Research On Plastic Color Matching Expert System Based On K-Nearest Neighbor Algorithm Ensemble Learning

Plastic with its light,cheap,superior performance and other characteristics,has become the highest utilization rate of materials in modern life.Plastic coloring can not only make plastic products get bright color,meet people’s needs,but also give new properties to materials.The key of plastic coloring lies in color matching,only the color configuration is appropriate,the color and lustre of products can be suitable.Under the current situation,plastic color matching depends on color matching staff according to experience,which not only requires color matching staff to have rich experience accumulation,professional technical knowledge,but also need good visual color recognition ability.Therefore,plastic manufacturers hope to find a new color matching method.In this paper,machine learning method is adopted to recommend the pigment formula,using the LAB value of the target color as input,and the amount of each pigment as output formula.In this paper,the effects of k-means,FCM and LI＿BIFT clustering algorithms on pigment selection were compared,and it was found that it was difficult to divide the formula classification accurately by one clustering method.The method of secondary clustering was used to further divide the data to improve the accuracy of the formula classification.In this paper,radial basis function(RBF)neural network and back propagation(BP)neural network are respectively used to predict LAB values in the learning process.KNN(K nearest neighbor)algorithm is used to classify the data,and then integrated neural network is used to train the data.This paper introduces the theory and results of multi-expert decision fusion and fuzzy decision making,and compares the final results.Multi-expert decision fusion can effectively reduce the misjudgment after clustering,while fuzzy decision is helpful to solve the indecision in decision,which can provide ideas for the colorist to correct the formula by suggesting a variety of possible data,and effectively reduce the workload of the colorist.Finally,this paper adopts KNN classifier integrated neural network and multi-expert fuzzy decision making to verify the effect in actual production.It is found that it can effectively help the colorist to complete the color matching process of the target color after three trial matches,and effectively improve the efficiency of plastic color matching in actual production.