Texture Classification System Research Articles

Steff: A Novel Spatio-Temporal System for Dynamic Texture Classification Using Efficientnet on Outdoor Scenes

Steff: A Novel Spatio-Temporal System for Dynamic Texture Classification Using Efficientnet on Outdoor Scenes

Transformation of soil texture schemes and determination of water-physical properties of soils

Measuring soil water-physical properties is laborious, time-consuming, and expensive. That provokes a lot of scientists to estimate them which action is troubled by the usage of different soil texture classification systems. The study proposes a rapid, reliable, and universally applicable methodology for soil textural transformations between different classification systems. The method of discrete mathematics is applied to make the conversion of particle-size classes from the Kachinsky system, which is used in Bulgaria to the International systems. Three different data sources were used to determine the water-physical properties of soils from textural data - 376 soil profiles from Bulgaria, extraction from the SoilGrids system for the Plovdiv district in Bulgaria and data from CanSIS/NSDB database. The relationship between the dependent variables field capacity (FC), wilting point (WP) and bulk density (BD), and independent variables sand, silt, and clay soil content was sought in the form of a regression equation. The applied stepwise regression procedure produces a close dependence between the soil texture and its water-physical properties.

Laser Diffraction as An Innovative Alternative to Standard Pipette Method for Determination of Soil Texture Classes in Central Europe

The paper presents the comparison of soil particle size distribution determined by standard pipette method and laser diffraction. Based on the obtained results (542 soil samples from 271 sites located in the Nitra, Váh and Hron River basins), regression models were calculated to convert the results of the particle size distribution by laser diffraction to pipette method. Considering one of the most common soil texture classification systems used in Slovakia (according to Novák), the emphasis was placed on the determination accuracy of particle size fraction <0.01 mm. Analysette22 MicroTec plus and Mastersizer2000 devices were used for laser diffraction. Polynomial regression model resulted in the best approximation of measurements by laser diffraction to values obtained by pipette method. In the case of particle size fraction <0.01 mm, the differences between the measured values by pipette method and both laser analyzers ranged in average from 3% up to 9% and from 2% up to 11% in the case of Analysette22 and Mastersizer2000, respectively. After correction, the differences decreased to average 3.28% (Analysette22) and 2.24% (Mastersizer2000) in comparison with pipette method. After recalculation of the data, laser diffraction can be used alongside the sedimentation methods.

A structural based feature extraction for detecting the relation of hidden substructures in coral reef images

In this paper, we present an efficient approach to extract local structural color texture features for classifying coral reef images. Two local texture descriptors are derived from this approach. The first one, based on Median Robust Extended Local Binary Pattern (MRELBP), is called Color MRELBP (CMRELBP). CMRELBP is very accurate and can capture the structural information from color texture images. To reduce the dimensionality of the feature vector, the second descriptor, co-occurrence CMRELBP (CCMRELBP) is introduced. It is constructed by applying the Integrative Co-occurrence Matrix (ICM) on the Color MRELBP images. This way we can detect and extract the relative relations between structural texture patterns. Moreover, we propose a multiscale LBP based approach with these two schemes to capture microstructure and macrostructure texture information. The experimental results on coral reef (EILAT, EILAT2, RSMAS, and MLC) and four well-known texture datasets (OUTEX, KTH-TIPS, CURET, and UIUCTEX) show that the proposed scheme is quite effective in designing an accurate, robust to noise, rotation and illumination invariant texture classification system. Moreover, it makes an admissible tradeoff between accuracy and number of features.

Assessment of the Impacts of Land Use/Cover Change and Rainfall Change on Surface Runoff in China

Assessment of the impacts of land use/cover change (LUCC) and rainfall change on surface runoff depth can help provide an understanding of the temporal trend of variation of surface runoff and assist in urban construction planning. This study evaluated the impacts of LUCC and rainfall change on surface runoff depth by adopting the well-known Soil Conservation Service-Curve Number (SCS-CN) method and the widely used Long-Term Hydrologic Impact Assessment (L-THIA) model. National hydrologic soil group map of China was generated based on a conversion from soil texture classification system. The CN values were adjusted based on the land use/cover types and soil properties in China. The L-THIA model was configured by using the adjusted CN values and then applied nationally in China. Results show that nationwide rainfall changes and LUCC from 2005 to 2010 had little impact on the distribution of surface runoff, and the high values of runoff depth were mainly located in the middle and lower reaches of the Yangtze River. Nationally, the average annual runoff depths in 2005, 2010 and 2015 were 78 mm, 83 mm and 90 mm, respectively. For the 2015 land use data, rainfall change caused the variation of surface runoff depth ranging from −203 mm to 476 mm in different regions. LUCC from 2005 to 2015 did not cause obvious change of surface runoff depth, but expansion of developed land led to runoff depth increases ranging from 0 mm to 570 mm and 0 mm to 742 mm from 2005 to 2010 and 2010 to 2015, respectively. Potential solutions to urban land use change and surface runoff control were also analyzed.

A Miniaturized Force Sensor Based on Hair-Like Flexible Magnetized Cylinders Deposited Over a Giant Magnetoresistive Sensor

The detection of force with higher resolution than observed in humans (~1 mN) is of great interest for emerging technologies, especially surgical robots, since this level of resolution could allow these devices to operate in extremely sensitive environments without harming these. In this paper, we present a force sensor fabricated with a miniaturized footprint (9 mm2), based on the detection of the magnetic field generated by magnetized flexible pillars over a giant magnetoresistive sensor. When these flexible pillars deflect due to external loads, the stray field emitted by these will change, thus varying the GMR sensor resistance. A sensor with an array of five pillars with $200~\mu \text{m}$ diameter and 1 mm height was fabricated, achieving a 0 to 26 mN measurement range and capable of detecting a minimum force feature of $630~\mu \text{N}$ . A simulation model to predict the distribution of magnetic field generated by the flexible pillars on the sensitive area of the GMR sensor in function of the applied force was developed and validated against the experimental results reported in this paper. The sensor was finally tested as a texture classification system, with the ability of differentiating between four distinct surfaces varying between 0 and $162~\mu \text{m}$ root mean square surface roughness.

Analysis of Agronomic Categories in Different Soil Texture Classification Systems

Different soil texture classification systems are used in Poland. The system most widely used in agriculture is named after Polish Soil Science Society (PSSS) and is described in the soil classification norm BN-78/9180-11 (BN1978 standard). The last edition of soil classification system and soil texture classes published by PSSS in 2008 (PSSS 2008 classification) is different from BN1978 standard. The aim of this paper is a quantitative and qualitative comparison of the compatibility of agronomic categories created according to the old textural classes (BN1978 standard) and the new textural classes (PSSS 2008 classification). The representative set of soil samples (n=316) for arable mineral soils in Poland were divided into agronomic categories according to these two soil classification systems. The agronomic categories, which comply with soil classification systems PSSS 1978 are widely used in agricultural advisory. The results of the study showed differences in the amount of soil samples classified for the corresponding agronomic category. The study also showed discrepancies in the fine particle (<0.02 mm) and colloidal fraction (<0.002 mm) content in the corresponding categories. The differences may affect the assessment of soil fertility in nutrients(abundance) such as potassium, magnesium and of soil liming needs, as well as appropriate determination of fertiliser doses.

Detection and prediction of osteoarthritis in knee and hand joints based on the X-ray image analysis

Abstract Current assessment of osteoarthritis (OA) is primary based on visual grading of joint space narrowing and osteophytes present on radiographs. The approach is observer-dependent, not sensitive enough for the detection of the early stages of OA and time consuming. A promising solution is through fractal analysis of trabecular bone (TB) textures on radiographs. The goal is to develop an automated decision support system for the detection and prediction of OA based on TB texture regions selected on knee and hand radiographs. In this review, we describe our progress towards this development which was conducted in five stages, i.e., (i) development of automated methods for the selection of TB texture regions on knee and hand radiographs (ii), development of fractal signature methods for TB texture analysis, (iii) applications of the methods in the analysis of x-ray images of knees and hands, (iv) development of TB texture classification system, and (v) development of ReadMyXray website for knee x-ray analysis. The results achieved so far are encouraging and it is hoped, that once the system is fully developed and evaluated, it will be used to aid medical practitioners in the decision-making, i.e., in designing OA preventative measures, treatments and monitoring the OA progression.

BiomassID: A biomass type identification system for mobile devices

Biomass quality assessment is of great importance when one in the biomass industry needs to produce another energy product, such as biofuel or bioenergy, for instance. Usually, the biomass quality is determined using expensive devices, such as mass spectrometers, or complex chemical tests that may need several days to complete. Because of the high costs of such methods, people tend to forsake biomass quality assessment and move on to directly produce bioproducts from any kind of biomass. In this paper, a cheap and fast solution for biomass type identification is proposed and investigated. The quality of biomass can be inferred at a coarse level from the type of biomass. In the proposed approach, biomass type identification is treated as a texture classification problem. A texture classification system developed for mobile devices which is able to distinguish between four types of biomass texture images is presented in this paper. Several state of the art texture classification systems based on machine learning are evaluated in a series of experiments on a data set of biomass texture images. The experiments are conducted to determine the system that can offer the best trade-off between accuracy and speed, since the goal is to implement it on a mobile device with limited processing power and memory. In the end, the selected system can identify the type of biomass from pictures taken with a mobile device camera in a few seconds directly on the respective mobile device. The utility of the system is demonstrated through an iOS application that is freely available for download in the App Store at http://appstore.com/biomassid.

Fusing Sorted Random Projections for Robust Texture and Material Classification

This paper presents a conceptually simple, and robust, yet highly effective, approach to both texture classification and material categorization. The proposed system is composed of three components: 1) local, highly discriminative, and robust features based on sorted random projections (RPs), built on the universal and information-preserving properties of RPs; 2) an effective bag-of-words global model; and 3) a novel approach for combining multiple features in a support vector machine classifier. The proposed approach encompasses the simplicity, broad applicability, and efficiency of the three methods. We have tested the proposed approach on eight popular texture databases, including Flickr Materials Database, a highly challenging materials database. We compare our method with 13 recent state-of-the-art methods, and the experimental results show that our texture classification system yields the best classification rates of which we are aware of 99.37% for Columbia–Utrecht, 97.16% for Brodatz, 99.30% for University of Maryland Database, and 99.29% for Kungliga Tekniska hogskolan-textures under varying illumination, pose, and scale. Moreover, the proposed approach significantly outperforms the current state-of-the-art approach in materials categorization, with an improvement to classification accuracy of 67%.

Conversion between Soil Texture Classification Systems using the Random Forest Algorithm

This study focuses on the reclassification of a soil texture system following a hybrid approach in which the conventional particle-size distribution (PSD) models are coupled with a random forest (RF) algorithm for achieving more generally applicable and precise outputs. The existing parametric PSD models that could be used for this purpose have various limitations; different models frequently show unequal degrees of precision in different soils or under different environments. The authors present in this article a novel ensemble modeling approach in which the existing PSD models are used as ensemble members. An improvement in precision was proved by better statistical indicators for the results obtained, and the article documents that the ensemble model worked better than any of its constituents (different existing parametric PSD models). This study is verified by using a soil dataset from Slovakia, which was originally labeled by a national texture classification system, which was then transformed to the USDA soil classification system. However, the methodology proposed could be used more generally, and the information provided is also applicable when dealing with the soil texture classification systems used in other countries.

Improving texture categorization with biologically-inspired filtering

Within the domain of texture classification, a lot of effort has been spent on local descriptors, leading to many powerful algorithms. However, preprocessing techniques have received much less attention despite their important potential for improving the overall classification performance. We address this question by proposing a novel, simple, yet very powerful biologically-inspired filtering (BF) which simulates the performance of human retina. In the proposed approach, given a texture image, after applying a difference of Gaussian (DoG) filter to detect the edges, we first split the filtered image into two maps alongside the sides of its edges. The feature extraction step is then carried out on the two maps instead of the input image. Our algorithm has several advantages such as simplicity, robustness to illumination and noise, and discriminative power. Experimental results on three large texture databases show that with an extremely low computational cost, the proposed method improves significantly the performance of many texture classification systems, notably in noisy environments.The source codes of the proposed algorithm can be downloaded from https://sites.google.com/site/nsonvu/code.

Classifying Similarity and Defect Fabric Textures based on GLCM and Binary Pattern Schemes

Textures are one of the basic features in visual searching,computational vision and also a general property of any surface having ambiguity. This paper presents a texture classification system which has high tolerance against illumination variation. A Gray Level Co-occurrence Matrix (GLCM) and binary pattern based automated similarity identification and defect detection model is presented. Different features are calculated from both GLCM and binary patterns (LBP, LLBP, and SLBP). Then a new rotation-invariant, scale invariant steerable decomposition filter is applied to filter the four orientation sub bands of the image. The experimental results are evaluated and a comparative analysis has been performed for the four different feature types. Finally the texture is classified by different classifiers (PNN, K-NN and SVM) and the classification performance of each classifier is compared. The experimental results have shown that the proposed method produces more accuracy and better classification accuracy over other methods.

Particle size distribution models for soils of the humid tropics

Standardisation of particle size distribution (PSD) is a prerequisite to achieve compatibility of soil data among various countries with different texture classification systems. Therefore, several mathematical models have been proposed to accurately represent PSD. Previous studies evaluated the performance of such models to describe PSD of soils from temperate regions. This study aims at evaluating the performance of models for describing PSD of soils from the humid tropics based on a large dataset. A dataset of 1,412 soils from Central Africa representing 11 different FAO Soil Groups was used. Ten PSD models with two to four fitting parameters were selected: simple log-normal (LN_2p), van Genuchten-type1 (VG_2p), van Genuchten-type2 (vG_3p), Fredlund-type1 (F_3p), Fredlund-type2 (F_4p), Weibull (W_3p), Skaggs (Sk_3p), Gompertz-type1 (G_2p), Gompertz-type2 (G_4p) and Andersson (A_4p). The fitting performance of the PSD models was evaluated by three statistical indices: the adjusted coefficient of determination, the Akaike information criterion and the relative error. Clustered columns and box plots were also used to get more insights. The predictive ability of the best PSD models was tested using a leave-one-out method and 1:1 plots. A table of initial values for the fitting parameters of each PSD model was provided for future applications. Some models like VG_2p, VG_3p, Sk_3p and G_4p were not suitable to describe PSD of soils in the humid tropics. On the other hand, F_3p, F_4p, W_3p and A_4p models showed outstanding fitting performance. The fitting performance of PSD models was also dependent of the textural class, the broad textural group and the bimodal character of the soil. For the most frequent textural classes in the dataset, the F_3p and A_4p models were the best closely followed by the W_3p model. While the F_3p model performed better than the A_4p model for coarse-textured soils, the opposite was observed for fine-textured soils. The W_3p model showed acceptable fitting performance for fine, medium and coarse-textured soils. The performance of the PSD models was found to be better for bimodal soils, which are common in the humid tropics, than for unimodal soils. Great differences in fitting and prediction performance were found between the PSD models. Soil texture as well as the bimodal character of the soil significantly affect their respective performance. Some models like VG_2p, VG_3p, Sk_3p and G_4p are not suitable to describe PSD of soils of the humid tropics. On the other hand, F_3p, F_4p, W_3p and A_4p models showed outstanding fitting performance. Therefore, they are highly recommended in order to get a better description of the PSD of soils of the humid tropics.

조명 및 카메라 이동속도가 토양 영상에 미치는 영향

Soil texture has an important influence on agriculture such as crop selection, movement of nutrient and water, soil electrical conductivity, and crop growth. Conventionally, soil texture has been determined in the laboratory using pipette and hydrometer methods requiring significant amount of time, labor, and cost. Recently, in-situ soil texture classification systems using optical diffuse reflectometry or mechanical resistance have been reported, especially for precision agriculture that needs more data than conventional agriculture. This paper is a part of overall research to develop an in-situ soil texture classification system using image processing. Issues investigated in this study were effects of sensor travel speed and light source and intensity on image quality. When travel speed of image sensor increased from 0 to 10 mm/s, travel distance and number of pixel were increased to 3.30 mm and 9.4, respectively. This travel distances were not negligible even at a speed of 2 mm/s (i.e., 0.66 mm and 1.4), and image degradation was significant. Tests for effects of illumination intensity showed that 7 to 11 Lux seemed a good condition minimizing shade and reflection. When soil water content increased, illumination intensity should be greater to compensate decrease in brightness. Results of the paper would be useful for construction, test, and application of the sensor.

FPGA Implementation of Generalized Hebbian Algorithm for Texture Classification

This paper presents a novel hardware architecture for principal component analysis. The architecture is based on the Generalized Hebbian Algorithm (GHA) because of its simplicity and effectiveness. The architecture is separated into three portions: the weight vector updating unit, the principal computation unit and the memory unit. In the weight vector updating unit, the computation of different synaptic weight vectors shares the same circuit for reducing the area costs. To show the effectiveness of the circuit, a texture classification system based on the proposed architecture is physically implemented by Field Programmable Gate Array (FPGA). It is embedded in a System-On-Programmable-Chip (SOPC) platform for performance measurement. Experimental results show that the proposed architecture is an efficient design for attaining both high speed performance and low area costs.

Influence of no tillage controlled traffic system on soil physical properties in double cropping area of North China plain

An experiment was conducted to determine the effects of tillage on soil properties in the field of maize ( Zea mays L.) and winter wheat ( Triticum aestivum L.) annual double cropping region in North China Plain. Measurements were made following six years (2005 to 2010) of three tillage treatments; no till with controlled traffic (NTCT), no till random trafficking (NTRT) and conventional tillage (CT) on a silt loam according to the USDA texture classification system soil in Daxing district, which lies in the suburb of Beijing. Long term no till with controlled traffic significantly (P < 0.05) increased macro-aggregates, infiltration rate, soil moisture, together with reductions in soil bulk density, soil compaction in different layers compared with the no till random traffic and traditional mould board tillage treatment currently used in this region. Consequently, mean winter wheat and summer maize yields for the NTCT treatment were improved by 2.8 and 7.1% when compared with the soils under no till random traffic, while huge improvement was found when it was compared with conventional ploughing management (4.2 and 12.08% for wheat and maize, respectively). The long-term experiment demonstrated that no-tillage controlled traffic with residues retained, offers a potentially significant improvement over the current farming systems in annual double cropping areas of North China Plain. Key words : No tillage, controlled traffic, soil physical properties, North China Plain.

Efficient hardware architecture based on generalized Hebbian algorithm for texture classification

The objective of this paper is to present an efficient hardware architecture for generalized Hebbian algorithm (GHA). In the architecture, the principal component computation and weight vector updating of the GHA are operated in parallel, so that the throughput of the circuit can be significantly enhanced. In addition, the weight vector updating process is separated into a number of stages for lowering area costs and increasing computational speed. To show the effectiveness of the circuit, a texture classification system based on the proposed architecture is designed. It is embedded in a system-on-programmable-chip (SOPC) platform for physical performance measurement. Experimental results show that the proposed architecture is an efficient design for attaining both high speed performance and low area costs.

Noise- and compression-robust biological features for texture classification

Texture classification is an important aspect of many digital image processing applications such as surface inspection, content-based image retrieval, and biomedical image analysis. However, noise and compression artifacts in images cause problems for most texture analysis methods. This paper proposes the use of features based on the human visual system for texture classification using a semisupervised, hierarchical approach. The texture feature consists of responses of cells which are found in the visual cortex of higher primates. Classification experiments on different texture libraries indicate that the proposed features obtain a very high classification near 97%. In contrast to other well-established texture analysis methods, the experiments indicate that the proposed features are more robust to various levels of speckle and Gaussian noise. Furthermore, we show that the classification rate of the textures using the presented biologically inspired features is hardly affected by image compression techniques.

Assessment of texture measures susceptibility to noise in conventional and contrast enhanced computed tomography lung tumour images

Noise is one of the major problems that hinder an effective texture analysis of disease in medical images, which may cause variability in the reported diagnosis. In this paper seven texture measurement methods (two wavelet, two model and three statistical based) were applied to investigate their susceptibility to subtle noise caused by acquisition and reconstruction deficiencies in computed tomography (CT) images. Features of lung tumours were extracted from two different conventional and contrast enhanced CT image data-sets under filtered and noisy conditions. When measuring the noise in the background open-air region of the analysed CT images, noise of Gaussian and Rayleigh distributions with varying mean and variance was encountered, and Fishers’ distance was used to differentiate between an original extracted lung tumour region of interest (ROI) with the filtered and noisy reconstructed versions. It was determined that the wavelet packet (WP) and fractal dimension measures were the least affected, while the Gaussian Markov random field, run-length and co-occurrence matrices were the most affected by noise. Depending on the selected ROI size, it was concluded that texture measures with fewer extracted features can decrease susceptibility to noise, with the WP and the Gabor filter having a stable performance in both filtered and noisy CT versions and for both data-sets. Knowing how robust each texture measure under noise presence is can assist physicians using an automated lung texture classification system in choosing the appropriate feature extraction algorithm for a more accurate diagnosis.