Research On Performance Variability Analysis And Simulation Of Cloud Network For Big Data Application

Manual detection of metal parts may cause defects such as easy to scratch the surface of metal parts,low efficiency and instability.Modern detection technology based on machine vision has advantages of stability,real-time and speed.In order to ensure that the industrial camera will not affect the smooth operation of the production line,its construction of hardware needs to be far away from the object to be detected,which reduces the image resolution of the object to be identified.In thesis,the image interpolation technology will be used to improve the detection accuracy of the metal part image captured by the industrial camera,and then the active contour detection method will be used to extract the image contour.In this way,the area of the object to be detected can be obtained in a non-contact way,so that it can be used in machine learning applications such as target location and defect detection.The main contents of thesis are as follows:Firstly,an improved image interpolation method based on edge information is proposed for image preprocessing.Using the existing interpolation technology to simulate the industrial image,it is found that there are some problems such as blur and sawtooth after image interpolation,which is not conducive to the research of contour detection in this subject.In order to make up for the average problem of undifferentiated gray information brought by the traditional interpolation technology,the image edge is obtained to distinguish its region.The edge direction information of the points in the edge region is calculated according to the adjacent pixels with different positions and distances around,and the points with determined directions are interpolated in different ways.After experimental verification and analysis of non-edge regions,linear interpolation method is adopted,and finally an improved image interpolation method based on edge information is obtained.Then for contour detection,an improved active contour model method based on local information and global information is proposed.The existing technology is used to interpolate the traditional image of industrial images.The experimental simulation results show that the traditional contour detection method is not universal for metal parts detection.Considering the particularity of metal part contour detection,this thesis constructs an improved contour detection model.Taking advantage of the SPF in model SBGFRLS,the SPF is improved to represent the local and global information in the image.Therefore,the constructed function can not only process the image with uneven gray distribution,but also make it not affected by the size of the initial contour position.At the same time,the weight coefficient is constructed to weigh the relationship between the two kinds of information,an improved active contour model based on local information and global information is obtained.Finally,experimental simulation is carried out based on the above improved model.By testing the improved methods,the indexes show that the improved interpolation method can effectively weaken the influence of background on contour detection and highlight the edge information of contour;The improved contour method effectively reduces the interference of texture noise,and the detection results are clear and complete.The improved contour detection method based on image interpolation has improved the overall performance index results compared with the detection method without interpolation,and the processing effect has been significantly improved,accounting for up to 98%.The utility model solves the problem of incomplete image processing in the prior art,and is better applicable to the contour detection of metal parts in industrial production.