CN113744184A - Snakehead ovum counting method based on image processing - Google Patents

Abstract

The invention discloses a snakehead ovum counting method based on image processing, which comprises the following steps of: s1, collecting images of the snakehead ovum targets needing counting, and transmitting the images into a computer; s2, carrying out gray processing on the collected image; s3, performing image enhancement by adopting a top-hat transformation algorithm, firstly designing a circular structural element, performing an opening operation on an image, and subtracting the image obtained by the opening operation from the original image to realize an enhancement effect; s4, performing image filtering processing by adopting a guide filtering method; s5, image segmentation; and S6, identifying and counting the ovum. The snakehead ovum picture is obtained by directly imaging the snakehead ovum, and the egg counting task is completed on the non-contact premise through the processing flow provided by the invention, so that the precision is high, and the original living environment of the ovum cannot be changed.

Description

Snakehead ovum counting method based on image processing

Technical Field

The invention relates to the field of intersection of image processing and aquaculture, in particular to a snakehead egg counting method based on image processing.

Background

The snakehead has big body, fast growth and rich nutrition, and has higher protein content than chicken and beef. The snakehead fish culture medium has the effects of removing blood stasis, promoting tissue regeneration, nourishing, recuperating and the like when being used as a medicine, so that the snakehead fish culture medium has higher economic value. Snakeheads inhabit at the muddy bottom of lake, pond and river, and their spawning site is also used. The estimation of the spawning time and the spawning amount is beneficial to the artificial culture in the later period.

The conventional counting method for the snakehead eggs has four methods. Direct counting is the most original, but the counting is time-consuming, especially the batch seedling growing quantity is large, and the counting one by one is impossible. Area, volume or measurement methods may also be used. Since the eggs of the snakehead belong to floating eggs, the produced eggs are not disturbed to scatter, are arranged regularly, tightly and intensively and float on the surface layer of water, and an area counting method can be adopted. The area counting method is to count the number of eggs by taking an egg per unit area and then measuring the area of each litter to obtain the number of eggs per litter. The volume method is that a standard container is filled with fertilized eggs, and then the number of the eggs in the volume is counted. The total number of eggs can be calculated by measuring the number of cups by using the standard container.

In general, the traditional snakehead ovum counting method has two main problems, one is that the traditional counting method is not only a manual direct counting method, but also an area method, a volume method and a weighing method are contact counting methods, and the ovum needs to be fished, so that the original environment is damaged; secondly, because the methods adopt a mode of estimating the area, the volume or the weight and then equally dividing, certain errors exist in the difference among individuals, and the errors are difficult to eliminate.

Disclosure of Invention

The purpose of the invention is as follows: in order to overcome the defects of the background technology, the invention discloses a snakehead egg counting method based on image processing, which completes an egg counting task on the premise of non-contact, has high precision and does not change the original living environment of the eggs.

The technical scheme is as follows: the invention relates to a snakehead ovum counting method based on image processing, which comprises the following steps of:

s1, collecting images of the snakehead ovum targets needing counting, and transmitting the images into a computer;

s2, carrying out gray processing on the collected image;

s3, performing image enhancement by adopting a top-hat transformation algorithm, firstly designing a circular structural element, performing an opening operation on an image, and subtracting the image obtained by the opening operation from the original image to realize an enhancement effect;

s4, performing image filtering processing by adopting a guide filtering method;

s5, image segmentation

S5.1, filtering the image in the horizontal and vertical directions by using a Sobel edge operator, and then solving a module value to obtain a gradient image;

s5.2, classifying all pixels in the gradient image according to the gray value, and setting a geodesic distance threshold;

s5.3, finding out the pixel point with the minimum gray value as a starting point, and increasing the threshold from the starting point;

s5.4, in the process of increasing the horizontal plane, the horizontal plane can touch surrounding neighborhood pixels, the geodesic distance from the pixels to a starting point is measured, if the geodesic distance is smaller than a set threshold value, the pixels are submerged, otherwise, a dam is arranged on the pixels, and the neighborhood pixels are classified in this way;

s5.5, setting more higher dams as the horizontal plane is higher until the maximum value of the gray value is reached, wherein all the areas meet on the watershed line, and all the dams partition the pixels of the whole image;

s6, identifying and counting ova

S6.1, corroding the target areas in the segmentation result by using circular structural elements with the radius of 2, wherein the distance between the segmented areas is increased, and the mutual influence of the adjacent target areas is avoided;

s6.2, calling a find contacts () function in an Open CV library to extract the outline of each area in the particle segmentation result;

s6.3, sequentially traversing each region contour, and solving a minimum circumcircle of a point set of the region contour which can be surrounded by a min Enclosing Circle () function, wherein the radius r and the Circle centers (a and b) of the Circle are approximate to the center point of the particle;

s6.4, as the target is corroded by the S6.1, increasing the radius of the solved circumscribed circle by 2, and drawing a circle with the circle center of (a, b) and the radius of r +2 on the original graph;

s6.5, counting the area of all circles as A₁，A₂…A_NAnd (4) obtaining the area median of S _ mean, and deleting circles with the area larger than or smaller than 0.2S _ mean, wherein the circles are too large or too small in area and should not count the number of the eggs. And counting the remaining circles, wherein the number of the remaining circles is the final number of the eggs.

Wherein, the number of effective pixel points of a single egg target in the image collected in the S1 is not less than 20 x 20.

Further, in the gradation processing in S2:

Gray＝0.299R+0.587G+0.114B

wherein, R, G and B are respectively red, green and blue channels for collecting color images.

Further, S3 specifically includes the following steps:

s3.1, designing a circular structural element b;

s3.2, performing gray scale corrosion and expansion operation on the original image by using the structural element b;

the etching process is to subtract each pixel point in the gray image f (x, y) from the corresponding point in the structural element b (x, y), and the etching result is the minimum value:

the expansion process is a corrosion process dual operation, each pixel point in the gray image f (x, y) is added with the corresponding point in the structural element b (x, y), and the corrosion result is the maximum value:

after the image is subjected to corrosion operation and then expansion operation, the opening operation is completed, and an operation result graph f is obtained_open；

S3.3, subtracting the result of the operation from the original image:

f_top-hat＝f(x,y)-f_open。

further, the guiding filtering in S4 assumes that there is a local linear relationship between the guiding image and the output image, as shown in the following equation:

wherein I is a guide image, q is a filtered output result, and w_kIs a square window with pixel point k as center, a_k、b_kThe constant coefficient in the window takes the following values:

wherein, mu_kAnd

for directing the image I at the window w_kIs the mean and variance of, | w | is the window w_kThe total number of the pixel points in (1),

is that the input image p is in the window w_kAverage value of the middle pixel points.

Has the advantages that: the snakehead ovum picture is obtained by directly imaging the snakehead ovum, and the egg counting task is completed on the non-contact premise through the processing flow provided by the invention, so that the precision is high, and the original living environment of the ovum cannot be changed.

Drawings

FIG. 1 is a flow chart of the present invention.

Detailed Description

The technical solution of the present invention is further described below with reference to the accompanying drawings and examples.

Reading hyperspectral remote sensing image data and registered high-resolution image data by a computer, firstly establishing a preliminary objective function based on nonnegative matrix decomposition, then adding a spectrum constraint term on the basis to obtain a final fusion model, solving an extreme value function of the fusion model to obtain a standard image and a weight matrix, multiplying the high-resolution image by the standard image and the weighted image of the high-resolution image, and obtaining a result, namely the solved hyperspectral fusion image.

As shown in figure 1, the snakehead egg counting method based on image processing adopts a computer configuration that: an Intel dual-core processor with a main frequency of 1.6GHz and a memory of 2GB, wherein an operating system is Windows Vista Home Basic, and a programming environment is Matlab7.1. The method comprises the following steps:

s1: image acquisition

And adopting an image acquisition system carrying a high-definition camera to acquire images of the snakehead ovum targets needing counting, and transmitting related images into a computer, wherein the number of effective pixel points of each ovum target in the images is not less than 20 x 20.

S2: graying processing

And the color image is converted into a gray image, so that the subsequent processing is facilitated.

Gray＝0.299R+0.587G+0.114B

Wherein, R, G and B are respectively red, green and blue channels for collecting color images.

S3: image enhancement pre-processing

A top-hat transformation algorithm is adopted to solve the problem of inconsistent image illumination intensity, a circular structural element is designed firstly, an opening operation is carried out on an image, the image obtained by the opening operation is subtracted from an original image, and an enhancement effect is achieved.

S3.1, designing a circular structural element b:

the circular structural element b with the radius of 5 is designed as follows:

s3.2, performing gray scale corrosion and expansion operation on the original image by using the structural element b:

the etching process is to subtract each pixel point in the gray image f (x, y) from the corresponding point in the structural element b (x, y), and the etching result is the minimum value:

the expansion process is a corrosion process dual operation, each pixel point in the gray image f (x, y) is added with the corresponding point in the structural element b (x, y), and the corrosion result is the maximum value:

and after the image is subjected to corrosion operation and then expansion operation, opening operation is completed. The obtained operation result graph f_openThe overall brightness is relatively uniform.

S3.3 subtracting the result of the opening operation from the original image

f_top-hat＝f(x,y)-f_open

And subtracting the operation result from the original image to obtain a top-hat calculation result.

S4: image filtering

And (4) performing filtering processing on the enhanced image obtained in the step (S3) by adopting a guide filtering method, eliminating noise and obtaining an image with clear edge texture. The guided filtering assumes a local linear relationship between the guided image and the output image, as shown by:

here, I is the guide image, and q is the filtered output result. w is a_kIs a square window with pixel point k as center, (a)_k,b_k) The constant coefficient in the window takes the following values:

μ_kand

for directing the image I at the window w_kIs the mean and variance of, | w | is the window w_kThe total number of the pixel points in (1),

is that the input image p is in the window w_kAverage value of the middle pixel points.

S5: image segmentation

S5.1, filtering the image in the horizontal and vertical directions by using a Sobel edge operator, then solving a module value to obtain a gradient image, wherein the image filtered by the Sobel operator shows a larger value at a boundary and a smaller value at a position without the boundary.

S5.2, classifying all pixels in the gradient image according to the gray values, and setting a geodesic distance threshold.

S5.3, finding out the pixel points with the minimum gray value (the default mark is the lowest gray value), and increasing the threshold from the minimum value, wherein the points are the starting points.

S5.4, in the process of increasing the horizontal plane, the horizontal plane can touch surrounding neighborhood pixels, the geodesic distance from the pixels to a starting point (the lowest point of the gray value) is measured, if the geodesic distance is smaller than a set threshold value, the pixels are submerged, otherwise, a dam is arranged on the pixels, and the neighborhood pixels are classified.

S5.5, as the horizontal plane is higher and higher, more and higher dams are arranged, until the maximum value of the gray value is reached, all the areas meet on the watershed line, and the dams partition the pixels of the whole image.

S6: ovum identification and counting

After splitting the adhered overlapping particles, the final recognition result is marked on the original for easy observation. The method comprises the following specific steps:

s6.1, corroding the target areas in the segmentation result by using circular structural elements with the radius of 2, wherein the distance between the segmented areas is increased, and the mutual influence of the adjacent target areas is avoided.

S6.2 calls the find contacts () function in the Open CV library to extract the outline of each area in the particle segmentation result.

S6.3, traversing each region outline in sequence, and solving the smallest circumcircle which can surround the region outline point set by using a min Enclosing Circle () function, wherein the radius r and the Circle centers (a, b) of the circles can be approximate to the center point of the particle.

S6.4, because the target is slightly corroded in the step 1, the radius of the solved circumscribed circle is increased by 2, and finally, a circle with the circle center of (a, b) and the radius of r +2 is drawn on the original graph.

S6.5, counting the area of all circles as A₁，A₂…A_NAnd (4) obtaining the area median of S _ mean, and deleting circles with the area larger than or smaller than 0.2S _ mean, wherein the circles are too large or too small in area and should not count the number of the eggs. Then countThe number of the remaining circles is the final number of eggs.

The method of the invention well solves the problem of snakehead ovum counting based on image processing, so the method can be applied to automatic snakehead breeding and has wide application prospect and value.

Claims (5)

1. A snakehead egg counting method based on image processing is characterized by comprising the following steps:

s1, collecting images of the snakehead ovum targets needing counting, and transmitting the images into a computer;

s2, carrying out gray processing on the collected image;

s3, performing image enhancement by adopting a top-hat transformation algorithm, firstly designing a circular structural element, performing an opening operation on an image, and subtracting the image obtained by the opening operation from the original image to realize an enhancement effect;

s4, performing image filtering processing by adopting a guide filtering method;

s5, image segmentation

S5.1, filtering the image in the horizontal and vertical directions by using a Sobel edge operator, and then solving a module value to obtain a gradient image;

s5.2, classifying all pixels in the gradient image according to the gray value, and setting a geodesic distance threshold;

s5.3, finding out the pixel point with the minimum gray value as a starting point, and increasing the threshold from the starting point;

s5.4, in the process of increasing the horizontal plane, the horizontal plane can touch surrounding neighborhood pixels, the geodesic distance from the pixels to a starting point is measured, if the geodesic distance is smaller than a set threshold value, the pixels are submerged, otherwise, a dam is arranged on the pixels, and the neighborhood pixels are classified in this way;

s5.5, setting more higher dams as the horizontal plane is higher until the maximum value of the gray value is reached, wherein all the areas meet on the watershed line, and all the dams partition the pixels of the whole image;

s6, identifying and counting ova

S6.1, corroding the target areas in the segmentation result by using circular structural elements with the radius of 2, wherein the distance between the segmented areas is increased, and the mutual influence of the adjacent target areas is avoided;

s6.2, calling a find contacts () function in an Open CV library to extract the outline of each area in the particle segmentation result;

s6.3, sequentially traversing each region contour, and solving a minimum circumcircle of a point set of the region contour which can be surrounded by a min Enclosing Circle () function, wherein the radius r and the Circle centers (a and b) of the Circle are approximate to the center point of the particle;

s6.4, as the target is corroded by the S6.1, increasing the radius of the solved circumscribed circle by 2, and drawing a circle with the circle center of (a, b) and the radius of r +2 on the original graph;

s6.5, counting the area of all circles as A₁，A₂…A_NAnd obtaining the area median value S _ mean, deleting the circles with the area larger than or smaller than 0.2S _ mean, counting the number of eggs, and counting the number of the remaining circles to obtain the final number of the eggs.

2. The method for counting snakehead eggs based on image processing according to claim 1, wherein: and the number of effective pixel points of a single egg sub target in the image collected in the S1 is not less than 20 x 20.

3. The method for counting snakehead eggs based on image processing according to claim 1, wherein: in the gradation processing in S2:

Gray＝0.299R+0.587G+0.114B

wherein, R, G and B are respectively red, green and blue channels for collecting color images.

4. The method for counting snakehead eggs according to claim 1, wherein S3 specifically comprises the following steps:

s3.1, designing a circular structural element b;

s3.2, performing gray scale corrosion and expansion operation on the original image by using the structural element b;

the etching process is to subtract each pixel point in the gray image f (x, y) from the corresponding point in the structural element b (x, y), and the etching result is the minimum value:

(fΘb)＝min{f(s+x,t+y)-b(x,y)|(s+x),(t+y)∈D,(x,y)∈S}

the expansion process is a corrosion process dual operation, each pixel point in the gray image f (x, y) is added with the corresponding point in the structural element b (x, y), and the corrosion result is the maximum value:

(fΘb)＝min{f(s-x,t-y)-b(x,y)|(s-x),(t-y)∈D,(x,y)∈S}

after the image is subjected to corrosion operation and then expansion operation, the opening operation is completed, and an operation result graph f is obtained_open；

S3.3, subtracting the result of the operation from the original image:

f_top-hat＝f(x,y)-f_open。

5. the method for counting snakehead eggs based on image processing according to claim 1, wherein: the guiding filtering in S4 assumes that there is a local linear relationship between the guiding image and the output image, as shown by the following equation:

wherein I is a guide image, q is a filtered output result, and w_kIs a square window with pixel point k as center, a_k、b_kThe constant coefficient in the window takes the following values:

wherein, mu_kAnd

for directing the image I at the window w_kIs the mean and variance of, | w | is the window w_kThe total number of the pixel points in (1),

is that the input image p is in the window w_kAverage value of the middle pixel points.

Images