Research Of Technology On Surface Defects Detection For Fireproof Boards Based On Machine Vision

Fireproof board,as a refractory building material for surface decoration,is widely used in the fields of architectural engineering,home decoration and so on.With the development of multi-style,stereoscopic and light type,the surface appearance quality of the fireproof boards is also concerned.The surface quality of fireproof boards will not only affect the surface beauty,but also further affect the overall quality.Therefore,we must strengthen the detection the surface quality of fireproof boards.At present,the detection of the surface quality of fireproof boards is mainly manual observation and inspection,which has great limitations,low detection efficiency and high false detection rate.Therefore,the research of efficient and accurate inspection technology for the surface quality of fireproofing boards has become the focus of enterprises.This paper studies a fireproofing boards surface quality inspection technology based on machine vision,analyzes and compares four common surface defect images of fireproof boards including bare substrate defect,dirt defect,frosting defect and impregnated paper crack defect from the fireproofing boards of Shandong Penglai Shenghua Industry.In view of this,this paper proposes an improved PCNN(Pulse Coupled Neural Network)algorithm for image segmentation,and BP(BackPropagation)neural network for the classification and identification of defects,can accurately detect defects and classification.The specific research content of this article includes:(1)The fireproof board surface defect detection system was designed.The hardware system mainly includes the selection of industrial camera,lens,lighting equipment and lighting mode.The software system mainly includes software configuration,interface design,image preprocessing,image segmentation,image feature analysis,image classification and recognition.(2)For the characteristics of four kinds of fireproof board surface defects,gray histogram equalization and median filtering methods are used to preprocess the image,which effectively enhances the contrast of the image and retains more useful information.An improved PCNN image segmentation method based on the minimum Tsallis cross-entropy is proposed and compared with the traditional image segmentation methods.The experimental results show that the proposed algorithm improves the self-adaptiveness of image segmentation.The segmented image has higher regional contrast parameters and regional consistency parameters and can effectively segment four defect images.(3)The surface quality images of a large number of segmented fireproof boards were analyzed,and the surface feature parameters were extracted,including five geometric feature parameters,six gray feature parameters and five kinds of texture feature parameters.The extracted image feature parameters are listed,and it can be seen that the selected image feature parameters are representative and can effectively reflect the features of the fireproof board surface defect images.(4)In order to solve the problem that the traditional BP neural network is easy to fall into a local minimum and slow convergence,a BP neural network is optimized by a genetic algorithm(GA-Genetic Algorithm),a GA-BP neural network is constructed.The extracted 16 feature parameters are used as the input of the GA-BP neural network,and the GA-BP neural network parameters are set reasonably.After training,the global optimal solution can be quickly obtained,and the number of iterations of the algorithm is greatly reduced,and the accuracy of the objective function is higher.