriting character recognition. Specifically, we design an additional classifier component in GAC, leading to a novel three-player GAN model which is not only able to generate high-quality super-resolved images, but also favorable for classification. Experimental results show that our proposed method can obtain remarkable performance in handwriting characters with 8 × super-resolution, achieving new state-of-the-art on benchmark dataset CASIA-HWDB1.1, and MNIST. [ABSTRACT FROM AUTHOR]

Evapotranspiration (ET), which is comprised by evaporation from soil surface ( E ), transpiration ( T ) and evaporation from the intercepted water by canopy (EI), plays an important role in maintaining global energy balance and regulating climate. Quantifying the spatiotemporal variations of T /ET (the ratio of T to ET) can improve our understandings on the role of vegetation ecophysiological processes in climate regulation. Using eddy covariance measurements at three forest ecosystems (Changbaishan temperate broad-leaved Korean pine mixed forest (CBS), Qianyanzhou subtropical coniferous plantation (QYZ) and Dinghushan subtropical evergreen mixed forest (DHS)) in north–south transect of Eastern China (NSTEC), we run the revised Shuttleworth–Wallace model (S–W model), validated its performance with the water vapor fluxes measured at two layers, and quantified the spatiotemporal variations of T /ET. The S–W model performed well in simulating ET and T /ET. The mean value of annual T /ET at three forests during the observation period all exceeded 0.6. The diurnal variation of canopy stomal conductance ( G c ) dominated that of T /ET. The seasonal dynamics of T /ET was mainly shaped by that of leaf area index (LAI), vapor pressure deficit (VPD) and air temperature ( T a ) through altering G c and the portion that the energy absorbed by canopy ( P EC ) at temperate forest (CBS), while the seasonal dynamics of T /ET at subtropical forests (QYZ and DHS) were mainly affected by T a , net radiation, VPD, and soil water content through altering G c and soil surface conductance ( G s ). The variation of mean annual G c governed the interannual varaition and spatial variation of T /ET. Therefore, forests in Eastern China played an important role in regulating climate through T and G c primarily affected the spatial and temproal variations of the role of forest T in regulating climate. [ABSTRACT FROM AUTHOR]

Accurate quantifying the magnitudes and distributions of carbon budgets is helpful for strategies in mitigating global climate change. Based on spatial patterns of carbon fluxes (gross ecosystem productivity (GEP), ecosystem respiration (ER) and net ecosystem productivity (NEP)) and their drivers, we constructed geographical statistical assessment schemes and quantified the magnitudes of carbon fluxes in China. The optimal assessment scheme was then validated with observed eddy covariance data to analyze the spatial distributions of carbon fluxes. Using climate-based geographical statistical assessment schemes, our estimates of GEP, ER and NEP in China during 2000s were 7.51±0.51, 5.82±0.16 and 1.91±0.15PgCyr−1, corresponding to 4.29%–6.80%, 5.65%–6.06% and 9.10%–12.73% of global annual carbon fluxes, respectively. The spatial distributions of GEP, ER and NEP, generated from the optimal scheme, were similar, following a southeast–northwest decreasing gradient. The maximum values for GEP, ER and NEP were 1790, 1300 and 490gCm−2 yr−1, respectively, which occurred in Central subtropics and Southern subtropics. Climate-based geographical statistical assessment schemes provided an independent dataset for the regional carbon budget assessment, which can be deemed as the potential carbon fluxes. Meanwhile, most areas in China were potential carbon sink especially Eastern China and the largest potential carbon sink appeared in Central subtropics and Southern subtropics. [ABSTRACT FROM AUTHOR]

Understanding the dynamics and underlying mechanism of carbon exchange between terrestrial ecosystems and the atmosphere is one of the key issues in global change research. In this study, we quantified the carbon fluxes in different terrestrial ecosystems in China, and analyzed their spatial variation and environmental drivers based on the long-term observation data of China FLUX sites and the published data from other flux sites in China. The results indicate that gross ecosystem productivity ( GEP), ecosystem respiration ( ER), and net ecosystem productivity ( NEP) of terrestrial ecosystems in China showed a significantly latitudinal pattern, declining linearly with the increase of latitude. However, GEP, ER, and NEP did not present a clear longitudinal pattern. The carbon sink functional areas of terrestrial ecosystems in China were mainly located in the subtropical and temperate forests, coastal wetlands in eastern China, the temperate meadow steppe in the northeast China, and the alpine meadow in eastern edge of Qinghai-Tibetan Plateau. The forest ecosystems had stronger carbon sink than grassland ecosystems. The spatial patterns of GEP and ER in China were mainly determined by mean annual precipitation ( MAP) and mean annual temperature ( MAT), whereas the spatial variation in NEP was largely explained by MAT. The combined effects of MAT and MAP explained 79%, 62%, and 66% of the spatial variations in GEP, ER, and NEP, respectively. The GEP, ER, and NEP in different ecosystems in China exhibited 'positive coupling correlation' in their spatial patterns. Both ER and NEP were significantly correlated with GEP, with 68% of the per-unit GEP contributed to ER and 29% to NEP. MAT and MAP affected the spatial patterns of ER and NEP mainly by their direct effects on the spatial pattern of GEP. [ABSTRACT FROM AUTHOR]

Abstract: Recently, the Institute of Automation of Chinese Academy of Sciences (CASIA) released the unconstrained online and offline Chinese handwriting databases CASIA-OLHWDB and CASIA-HWDB, which contain isolated character samples and handwritten texts produced by 1020 writers. This paper presents our benchmarking results using state-of-the-art methods on the isolated character datasets OLHWDB1.0 and HWDB1.0 (called DB1.0 in general), OLHWDB1.1 and HWDB1.1 (called DB1.1 in general). The DB1.1 covers 3755 Chinese character classes as in the level-1 set of GB2312-80. The evaluated methods include 1D and pseudo 2D normalization methods, gradient direction feature extraction from binary images and from gray-scale images, online stroke direction feature extraction from pen-down trajectory and from pen lifts, classification using the modified quadratic discriminant function (MQDF), discriminative feature extraction (DFE), and discriminative learning quadratic discriminant function (DLQDF). Our experiments reported the highest test accuracies 89.55% and 93.22% on the HWDB1.1 (offline) and OLHWDB1.1 (online), respectively, when using the MQDF classifier trained with DB1.1. When training with both the DB1.0 and DB1.1, the test accuracies on HWDB1.1 and OLHWDB are improved to 90.71% and 93.95%, respectively. Using DFE and DLQDF, the best results on HWDB1.1 and OLHWDB1.1 are 92.08% and 94.85%, respectively. Our results are comparable to the best results of the ICDAR2011 Chinese Handwriting Recognition Competition though we used less training samples. [Copyright &y& Elsevier]

Abstract: As a weather element, clouds can affect CO2 exchange between terrestrial ecosystems and the atmosphere by altering environmental conditions, such as solar radiation received on the ground surface, temperature, and moisture. Based on the flux data measured at five typical ecosystems of China during mid-growing season (June–August) from 2003 to 2006, we analyzed the responses of net ecosystem exchange of carbon dioxide (NEE), light use efficiency (LUE, defined as Gross ecosystem photosynthesis (GEP)/Photosynthetically active radiation (PAR)), and water use efficiency (WUE, defined as GEP/Evapotranspiration (ET)) to the changes in cloudiness. The five ecological sites included Changbaishan temperate mixed forest (CBS), Dinghushan subtropical evergreen broad-leaved forest (DHS), Xishuangbanna tropical rainforest (XSBN), Inner Mongolia semi-arid Leymus chinensis steppe (NMG), and Haibei alpine frigid Potentilla fruticosa shrub (HB). Our analyses show that cloudy sky conditions with cloud index (k t) values between 0.4 and 0.6 increased NEE, LUE, and WUE of the ecosystems at CBS, DHS, NMG and HB from June to August. The LUE of tropical rainforest at XSBN was higher under cloudy than under clear sky conditions, but NEE and WUE did not decrease significantly under clear sky conditions from June to August. The increase in GEP with increasing diffuse radiation received by ecosystems under cloudy skies was the main reason that caused the increases in LUE and net carbon uptake in forest ecosystem at CBS, DHS, and alpine shrub ecosystem at HB, compared with clear skies. Moreover, for the ecosystem at CBS, DHS, and HB, when sky condition became from clear to cloudy, GEP increased and ET decreased with decreasing VPD, leading to the increase in WUE and NEE under cloudy sky conditions. The decrease in R e with decreasing temperature and increase in GEP with decreasing VPD under cloudy skies led to the increase in LUE, WUE, and net carbon uptake of semi-arid steppe at NMG, compared to clear skies. These different responses among the five ecosystems are attributable to the differences in canopy characteristics and water conditions. From June to August, the peaks of the k t frequency distribution in temperate ecosystems (e.g., CBS, NMG, and HB) were larger than 0.5, but they were smaller than 0.4 in subtropical/tropical forest ecosystems (e.g., DHS and XSBN). These results suggest that the pattern of cloudiness during the years from 2003 to 2006 in the five ecosystems was not the best condition for their net carbon uptake. This study highlights the importance of cloudiness factor in the prediction of net carbon absorption in the Asia monsoon region under climate change. [Copyright &y& Elsevier]

Net ecosystem productivity (NEP) was continuously measured using the eddy covariance (EC) technique from 2003 to 2005 at three forest sites of ChinaFLUX. The forests include Changbaishan temperate mixed forest (CBS), Qianyanzhou subtropical coniferous plantation (QYZ), and Dinghushan subtropical evergreen broad-leaved forest (DHS). They span wide ranges of temperature and precipitation and are influenced by the eastern Asian monsoon climate to varying extent. In this study, we estimated ecosystem respiration (RE) and gross ecosystem productivity (GEP). Comparison of ecosystem carbon exchange among the three forests shows that RE was mainly determined by temperature, with the forest at CBS exhibiting the highest temperature sensitivity among the three ecosystems. The RE was highly dependent on GEP across the three forests, and the ratio of RE to GEP decreased along the North–South Transect of Eastern China (NSTEC) (i.e. from the CBS to the DHS), with an average of 0.77 ± 0.06. Daily GEP was mainly influenced by temperature at CBS, whereas photosynthetic photon flux density was the dominant factor affecting the daily GEP at both QYZ and DHS. Temperature mainly determined the pattern of the interannual variations of ecosystem carbon exchange at CBS. However, water availability primarily controlled the interannual variations of ecosystem carbon exchange at QYZ. At DHS, NEP attained the highest values at the beginning of the dry seasons (autumn) rather than the rainy seasons (summer), probably because insufficient radiation and frequent fog during the rainy seasons hindered canopy photosynthesis. All the three forest ecosystems acted as a carbon sink from 2003 to 2005. The annual average values of NEP at CBS, QYZ, and DHS were 259 ± 19, 354 ± 34, and 434 ± 66 g C m−2 yr−1, respectively. The slope of NEP that decreased with increasing latitude along the NSTEC was markedly different from that observed on the forest transect in the European continent. Long-term flux measurements over more forest ecosystems along the NSTEC will further help verify such a difference between the European forest transect and the NSTEC and provide insights into the responses of ecosystem carbon exchange to climate change in China. [ABSTRACT FROM AUTHOR]

The study presents a theoretical basis of a stomatal behavior-based coupled model for estimating photosynthesis, A, and transpiration, E. Outputs of the model were tested against data observed in a maize (Zea mays L.) field. The model was developed by introducing the internal conductance, g ic, to CO2 assimilation, and the general equation of stomatal conductance, g sw, to H2O diffusion, into models of CO2 and H2O diffusion through the stomata of plant leaves. The coupled model is easier for practical use since the model only includes environmental variables, such as ambient CO2 concentration, leaf temperature, humidity and photosynthetic photon flux received at the leaves within the canopy. Moreover, concept of g ic, and factors controlling A and E were discussed, and applicability of the model was examined with the data collected in the maize field. [ABSTRACT FROM AUTHOR]