Research On The Roughness Recognition Based On Colour Distribution Statistical Matrix And Variable Predictive Model Based Class Discriminate

The surface roughness has a great influence on the performance and life of the mechanical parts. With the development of image processing technology, mechine vision has become the part of the industrial automation test which should not be overlooked. The research on roughness recognition based on vision has a vital theoretical and practical significance. However, there are two problems for rough recognition technology, one is the image quality index to measure the roughness of workpiece; the other is the pattern recognition methods to complete the identification of roughness.Until now, there has not been a standard relationship to describe the connectionbetween image quality indexes and roughness. The method widely used is to shootthe workpiece surface texture firstly and calculate the gray level co-occurrencematrix of the pictures whose matrix index could complete the identification ofroughness. However, the gray level co-occurrence matrix algorithm istime-consuming and parameters are very difficult to control. In terms of patternrecognition, the methods commonly used are neural network and support vectormachine which have also some limitations and do not make full use of the internalrelations of the characteristic variables of the image in each category.Colour distribution statistical matrix is usually using the custom colors of light sources irradiation on the workpiece surface in imaging procedure and the statistical matrix is obtained by calculating out the luminance information of the red and green components from the images. While informations contained in the matrix are various due to the different reflection performance of the rough surface. In recent years, Variable Predictive Model Based Class Discriminate(VPMCD), as a new pattern recognition algorithm, is used to realize the classification. It exploits the mutual relationship of each feature and establishes their corresponding Variable Predictive Model(VPM) respectivelyUnder the support of the National Natural Science Foundation of China(Grant No. 71271078), this thesis has taken deeply and systematically research on roughness classification based on machine vision using the index of the color distribution statistic matrix combining VPMCD method. The main research contents and innovation points are as follows:(1)Since it is proposed,color distribution statistics matrix have been proved to have good advantage on roughness measurement, but the single index used can’t express the large image information which would bring out an inaccurate measurement. This thesis designs five matrix indexs for the color distribution statistics matrix which include the number of nonzero value, contrast, homogeneity, information entropy and energy respectively. The thesis presented the theory foundation of these indexs and the results showed that it was feasible to represent the roughness using the matrix indexs.(2)The thesis studied on the basic principle and specific algorithm of VPMCD pattern recognition. According to the correlations of the color image statistical distribution matrix indexs, this thesis built VPM models for the different categories and tested their effectiveness in order to find the relationship between the matrix indexs by using the proposed experimental samples. The experimental results showed that VPMCD method was efficient and feasible to indentify the sample pieces with different roughness.(3)Lab VIEW has the advantages of high integration and it is convenient to design the interface. Matlab has the powerful function of matrix calculation. Therefore, combined with the advantages of Lab VIEW and MATLAB platform, this thesis designed and implemented an offline equipment that could classify the surface roughness level. First of all, pictures were collected through the Lab VIEW. Then the image features were extracted and the roughness category classification was completed through the Matlab. At last, the results would been displayed based on the Lab VIEW. In order to have an objective understanding of the equipment, this thesis selected accuracy and response time to complete the preliminary evaluation of the equipment based on the 1094-2002 JJF standard.Based on the platform of Lab VIEW and MATLAB, a roughness recognition device based on machine vision was developed. The device could calculate the interest part of the capture images’ s color distribution statistic matrix and obtain the matrix indexes. Combined with the image indexes and VPMCD method, the roughness recognition of the workpiece surface will be realized.