Study On Caragana Stenophylla Pojark Leaf Anatomical Structure Analysis Using Image Processing Technology

Grassland degradation is one of the main manifestations of grassland desertification,which is a retrogressive succession process of grassland ecosystem caused by human activities or adverse natural conditions.In recent years,due to the impact of climate change,human activities and other factors,grassland degradation is very critical,and the degradation trend is increasingly,so the prevention and control of grassland degradation become urgent.With the rapid development of the field of life science,and the deepening of plant anatomy,the study of the relationship between the anatomical structure of plants and the degeneration of grassland is increasing.With the rapid development of the field of life science,and the deepening of plant anatomy,the study of the relationship between the anatomical structure of plants and the degeneration of grassland is increasing.However,the traditional methods of acquiring the anatomical information is relatively backward,which is also with the higher requirements for the professional of experimental personnel,and the heavier testing works,as well as the poor consistency and reproducibility of the results,therefore the development of this field is limited.Aiming at the above problems,in this paper,the desert grassland of Ordos is chosen as the experimental area,divided different degradation gradient accoding according to field investigation and expert experience.The typical plant Caragana microphylla was selected as the test object and the leaves were sliced.Analysed the Caragana Stenophylla Pojark Leaf Anatomical structure with the image processing technology,studied and improved the related image processing algorithms,the correlation between the structural indexes of extraction and measurement and the degradation gradient of grassland were analyzed,the main research works are as follows:1.In the year 2014-2016,the degraded gradient samples were surveyed in July and August of each year to obtain the basic information of the grassland in the experimental area.Caragana Stenophylla Pojark was used as the follow-up experimental plant.According to the requirements of ecology and statistics,the leaves of Caragana microphylla were sampled and sliced,and all anatomical indexes were captured and measured by computer,microscope and the Touptek Toupview software;2.Based on the image characteristics of Caragana Stenophylla Pojark leaf structure,the pretreatment,segmentation and feature extraction of images of Caragana Stenophylla Pojark leaf anatomical structure were studied,extracted some characteristic parameters of leaf anatomical structures.The specific process and the results are analyzed as follows:?1?Image preprocessinga.Compared the effects of linear transform of gray,nonlinear transformation,histogram equalization,contrast limited adaptive histgram equalization and histogram specification on image enhancement,the results show that the contrast limited adaptive histgram equalization has obvious effect on image enhancement and higher degree of automation;b.According to the principle of homomorphic filtering,the traditional dynamic Butterworth filter is improved.Since the cut-off frequency D₀ of traditional homomorphic filter is often chosen by a large number of practice,the effect is poor.By controlling the filter function slope equation D?u,v?frequency domain to take the value,improved homomorphic filtering quickly determine the value to achieve the best brightness uneven correction effect.Compared with the traditional homomorphic filtering method,background subtraction and bottoming transformation processing results,visual effect and the brightness unevenness of the image are obviously improved.Subsequently,applying improved high frequency emphasis Butterworth filter to image sharpening,compared with Prewitt operator,Sobel operator and Laplacian,used improved high frequency emphasis Butterworth operator,image sharpening effect is better,the image details and edges are very clear,not sensitive to noise;c.A new image denoising method was presented combined blind denoising algorithm.Firstly,used the squared reconstruction method and fitting algorithm to estimate the image noise type of Caragana Stenophylla.Compared with the traditional method,the proposed algorithm can accurately evaluate the noise type and guide the selection of the subsequent image denoising method.Then,the SVD domain image noise estimation method based on block strategy is used to estimate the noise intensity.After image segmentation,the SVD domain decomposition calculation is reduced significantly,and the precision,speed and stability were higher.Finally,combined geometric mean filtering?GMF?and block-matching and 3-D filtering?BM3D?denoising method,achieved image denoising accurately.It can not only remove the noise well,but also can keep the image edge,texture and other details.Compared with the traditional BM3D?block-matching and 3-D filtering,BM3D?algorithm,the denoising effect is equivalent,but it takes only about 1/9 of the BM3D algorithm.It was better than WT?wavelet threshold?denoising in BRISQUE,which was lowered by about 4.The image quality is improved obviously.BRISQUE of the processed image was 10,equivalent to half of that of the original image.So,with the noise type estimation-intensity estimation-denoising process and algorithm in this paper obtained good results for the image denoising of Caragana microphylla.?2?Image segmentationa.Using the maximum interclass variance?Otsu?and iteration,combined with the mathematical morphology,realized the partial segmentation of leaf slices.Compared with the result of single algorithm,the precision of segmentation was improved obviously,the average segmentation error is 0.765,the over-division error is 0.054,and the under-division error is 0.502;b.Introduced the binary value K-means clustering method with maximum value as clustering center based on traditional K-means,overcomed the shortcomings of K-means clustering algorithm and FCM clustering algorithm,improved the clustering speed,and achieve the xylem cells accurate and efficient division.Compared with the results of traditional K-means clustering and fuaay C-means clustering?FCM?algorithm,the segmentation accuracy was similar,but the running speed improved obviously.When the number of clusters was 2,the average time was 3.523s.When the number of clusters was3,the average time was 5.112s.With the improved pitting method,the watershed metho d and the proposed ring structure extraction method to separate the adherent xylem cells after the initial separation.Compared the effect of the separation,the circular structure extraction method was more accurate and faster for the adhesion of the xyle m cells;c.Using the Lazy Snapping algorithm,the GrabCut algorithm and the active contour algorithm based on the variational level set to segment the epidermal cells,compared with other interactive segmentation algorithms,the results showed that the active contour algorithm based on variational level sets took 30.228s averagely,but the operation speed was relatively slow.However,the accuracy of segmentation for epidermal cells was high,the average segmentation error was 5.670,the over-division error was 0.589,and the under-segmentation error was 5.900.?3?Feature extractiona.According to the characteristics of interactive measurement of structural indicators,a leaf width?K?feature extraction algorithm was proposed.Applying the fast skeleton extraction algorithm to extract the skeleton of the whole part of the segmented leaf slice,using the Gaussian curve fitting skeleton,we determined intersection point between the skeleton fitting curve and the edge by the minimum distance method,and the distance between the two intersections was K.Compared with the traditional skeleton extraction algorithm,the results show that the fast-moving skeleton extraction algorithm eliminated the defects such as burr,bifurcation and multi-pixel width under the premise of ensuring connectivity.Compared with the polynomial curve,the Gaussian curve was more accurate.The average root mean square error of the frame was only 0.796 9,and the average root mean square error of the edge fitting was only 0.781 4.Compared with the simultaneous equation,the minimum distance was easy to calculate and run fast.The average relative error of the algorithm was 0.42%,the mean square error was 3.064 5,the average time was 4.31s,and the variance was 0.322 9;b.Using improved pouch matching method to extract feature of leaf thickness?H?,compared with interactive measurement of mean,the average relative error of the algorithm was 1.49%,the mean square error was 4.611 5,the average time was 2.83s,and the variance was 0.083 1;c.Calculating the number of rectangular boxes after the segmentation algorithm,we realized the xylem cell count.Compared with the mean value of the interaction,the average relative error of the algorithm was 0.67%,the root mean square error was 0.5352,the average time was 0.14s,and the variance was 0.322 9;d.Using extraction algorithm based on the fast-moving skeleton and Sobel operator,the epidermal cytoskeleton and the edge were extracted.We determined intersection point between the local Gaussian fitting center vertical line and the edge by the minimum distance method.After continuous iteration,the epidermal cells Thickness was measured.Compared with the mean value of the interaction,the average relative error of the algorithm was 3.27%,the mean square error was 0.982 6,the average time was 19.36s,and the variance was 0.007 4.3.Analyzed the results of image processing and partial structural indicators with statistics.In the anatomical structure of Caragana microphylla,without the size of Vascular bundle sheath cell?WGQ?,the thickness of phloem fiber?RPX?,the thickness of phloem?RPB?and the number of rosette?HHG?,the remaining 11 structural indexes were significantly different with the degree of grassland degradation by single factor analysis of variance.Based on the correlation analysis between the 11 significant structural indexes and the grassland wetting coefficient,the results showed that the leaf width?K?was positively correlated with the wetting coefficient,and the other 11 indexes were negative.By analyzing the 10 structural indexes except the xylem?MZG?with the corresponding analysis method,the results showed that the leaf width?K?was larger,while the rest of the structure dimensions were smaller and the number of xylem cells?MZG?was relatively smaller in the condition of comparison?CK?and mild degradation?LD?than in others.The thichness of leaf?H?,the thickness of adaxial/abaxial palisade tissue?XBZ?SBZ?;the diameter of upper/lower epidermis?SBX?XBX?,the thickness of upper/lower epidermal cell wall?SBB,XBB?,the thickness of upper/lower epidermal stratum corneum?SBJ,XBJ?were larger,MZG relatively more under the condition of moderately degraded?MD?and heavy degraded?HD?.The application of image processing technology to extract the characteristics of Caragana Stenophylla Pojark leaf anatomical structure which have strong correlations with grassland degradation had the advantages of high precision,good consistency,high reproducibility and high efficiency.The stage results that we have obtained in this paper can provide theoretical basis and technical support for the conservation and utilization of Caragana Stenophylla Pojark,the degradation and control of desert grassland and plant anatomy.