Study On Intelligent Algorithm Of Colony Counting And Classification

Colony counting is widely used in detection of food, hygiene, medical treatment, environmental protection and security. The quantity demanded of Colony counting work is very much more especially in recent years for strengthen supervising in food security and environmental protection. However the traditional method, which rely on manual counting not only is tedious and time-consuming, but also has high subjective error rate. Aiming at this problem, the paper studies the image processing algorithm to realize automatic colony counting. In order to count the total number of the colony on ordinary growth media or the number of each kind of colony on selective media, which are both the usual problem in this field, the paper consults many references and relevant standard. At last, the paper designs the algorithm to realize colony counting and classification, by using development tool of OpenCV.In the counting algorithm, the paper uses Canny operator to detect the edge of colony image after preprocessing. The Hough transform is used to ensure the location of the petri dish, and then, the background outside the petri dish can be removed. Dynamic local threshold segmentation is used to separate the colony from the media. Iterative erosion is used to transform the colony into the seed points to avoid colony adhesion, and the seed points counting and marking are completed by neighborhood searching.In the classification algorithm, the design starts with the seed points. At first, region growing and Graph Cut are used to segment the colony region accurately. At second, color, shape and texture of each colony region are extracted and quantized into feature vectors. First moment, second moment and third moment of the three color channels are used to describe the color information of the colony. Freeman chain code is used to measure the perimeter, area and circularity of the colony, long and short axis can be obtained by Hough transform. All the parameters can describe the shape information of the colony. At last, GLCM is used to obtain energy, contrast, inverse difference moment and correlation to describe the texture information of the colony. At last, SVM is used to find the best hyperplane based on feature vectors to realize colony classification.The paper captures 50 images of colony to check the effect of the algorithm. The results show that the mean and highest error rate of the counting algorithm are 4.43% and 10.25%, while the classification algorithm are 2.76% and 5.08%, which can satisfy the national standard GB4789.2-2010. The algorithm can realize colony counting and classification well.