Research On Detection Technology Of Small Impurities In Transparent Liquid

In the production line of transparent liquid,if there are impurities in the liquid,it will seriously affect the quality of products and may cause great harm to our health.Compared with traditional manual detection,liquid impurity detection based on image processing technology has the advantages of fast speed,high accuracy and low cost.In the process of liquid impurity detection based on image processing technology,the components of the liquid image can be divided into five categories: impurities,bubbles,noise,background and bottle body trace.The influence of background on impurity detection is relatively small,the bottle body trace is easy to eliminate,and the remaining noise and bubbles are the two difficulties of liquid impurity detection.And the existing liquid impurity detection algorithm cannot effectively eliminate the influence of bubbles on the detection results,especially when a large number of bubbles exist.To solve this problem,this paper presents a feature-based classification algorithm for detecting impurities in liquids,which can suppress noise and distinguish impurities from bubbles in the presence of a large number of bubbles in the image of liquid impurities.The main works of this paper are as follows:Firstly,an image acquisition system for liquid impurity detection is designed to achieve high quality liquid impurity image acquisition.At the same time,aiming at the problem of image noise,we compared the collected images with various pretreatment methods,and selected bilateral filtering which can remove noise and keep edge as the pretreatment algorithm of liquid impurity detection.Secondly,aiming at the problem of liquid impurity image edge detection,the multi-scale wavelet transform edge detection algorithm is deeply analyzed and optimized by the method of double threshold.By choosing high and low thresholds in edge pixel determination and connecting edge pixels,the pseudo-edge pixels caused by noise and color change are effectively suppressed,and the positioning accuracy and contour clarity of edges are improvedThirdly,aiming at the whole process of liquid impurity detection,a liquid impurity detection algorithm based on feature classification is proposed.The original image is preprocessed by bilateral filtering,and the optimized multi-scale wavelet transform is used to detect the edge of the target.Then the roundness,average gray level and variance of the target feature are extracted.Finally,the three features are combined to distinguish the impurities and bubbles in the liquid,so as to eliminate the influence of bubbles on the detection results and obtain the final impurities.The proposed algorithm effectively suppresses noise and eliminates the influence of bubbles on impurity detection results.In the complex situation with a large number of bubbles,the detection accuracy is significantly improved.