Multicolor colonies are a phenomenon often encountered in microbial detection. Colonies of different colors often represent different bacteria, and the classification and detection of multi-colored colonies becomes an important task.

At present, the automatic classification of multi-color colonies is generally carried out by the method of colored cubes. This method is simple to operate, but the effect is often not good. This is because:

(1) The surface color of each colony belonging to the same color is not exactly the same, and sometimes there are often quite different. For example, in Figures 1-a and 1-b below, the red colonies have dark red and light red, but they obviously belong to the same category; in Figure 1-c below, the two red colonies have different color shades, and the surface is also caused by reflections Stains.

(2) For a single colony, its surface color is not static, often the middle part is darker in color, while the other parts are slightly lighter in color, and there is even a lighter circle of color caused by diffusion around, as shown in Figure 1- The blue colony in a.

For the above reasons, traditional methods such as colored cubes often divide a complete colony into many fragments. Figure 2 shows the partial enlargement of the division of Figure 1 using this method. The pigment diffusion of the blue colony causes a colony to be over-segmented into The inner ring and the outer ring are two parts.

Figure 1 Multicolor colony image

Figure 2 Over-segmentation caused by uneven (or diffuse) color of the colony surface

On the other hand, the image segmentation method based on the level set active contour model has the advantages of strong noise resistance, good numerical solution stability, smooth and continuous segmentation boundary, etc., and will not cause over segmentation or many fragments. How to introduce color as a constraint to the level set active contour model to solve the above automatic classification of multicolor colonies is one of the goals of the research team of Xunshu Technology in recent years. It has achieved success and can solve the automatic multicolor colonies. Classification problem.

1. Multicolor clustering of level set active contour model based on RGB constraints

The basic principle of the level set active contour model is to embed the curve or surface in the higher one-dimensional level set function, and use a high-dimensional function to express the evolution process of the low-dimensional curve or surface. Establish a two-dimensional energy functional (ie CV model) with closed curve length and area smoothing constraints:

In the above formula, I (x, y) is the gray level of each pixel point in the image, and co and cb are the average gray level values ​​inside and outside the colony outline, respectively. The first two items of the above formula are used to control the smoothness of the colony contour curve, and the latter two items drive the contour line to shrink to the actual colony contour. Minimizing the energy function completes the separation of the colony image.

The traditional CV model only uses grayscale information, and it is inefficient to segment multicolor colonies with rich color information. In order to integrate color information into it, the traditional gradient indication function based on gray information is changed to the gradient indication function based on color information. Let I (x, y) represent the three R, G, and B vectors of the color image at point (x, y). Define E, F, G and fcolor as follows:

The energy function expression of Ecv (Ф) is:

The gradient descent flow is as follows:

Among them, I, c1, c2 are the image gray value, the average gray value inside and outside the contour line. In order to add color information, the traditional gray-scale difference is changed to R, G, B three channel difference weighted average. The expression is as follows:

Where c1i and c2i are the average of the three channels inside and outside the contour, and Ii corresponds to the intensity values ​​of the three channels. The curve evolution equation of the color active contour model based on gradient and CV model is as follows:

The above-mentioned integrated active contour model based on color gradient and multi-channel CV model can better solve the problem of automatic color classification of multi-color colony images.

2. Automatic classification test of multicolor colonies

Figures 3 and 4 show the effect of segmentation of multi-color colonies using the fast multi-color classification technology.

Figure 3 shows the automatic classification and detection of three color colonies. First, obtain the a priori knowledge of the three colonies of red, blue, and yellow from the colony image (Figure 3-a), and add it to the constraints of the active contour model of the CV level set through the RGB model. In this way, during the evolution of the CV model, only the specified color contour is contracted, and the classification results are shown in Figure 3-b.

Figure 3 The effect of automatic classification of multicolor colonies

Figure 4 shows the automatic classification of multicolor colonies that contain 2 color colonies but have significant pigment diffusion. The method is the same as above, first obtain the a priori knowledge of the red and green colonies from the colony image (Figure 4-a), and add it to the constraints of the active contour model of the CV level set through the RGB model. In this way, during the evolution of the CV model, only the specified color contour is shrunk. The classification results are shown in Figure 4-b.

Figure 4 The effect of automatic classification of multicolor colonies

3. Summary

The image segmentation method based on the level set active contour model has the advantages of strong noise resistance, good numerical solution stability, smooth and continuous segmentation boundary, and can deal with complex topological structures. It can solve many complex problems.

After more than two years of research and development, Xunshu Technology R & D team has not only mastered this advanced technology, but also aimed at the characteristics of microbial colonies. Based on the traditional level set active contour model, by introducing various constraint mechanisms and boundary conditions, Creative research has developed advanced image segmentation or automatic classification technology suitable for a variety of complex situations, providing practical and advanced technology for colony counting and microscopic image analysis.

Hangzhou Xunshu Technology Co., Ltd. R & D Department

Folding Shoe Rack:
Folding Shoe Rack is used to store shoes for easy management. It consists of a stationary frame and a moving frame so that you can fold it when you don't need it. It is made of antirust, strong and durable iron steel tube. Also, folding shoe rack is easy to assemble and clean. Folding shoe rack can be used in living rooms, bedrooms, balconies and so on, which can make your shoes in order. We have various of folding shoe stands for your selection. Welcome new and old customers contact us, and we will provide first-class products and quality service to you.

Folding Shoe Rack

Folding Shoe Rack,Folding Shoe Stand,Folding Shoe Holder

Xinhui Mingcheng Hardware Mamufacture Limited of Jiangmen City , https://www.gd-jmmc.com