Sugar Crystallization Status Detection And Analysis Based On Image Processing In Sugar Process

At present,the crystallization status in sugar process is often determined by operator’s observation.A lot of sugar mill judge by sample drawn out of material falling on the grass,and make observation of crystal size,content of mother liquor,absorption degree in the light and feelings of hand twisting.Through determining whether the current vacuum,temperature and degree of hammer of the crystal is in a good condition,then adjust the intake or into the water.This artificial sampling mode hinders the realization of automation control in the sugar boiling process.Aimed to tackle this problem,some key technologies of image processing and recognition for the sugar crystal of automatic sampling was deeply discussed in this dissertation.Theresearch work was organized as follows:(1)The device for automatic sampling and image formation of crystals was designed.It could take the crystal samples from the sugar pan and flat then out into a layer automatically,the observation platform will be automatically cleaned after imaging.(2)Image preprocessing was conducted based on gray level、transformation,PCNN algorithm for the noise adaptive median filtering and morphological optimization method.As to the mingled particles,a method for watershed optimization was put forward,which would separate the mingled particles into individual ones.At first,picked out the adhesion particles and area of threshold according to the shape factor,for adhesion particles distance transform to calculate the extremum of the seed point,then merged redundancy seed points,Watershed segmentation on adhesion particles.Finally,superimposed on individual particles.The results showed that the optimization algorithm was effective to prevent the over-segmentation phenomena of watershed algorithm.(3)The area of the mean,area of variance,difference of two extraction particle average size,the mean perimeter and other parameters were extracted based on the sugar crystal morphology.Those parameters were considered as an index for crystallization status identification.Due to the independence of these 7 characteristic parameters’ value,we utilized rough set to extract the features of parameters and reduce the dimension of them.Finally,the 7 original parameters defined as index for crystallization were cut into 4.(4)Aimed at the difficulty of recognition of sugar crystalline status in sugar process,a prediction classification method of crystalline status rooted in sugar process of gaussian process,based on classification model of gaussian process,was put forward.Through selecting the proper kernel function and hyper-parameter optimization,to reduction the characteristic values of the input,the crystalline status as the output,accurate prediction classification of crystalline status was realized in sugar process,which provided basis for realization of automatic control in sugar process.(5)At last,the crystal image processing system for sugar crystallization status detection and analysis was implemented in VC++6.0 environment and tested the image detection and analysis system on the process of boiling sugar monitoring platform.Test results demonstrated that the output of the system was almost identical to the real plant,which was applicable in real sugar production field.