Research On Image Analysis Based L-glutamic Crystallization Monitoring Technology

Crystallization process plays an important role in chemical industry,especially in fine chemical industry.It is widely used in the crystal production.In order to obtain the crystal products with the desired form,the crystal morphology during the crystallization process should be obtained for reasonably controlling,which requires a real-time technique to monitor the crystallization process.With the rapid development of high-speed cameras and other hardware devices,image analysis based techniques show a great potential in the field of crystallization monitoring.Crystal segmentation is the most important step in the image analysis based crystallization monitoring.However,it is hard to perfectly segment crystals from the image due to the interferences of droplets and particle shadow caused by the invasive imaging system.In addition,the complex segmentation algorithms also increase the time complexity of the operation of the crystal segmentation.In order to solve the problems above and achieve the purpose of real-time crystallization monitoring,this thesis studies the interferences of droplets and particle shadow as follows:(1)For the interferences of droplets caused by the invasive imaging system,this paper proposed a difference method based on average background modeling for droplets elimination.Firstly,an average background model is constructed based on the idea of average background modeling,and then process difference method between the particle image and the background model to eliminate the droplets and also the complex background;(2)For the interferences of particle shadow,this paper proposed a local threshold method based on graph-cut.Considering the particularity of crystallization,the algorithm designed a synthesized strategy of particle area separating.Concretely,divide the whole image into several small particle areas,and then perform the threshold process for every isolated area to eliminate particle shadow;(3)On the basis of(1)and(2),this paper presents a kind of crystal classification descriptor for the polycrystalline of L-glutamic acid called aspect ratio,and adopt image processing strategies to analyze the morphological information of the crystal after the segmentation to achieve the objective of obtaining the crystal visual information during the crystallization process.Through comparison experiments the proposed methods were convinced to be effective for eliminating the interferences of droplets and particle shadow and also cleanly segmenting the particles from the images.And further the real-time morphology visual information could be obtained.The time complexity of the proposed methods are pretty low,indicating that the proposed method can be efficiently applied to real-time monitoring of crystallization processes.