Research On The Missing Detecting Method Of Mask Belts Based On Machine Vision

Masks are common in air filtration products.In recent years,with the aggravation of the air pollution,the national attention to personal safety protection and the incidence of pneumoconiosis and other occupational diseases increased,the market space of masks is becoming biger and biger.However,in the production process of masks,there is often a missing problem of mask belts.If such problem p roducts fl ow into the market,it will affect the product image of the enterprise.At present,the detection of mask belt loss is artificial.It is time-consuming and laborious.Therefore,the use of machine vision method for the missing detection of mask belts has important application value,In the research,the key part is the attitude correction of the mask,the extraction of the feature of the mask and the training of the classifier.In view of these questions,this article launches the related research and the experimental analysis,the main research achievement is as follows:First need to correct the mask image normalization.Contrast and select the appropriate threshold processing algorithm,use threshold segmentation to the mask image,segment masks from background.Using morphological processing to generate two value image masks do corrosion expansion,removing holes and noise.Using Canny edge detection to extract the outline of the mask.Than do the the minimum external rectangle calculation to the mask contour,get the mask of the smallest external rectangular length and width and tilt angle,calculate the rotation of the affine transformation matrix.Using the matrix,the original mask image or pre processed mask image do attitude correction.Secondly,according to the characteristics of the texture features of the mask,the LBP and HOG features are used in the research.TO the different sizes of the original image and the two value of the enhanced processing,we use a variety of patterns and size of the LBP operator to extract features,and the different size of the test set to extract the HOG features.Finally,according to the type of real data sets and the resear-ch question,this paper chooses the support vector machine(SVM)as classification modell,for missing detection of mask belts,for the extracted LBP and hog feature sets training classification.Use the classifier to predict if the mask belt is missing.This topic using Microsoft Visual Studio 2013 integrated development environment,combined with OpenCV image processing library,developed a missing detection system of mask belts based on machine vision.Use the system to verify the effectiveness and robustness of the attitude correction algorithm.The parameters of the set data and the feature are discussed with experiments.The validity of the extracted feature set and the trained classifier is tested by 5-fold cross validation.