The Research Of Automatic Cell Recognition And Calculation System By Image Processing Method

At present, in many of the biological experiment and medical diagnosis fields, to recognize and calculate the number of cells is a regular testing project. But, now testing project is mainly by dying cells,then identifying and counting cells. It is time-consuming and inefficiency to do the project by this method. But the use of embedded cell image process devices can make it convenient and efficient. If this design can get the actual application, it will greatly improve the efficiency of work. Therefore, the application will have a good future.In software,by comparing the advantage and disadvantage of various algorithms of the image processing method and considering the specificity of cow’s cell image processing, we choose a set of appropriate image processing algorithms through image processing algorithms simulation by matlab to identify and count cells. In this paper,we can get cell characteristic parameters by the morphology,then use these parameters as the input parameters of BP neural network.At last, we can use BP neural network to identify and count cell.In addition,in order to improve the accuracy, we can use color image processing to ajust the number of cells.In hardware in this paper,we carryed median filter algorithms,gray histogram algorithms,threshold value algorithms,corrosion algorithms and dilation algorithms on the ADSP-BF533platform,then use characteristic parameters to identify and count cells.The program on the ADSP-BF533platform has high accuracy and run stably.Through the experiment,it proves that it is feasible to count cells by embedded devices. Because real-time image processing system is a complicated system including software and hardware,so if we want to design commonly used real-time image processing system, we have a lot of work to do in the optimization of hardware and software.