Research On Jujube Texture Detection Based On OpenCV

On the basis of summarizing the domestic and foreign research on texture detection,combined with the development of jujube industry in Xinjiang,a research subject about jujube texture detection was proposed based on the Open CV.As one of the geographical indications of Xinjiang,Hami jujube is able to grow in harsh environments and rich in nutrients.This paper takes Hami jujube as the research object,using Open CV image processing library as the core,combined with digital image processing technology,based on the external surface texture features of Hami jujube to determine classification.At the same time,this subject carried on the algorithm research through the MATLAB software,designed the digital image processing system based on OpenCV,the interface of which is friendly and it can satisfy the need of some digital image processing,lay the foundation for the following research.Jujube texture detection based on Open CV is helpful to improve the texture classification efficiency of Hami jujube appearance quality,liberating human and reducing the time from harvest packaging to mass market.It also can guarantee the quality of Hami jujube and enhance the market competitiveness.The main results of this paper are as follows:1?The Hami jujube image was denoised and enhanced by using the discrete wavelet transform.Single scale two-dimensional discrete wavelet transform was used on single scale decomposition and reconstruction of Hami jujube images.By it low frequency signal and high frequency signal can be got.Low frequency represents contour,high frequency responses details and mixed noise.By enhancing the low frequency part and weakening the high frequency part to enhance the low frequency and suppress the high frequency,achieving the purpose of denoising and enhancing Hami jujube image.2?The six parameters were normalized between 0 and 1,which are the maximum probability,correlation,contrast,energy,homogeneity and entropy.The results were used as input vectors of BP neural network and ANFIS.Three kinds of training methods of BP neural network were compared,which were gradient descent method,Quasi-newton method and conjugate gradient method.A model was established,by using the ANFIS to evaluate the external texture quality of Hami jujube,and threshold the prediction results.The results show that Quasi-newton method training single hidden layers BP Neural Network has the fastest convergence rate.The Quasi-newton method training double hidden layers BP Neural Network had a higher accuracy then the former,which is 89.66%.The ANFIS algorithm had an accuracy of 93.33%.3?Extracting the contour of Hami jujube binary images by means of morphology of corrosion and expansion.Using the circular structure with a diameter of 50 pixels can meet the requirements of the contour map.The centroid of the contour is extracted and labeled.In order to get the connected domain on central area of jujube and eliminate the edge effect,subtracted the contour map form binary map and then used open operation.Meanwhile,the centroid of each connected domain was signed.4?In order to evaluate Hami dried dates external quality,two different algorithms were proposed to quantitatively calculate the connected domain density,which were different form destructive mechanical grading.The first algorithm is taking the centroid of Hami dried dates as coordinate system origin and the second is taking the averages of horizontal and vertical coordinates of each connected domain center respectively as coordinate system origin.The connected domain was obtained by arithmetic of substr act between binary image and contour map of Hami dried dates and morphological operation.By using optimized SVM modeling,verify the accuracy of the two algorithms.The results show that the second algorithm is the best prediction model,and its prediction accuracy is 93%.5?Digital image processing system was developed by Qt and OpenCV and some basic image processing functions were implemented.As well as,it can carry out functional development and add-on functions' development.The OpenCV function library was used in the dynamic tracking of jujube.By the sum of row and column of grayscale,find the smallest position and with 10×10 pixels black square mark to lock the jujube position for dynamic tracking.