Research On Measuring Method Of Surface Roughness For Speckle Image

The surface roughness is an important parameter to measure the quality of the surface of the part,which affects its operational stability,wear resistance and service life.Therefore,it has become a research hotspot to accurately evaluate the surface roughness.Analysis of existing research found that these studies need to establish different corresponding models for different types of parts,and the specific types of features in the specific mathematical model are single.In order to establish a roughness detection model suitable for different types of parts,this paper proposes a set of parts surface roughness detection scheme based on MPML.First,preprocess the collected images and extract the-multi-features of the-processed images.Then,improve the classic RF classification algorithm,and apply the improved A-RF classification algorithm to the part process type identification.Finally,the multi-feature descriptors of this paper are constructed,and the roughness learning function is constructed with the classification parameters to achieve the roughness detection of the parts.The main contents of this paper are as follows:(1)Firstly,the speckle image is collected,and the influence of the laser incident angle during the acquisition process is analyzed to select the best angle,and the noise and dark areas in the collected image are preprocessed to improve the image quality.(2)Research on part process identification method based on A-RF algorithm.Firstly,the features of the image are extracted,and the A-RF classification algorithm is proposed by reducing the weak classifiers with high correlation and low classification accuracy in the classical RF model.Then the extracted image features are trained in the A-RF classification model.(3)Research on the Measurement of Part Surface Roughness Based on MPML Algorithm.First,a multi-feature descriptor is constructed by introducing the mutual information,which aiming at the single feature problem in the existing research.And the MPML model function is constructed by adding the classification parameters to the problems that need to be trained in the existing research.(4)Design a series of experiments to verify the validity and accuracy of this article.By comparing the results of the classification experiment,it is concluded that the A-RF algorithm has a high classification accuracy,and the algorithm parameters are compared and optimized.The effectiveness of the proposed method is verified by comparing the experimental results between MPML roughness model and the roughness model established by single features,which has low relative error rate and high accuracy.