The Research Of Key Technology On The Intelligent Recognition And Counting Of Tubercle Bacillus Of Dynamic Micrograph

Tuberculosis is a chronic disease,which is caused by mycobacterium tuberculosis,and it is also known as phtisis and "white plague".The number of Chinese tuberculosis patients ranked second in the world,with more than 130,000 deaths each year,which is severely restricting China's economy development.As the traditional methods of diagnosis have many shortcomings,which have a serious influence on the diagnosis of the efficiency,velocity and accuracy.Therefore,how to achieve standardization,intelligence and automation in the TB (Tubercle Bacillus) bacteriology conventional methods by using modern science and technology is the goal of disease control departments and research institutions of our country and many other countries.This paper deals with the application research on image process and recognition in the field of medical.Both the machine vision and image processing are applied in detecting TB in this paper so as to research how to realize a dynamic and intelligent detection on the continuous vision of the same sputum smear micrograph.The research includes two main parts.The first part is the study of auto-focusing system of TB of dynamic micrograph, which mainly includes the selection of image definition criterion function of auto-focusing in the range of low accuracy and high accuracy and the selection of searching strategy of auto-focusing.The second part is about intelligent recognition of TB of dynamic micrograph, which is the core of the whole paper.This part mainly includes the research of pre-processing algorithm for the complex background in the HSI color space,and the research on the algorithm of segmentation and feature extraction on the micrograph of TB.According to the feature extraction of Tubercle Bacillus,firstly 6 measurement parameters of TB,such as length,width,perimeter,area,rectangularity,extension,are extracted as morphological characteristics of TB in the connectivity region of microscopic image of TB.Considering the demand on counting of TB,we secondly extract 2 important characteristics of the number of long branches and the number of sub-branches as accessional characteristics of Tubercle Bacillus on the goal region of TB.As a result,the experiment shows that the 8 characteristics are of good classification effect.Finally,the accurate identification,classi- fication and counting on microscopic image of TB are researched through the BP artificial neural network algorithm.The intelligent detection of tuberculosis proposed in this article will greatly improve the accuracy and efficiency of TB detection,which will also provide a scientific basis for clinical diagnosis of tuberculosis as well as prevention and cure of tuberculosis.The intelligent detection of tuberculosis will have a seriously practical significance and great social significance to prevent and control tuberculosis in our country.