A method for assisting in identifying damage grade of micropterus salmoides iridovirus based on machine vision

Abstract

The application provides a method for assisting in identifying damage grades of Micropterus salmoides iridovirus based on machine vision, and belongs to the technical field of aquatic disease identification. The application is used for standardizing the measurement of the damage degree of Micropterus salmoides infected with iridovirus, assisting in analyzing early Micropterus salmoides body surface damage data based on visual machine, dividing the body surface damage area, measuring the body surface damage quantity and damage area, calculating the individual comprehensive damage score through data standardization and PCA dimension reduction, and constructing the body surface damage grade, so that a scientific basis is provided for the disease management of Micropterus salmoides.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of aquatic disease identification, and particularly relates to a method for identifying damage grades of rainbow trout virus of Micropterus salmoides based on machine vision assistance. BACKGROUND

[0002] Micropterus salmoides has become an important breeding variety in China due to its delicious meat, strong disease resistance, rapid growth and many other advantages. With the rapid development of Micropterus salmoides breeding industry, the adverse factors hindering the sustainable expansion of the industry have gradually increased. Among the fish disease problems caused by intensive breeding, the viral disease rainbow trout virus is the most difficult to prevent and control, and has the highest mortality.

[0003] The suitable temperature for rapid proliferation of Micropterus salmoides rainbow trout virus disease (LMBV) is 24-30℃, which is consistent with the optimal growth environment temperature (24-30℃) of Micropterus salmoides. This leads to the most susceptible rainbow trout virus infection in the best growth and reproduction season, and once the outbreak, the mortality rate is as high as 60%. Rainbow trout virus has multiple transmission routes and fast spreading speed, and has the characteristics of wide transmission range and outbreak of mass death. In the early stage of infection, the symptoms of viral disease are not obvious, but in the late stage of infection, it will cause mass death, accompanied by symptoms such as infectious spleen and kidney necrosis and ulcerative syndrome. At present, the effective control method of Micropterus salmoides rainbow trout virus disease has not been solved. In order to effectively control the infection of rainbow trout virus, it is necessary to start from the aspects of strengthening biological control measures, fry quarantine and fish farm management. Early identification of viral infection and quantification of damage degree are of great significance for disease control, breeding management optimization and fine loss control. However, the traditional evaluation method relies on experience observation, has strong subjectivity and is difficult to standardize. SUMMARY

[0004] Therefore, the purpose of the present application is to provide a method for identifying the damage grade of rainbow trout virus of Micropterus salmoides based on machine vision assistance, which can analyze the early surface damage data of Micropterus salmoides based on visual machine assistance, divide the surface damage area, measure the surface damage quantity and damage area, calculate the individual comprehensive damage score through data standardization and PCA dimension reduction, and construct the surface damage grade, thereby providing a scientific basis for the disease management of Micropterus salmoides.

[0005] The present application provides a method for identifying the damage grade of rainbow trout virus of Micropterus salmoides based on machine vision assistance, which comprises the following steps:

[0006] Under the assistance of visual machine, the surface damage data of Micropterus salmoides infected with rainbow trout virus is collected, and the damage segmentation area quantity, surface damage quantity and surface damage area of the front surface, back surface, dorsal surface and abdominal surface of the Micropterus salmoides to be tested are counted;

[0007] The damage segmentation area is divided into 12 damage segmentation areas by taking the gill cover and the end of the anal fin of the largemouth bass as the boundary line to divide the body surface front, the body surface back, the body surface back and the body surface abdomen;

[0008] The data of the number of damage segmentation areas, the number of body surface damage and the area of body surface damage of the body surface front, the body surface back, the body surface back and the body surface abdomen of the largemouth bass to be tested are normalized and substituted into the formula VI to calculate the comprehensive damage score.

[0009] The comprehensive damage score = 0.59 x damage segmentation area + 0.63 x number of body surface damage + 0.68 x area of body surface damage formula VI.

[0010] The higher the comprehensive damage score of the largemouth bass to be tested is, the higher the damage degree is.

[0011] Preferably, the damage segmentation area uses Min-Max normalization method to process data, and is calculated according to formula I.

[0012] Damage segmentation area = damage segmentation area / 10 formula I.

[0013] The number of body surface damage or the area of body surface damage is processed according to z-score normalization method and linear translation method.

[0014] The z-score normalized value of the number of body surface damage is calculated according to formula II or the z-score normalized value of the area of body surface damage is calculated according to formula III.

[0015] Z-score normalized value of the number of body surface damage = (number of body surface damage - μ) / σ formula II.

[0016] Z-score normalized value of the area of body surface damage = (area of body surface damage - μ) / σ formula III.

[0017] In formula II or formula III, μ is the average value of the number or area of damage; σ is the standard deviation of the number or area of damage.

[0018] The number of body surface damage is calculated according to formula IV and the area of body surface damage is calculated according to formula V.

[0019] Number of body surface damage = z-score normalized value of the number of body surface damage + |z-score minimum value| formula IV; area of body surface damage = (z-score normalized value of the number of body surface damage + |z-score minimum value|) / 4 formula V.

[0020] In formula IV, |z-score minimum value| is 0.882; in formula V, |z-score minimum value| is 1.261.

[0021] Preferably, the damage segmentation area is to divide the body surface front, body surface back, body surface back and body surface abdomen into 12 areas according to the following method:

[0022] The body surface front is divided into 3 areas, from the front end of the lower jaw to the gill cover rear along the definition of A area, the gill cover rear to the end of the anal fin definition of B area, the end of the anal fin to the end of the tail fin definition of C area;

[0023] The body surface back is divided into 3 areas, from the end of the tail fin to the end of the anal fin definition of D area, the end of the anal fin to the gill cover rear definition of E area, the gill cover rear from the front end of the lower jaw definition of F area;

[0024] The body surface back is divided into 3 areas, from the end of the tail fin to the front end of the second dorsal fin definition of G area, the dorsal fin front end to the rear end of the skull definition of H area, the rear end of the skull lower jaw definition of I area;

[0025] The body surface back is divided into 3 areas, from the end of the tail fin to the end of the anal fin definition of D area, the end of the anal fin to the gill cover rear definition of E area, the gill cover rear from the front end of the lower jaw definition of F area;

[0026] Preferably, the screening criteria for the rainbow trout infected with the iridovirus include the presence of spot ulcer, red bleeding point and cheek redness lesion damage wound on the body surface.

[0027] Preferably, the growth environment of the rainbow trout infected with the iridovirus is: temperature 25.5-26.5℃, dissolved oxygen ≥7.5mg / L, total ammonia nitrogen ≤1mg / L.

[0028] Preferably, the method of collecting the body surface damage data of the rainbow trout infected with the iridovirus by visual machine assistance is to take the complete body surface phenotype image of the body surface front, body surface back, body surface back and body surface abdomen of the rainbow trout to be tested, finely mark the damage area, and calculate the body surface damage number and body surface damage area of each damage area.

[0029] Preferably, the device for shooting includes a camera.

[0030] Preferably, the software for finely marking the damage area includes ImageJ software.

[0031] Preferably, after obtaining the comprehensive damage score of the rainbow trout to be tested, it further includes judging the damage degree of the rainbow trout to be tested according to the scoring standard;

[0032] The scoring standard is divided into the following three grades:

[0033] 0<comprehensive damage score≤0.53 is mild damage;

[0034] 0.53<comprehensive damage score≤1.13 is moderate damage;

[0035] 1.13<The comprehensive injury score is less than or equal to 3.65, which is a severe injury.

[0036] Preferably, the body surface morphology of the mild injury of the largemouth bass is that part of the red bleeding points appear on the body surface front and the body surface back, and there is no obvious ulcer-like injury.

[0037] The body surface morphology of the moderate injury of the largemouth bass is that obvious red bleeding points appear on the entire body surface, operculum and caudal region, and the number of ulcers increases.

[0038] The body surface morphology of the severe injury of the largemouth bass is that the number of red bleeding points on the entire body surface increases compared with the body surface morphology of the toxic injury of the largemouth bass, and is accompanied by ulcer lesions and muscle necrosis on the body surface.

[0039] The present application provides a method for identifying the injury grade of largemouth bass iridovirus based on machine vision assistance, comprising the following steps: collecting the body surface injury data of largemouth bass infected with iridovirus under the assistance of visual machine, counting the number of injury segmentation areas, the number of body surface injuries and the area of body surface injuries on the body surface front, the body surface back, the body surface back and the body surface abdomen of the largemouth bass to be tested; the injury segmentation area is divided into 12 injury segmentation areas by taking the operculum and the end of the anal fin of the largemouth bass as the boundary line; the data of the number of injury segmentation areas, the number of injuries and the area of injuries on the body surface front, the body surface back, the body surface back and the body surface abdomen of the largemouth bass to be tested are substituted into formula I to calculate the comprehensive injury score; the comprehensive injury score = 0.59 x injury segmentation area + 0.63 x body surface injury number + 0.68 x body surface injury area formula VI; the higher the comprehensive injury score of the largemouth bass to be tested, the higher the injury degree. The present application collects the body surface injury data of largemouth bass infected with iridovirus under the assistance of visual machine, and further analyzes the data of the number of body surface injury segmentation areas, the number of body surface injuries and the area of body surface injuries of the largemouth bass, determines the injury segmentation area as the independent variable, and the number of injuries and the area of injuries as the dependent variable, and the correlation coefficients are 0.9462 and 0.9453 respectively, which shows a high positive correlation; similarly, the correlation coefficient of the number of injuries and the area of injuries is also high, and the correlation coefficient is 0.7901. The three redundant variables of injury segmentation area, body surface injury number and body surface injury area are compressed into a single comprehensive index to obtain the comprehensive injury score. Then, the number of injury segmentation areas, the number of body surface injuries and the area of body surface injuries of the body surface front, the body surface back, the body surface back and the body surface abdomen of the largemouth bass to be tested are collected under the assistance of machine vision, and substituted into the comprehensive injury score, so as to accurately judge the injury degree of largemouth bass infected with iridovirus according to the results of the comprehensive injury score, which can greatly improve the scientific nature and prevention efficiency of disease management, and reduce the fine injury of aquaculture. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 Comparison results of growth data between healthy and diseased large-mouth bass individuals; wherein A is body weight, B is body length, C is body height, and D is body thickness;

[0041] Figure 2 An illustration of 12 segmentation areas on the body surface of large-mouth bass;

[0042] Figure 3 An illustration of body surface damage; wherein A is the frequency of individuals with damage in the 12 different segmentation areas; B is the extent of damage spread on the fish body; C is the frequency of the presence of 0-3 damage segmentation areas in the four body surfaces; D is the result between the number of damage segmentation areas and the number of damage; E is the result between the number of damage segmentation areas and the area of damage; F is the number of damage present on the four body surfaces of the body surface front, the body surface back, the body surface back, and the body surface abdomen;

[0043] Figure 4 An illustration of the body surface characteristics of 12 damage segmentation areas infected with iridovirus;

[0044] Figure 5 Correlation analysis results of the number of damage segmentation areas, the number of damage, and the area of damage; A is the linear correlation between the number of body surface damage segmentation areas and the number of body surface damage; B is the linear correlation between the number of body surface damage segmentation areas and the area of body surface damage; C is the linear correlation between the number of body surface damage and the area of body surface damage;

[0045] Figure 6 Damage grade discrimination criteria for large-mouth bass infected with iridovirus; A is the undamaged grade; B is the mild damage grade; C is the moderate damage grade; D is the severe damage. DETAILED DESCRIPTION

[0046] The present application provides a method for assisting in identifying the damage grade of large-mouth bass iridovirus based on machine vision, comprising the following steps:

[0047] Collecting body surface damage data of large-mouth bass infected with iridovirus under the assistance of visual machine, and counting the number of damage segmentation areas, the number of body surface damage, and the area of body surface damage on the body surface front, the body surface back, the body surface back, and the body surface abdomen of the large-mouth bass to be tested;

[0048] The damage segmentation areas are divided into 12 damage segmentation areas by taking the gill cover and the end of the anal fin of the large-mouth bass as the boundary line to divide the body surface front, the body surface back, the body surface back, and the body surface abdomen;

[0049] The data of the number of damage segmentation areas, the number of body surface damage, and the area of body surface damage on the body surface front, the body surface back, the body surface back, and the body surface abdomen of the large-mouth bass to be tested are normalized and substituted into the formula VI to calculate the comprehensive damage score;

[0050] Comprehensive injury score = 0.59 x injury segmentation area + 0.63 x body surface injury number + 0.68 x body surface injury area Formula VI

[0051] The higher the comprehensive injury score of the to-be-tested largemouth bass is, the higher the injury degree is.

[0052] In the present application, in order to understand the influence of iridovirus infection on the culture of largemouth bass, the immersion method of virus infection is used to simulate the natural infection of largemouth bass, and the influence of iridovirus on the growth and development of largemouth bass is evaluated by measuring the changes of growth indexes. The present application first isolates iridovirus from the tissues of largemouth bass naturally infected with iridovirus, and determines the TCID50 of the iridovirus sample by infecting the epithelioma cells of carp, so as to determine the virulence of the sample. 50 is 4 x 10 5 / mL. Largemouth bass is cultured in water containing iridovirus, so as to complete the infection culture. After 3 months of culture, by detecting the body weight, body length, body height and body thickness of largemouth bass, it is found that the body weight (124.93±33.27g), body length (16.79±1.37cm) and body height (5.04±0.50cm) of the largemouth bass are significantly decreased compared with healthy individuals, and the body thickness has no significant change. This shows that iridovirus can significantly reduce the growth performance of largemouth bass and affect the normal development of individuals. Therefore, in order to reduce the influence of iridovirus on the growth and development in the process of largemouth bass culture, timely detection and detection of largemouth bass with different degrees of iridovirus infection are beneficial to strengthen the biological control measures, seed quarantine, fish farm management and the like.

[0053] In the present application, in order to accurately divide the damage of the largemouth bass infected with iridovirus, a comprehensive damage score method is first established, the present application assists in analyzing early largemouth bass body surface damage data based on visual machine, divides the body surface damage area, measures the body surface damage quantity and body surface damage area, and calculates the individual comprehensive damage score through data standardization and PCA dimension reduction. The growth environment of the largemouth bass infected with iridovirus is preferably: temperature 25.5-26.5℃, dissolved oxygen ≥7.5mg / L, and total ammonia nitrogen ≤1mg / L. The largemouth bass infected with iridovirus is the object of damage data collection after being infected with iridovirus for 3 months. The screening standard of the largemouth bass infected with iridovirus preferably includes the presence of spot ulcer, red bleeding point and cheek redness lesion damage wound on the body surface. The largemouth bass body surface damage data preferably includes the damage segmentation area quantity, damage quantity and damage area of the body surface front, body surface back, body surface back and body surface abdomen of the largemouth bass. The body surface front, body surface back, back and abdomen are defined as follows: the largemouth bass is laid flat on the measurement plate, the head of the largemouth bass is directed to the left hand direction of the observer, and the body surface part from the lower jaw front end to the tail fin end is defined as the body surface front; the laid flat largemouth bass is horizontally turned over by 180 degrees and laid flat on the measurement plate, and the tail fin end to the lower jaw front end is defined as the body surface back; the back is turned up by 90 degrees, and the body surface back is defined as the body surface back from the tail fin end to the lower jaw front end; and the abdomen is turned up by 180 degrees, and the body surface abdomen is defined as the body surface abdomen from the tail fin end to the lower jaw front end. The damage segmentation area is preferably divided into 12 areas according to the following method: the body surface front is divided into 3 areas, the area from the lower jaw front end to the gill cover rear edge is defined as A area, the area from the gill cover rear edge to the anal fin end is defined as B area, and the area from the anal fin end to the tail fin end is defined as C area; the body surface back is divided into 3 areas, the area from the tail fin end to the anal fin end is defined as D area, the area from the anal fin end to the gill cover rear edge is defined as E area, and the area from the gill cover rear edge to the lower jaw front end is defined as F area; the body surface back is divided into 3 areas, the area from the tail fin end to the second dorsal fin front end is defined as G area, the area from the dorsal fin front end to the head bone rear end is defined as H area, and the area from the head bone rear end to the lower jaw front end is defined as I area; and the body surface abdomen is divided into 3 areas, the area from the tail fin end to the cloaca is defined as J area, the area from the cloaca to the gill cover rear edge is defined as K area, and the area from the gill cover rear edge to the lower jaw front end is defined as L area. The method for collecting the body surface damage data of the largemouth bass infected with iridovirus by visual machine assistance preferably comprises the following steps: shooting the complete body surface phenotype image of the body surface front, body surface back, body surface back and body surface abdomen of the largemouth bass to be measured, finely marking the damage area, and calculating the body surface damage quantity and damage area of each damage area. The shooting device preferably comprises a camera. In the embodiment of the present application, the camera is preferably a SONY ILCE-7RM3 camera, and the camera parameters during shooting are preferably wide circle F / 11, 75mm focal length and 1 / 200 seconds.The photographed environment is preferably a photograph of the damaged part of the large-mouth bass taken in a small soft light shooting box (GODOX-LST60). The number of photographed photographs is preferably 900-1100, which can be 1000. The fine marking of the damage area is preferably performed using software, including ImageJ software. The method for fine marking of the damage area using the ImageJ software is as follows: after defining the ruler distance according to the scale, the damage area is fine marked using the lasso tool, the number of body surface damage and the damage area of each damage area are calculated in detail, and the number of damage areas of each damage individual is counted.

[0054] In the present application, after obtaining the number of body surface damage areas, the number of body surface damages and the area of body surface damages, the three variables are normalized to eliminate the dimensional differences between them. The Min-Max normalization method is used for the segmentation area, the z-score method is used for the continuous variables of the number of damages and the area of damages, and the linear translation method is used to map the negative values to the non-negative interval to avoid negative values after normalization. The area of damages is normalized twice to maintain the data distribution pattern.

[0055] The damage segmentation area is processed using the Min-Max normalization method, and the formula I is used for calculation:

[0056] Damage segmentation area = damage segmentation area / 10 formula I

[0057] The number of body surface damages or the area of body surface damages is processed using the z-score normalization method and the linear translation method.

[0058] The z-score normalized value of the number of body surface damages is calculated according to formula II, or the z-score normalized value of the area of body surface damages is calculated according to formula III.

[0059] Z-score normalized value of the number of body surface damages = (number of body surface damages - μ) / σ formula II

[0060] Z-score normalized value of the area of body surface damages = (area of body surface damages - μ) / σ formula III

[0061] In formula II or formula III, μ is the average value of the number or area of damages, and σ is the standard deviation of the number or area of damages.

[0062] The number of body surface damages is calculated according to formula IV, and the area of body surface damages is calculated according to formula V.

[0063] The number of body surface injuries = z-score normalized value of the number of body surface injuries + |z-score minimum value| Formula IV; the area of body surface injuries = (z-score normalized value of the number of body surface injuries + |z-score minimum value|) / 4 Formula V;

[0064] In Formula IV, |z-score minimum value| is 0.882; in Formula V, |z-score minimum value| is 1.261. The correlation between the three variables is quantified by the Pearson correlation coefficient; when the correlation coefficient is greater than 0.8, it is considered to be redundant. In order to avoid the redundancy problem between variables, principal component analysis (PCA) is further used to compress the three redundant variables into a single comprehensive index to quantify the degree of body surface injury of each fish. The results show that the correlation coefficients of the number of injuries and the area of injuries are 0.9462 and 0.9453, respectively, showing a high positive correlation; at the same time, the correlation coefficient between the number of injuries and the area of injuries is also high, which is 0.7901, indicating that there is redundancy between the three variables. The damage segmentation area and the number and area of damage are standardized, and the covariance matrix and eigenvalue decomposition are calculated. According to the PC1 load score, the calculation formula of the comprehensive damage score is obtained.

[0065] In the present application, after obtaining the comprehensive damage score of the test largemouth bass, the damage degree of the test largemouth bass is preferably judged according to the scoring standard; the scoring standard is preferably divided into the following three levels:

[0066] 0< comprehensive damage score ≤0.53 is mild damage;

[0067] 0.53< comprehensive damage score ≤1.13 is moderate damage;

[0068] 1.13< comprehensive damage score ≤3.65 is severe damage.

[0069] In the present application, the body surface morphology of the largemouth bass with mild damage is preferably that part of the body surface front and body surface back appears red bleeding spots, and there is no obvious ulcer-like damage; the body surface morphology of the largemouth bass with moderate damage is preferably that the whole body surface, operculum and tail handle parts appear obvious red bleeding spots, and the number of ulcers increases; the body surface morphology of the largemouth bass with severe damage is preferably that the number of red bleeding spots on the whole body surface of the largemouth bass with moderate damage increases, and is accompanied by ulcer lesions and muscle necrosis. If a largemouth bass individual with intact body surface after being infected with iridovirus is obtained, it is determined as the undamaged level.

[0070] In the present application, based on the visual machine-assisted largemouth bass infected with iridovirus damage phenotype data, a largemouth bass iridovirus body surface damage level system is constructed, which not only provides a quantitative tool for disease research, but also provides reference data for early disease prevention and control of largemouth bass.

[0071] The method for identifying the damage grade of LMBV based on machine vision is described in detail below in combination with examples, but they should not be understood as limiting the protection scope of the present application.

[0072] Example 1

[0073] A technology for establishing a standard for quantitatively identifying the damage grade of LMBV based on machine vision. The main steps include the following:

[0074] I. Virus infection experiment

[0075] F1 generation population was selected based on the original population of LMBV introduced from the United States by the Freshwater Fisheries Research Center of Chinese Academy of Fishery Sciences, and F2 generation was obtained by self-breeding. Before the start of the infection experiment, the experimental fish were temporarily raised in a 4.5x8.8x1m indoor aquaculture pond for 2 weeks to adapt to the aquaculture environment. At the beginning of the experiment, 500 LMBV with good body condition, good appetite, similar size, and an average weight of (20.5±0.5) g were selected and transferred to a 2x4x1 cement pool. The water in the pool was filtered and disinfected, and was fully aerated to maintain the water temperature at 26±0.5℃. The immersion method was used to simulate natural infection. During the early breeding process, LMBV with natural disease were collected, and individuals with typical signs of Iridovirus disease were selected, including extensive ulceration on the body surface, fin base swelling and ulceration, and white liver, black spleen, and swollen kidney after dissection. Under sterile conditions, liver, spleen, and kidney samples were prepared into tissue homogenate, mixed, and used for virus identification, isolation, purification, and culture. After the sample was identified by Iridovirus nucleic acid detection kit and the sequencing of the amplification product, the individual was confirmed to be infected with Iridovirus after the sequence was consistent with that of Iridovirus.

[0076] After the sample homogenate was repeatedly frozen and thawed three times, it was centrifuged at 4℃ for 30 minutes at 4000 rpm per minute, and the supernatant was taken and filtered with a 0.22μm filter. The filtrate was stored at -80℃. Epithelioma of carp cells (EPC) were used to culture LMBV. After 200μL of filtrate was mixed with 800μL of M199 culture solution, it was inoculated into healthy EPC, adsorbed for 1h at 25℃, and then 4mL of M199 culture solution containing 2% FBS was added. After 1h, it was placed in a cell culture incubator containing 5% CO2 at 25℃ until the cell monolayer was 80% and lesions appeared. After the cells were frozen and thawed three times, the supernatant of the target virus was obtained by centrifugation. Based on the cytopathic effect of LMBV on EPC, the TCID 50 was determined to be 4x10 5 / mL. During the infection experiment, the cultured virus solution was evenly sprayed into the cement pool, and a low level of infection concentration was used to maintain the virus TCID50 Diluted to 40 / mL, 2 / 3 of the water was replaced every week, and the virus liquid was sprayed again. The water virus concentration was determined by TCID 50 method to keep the infection concentration stable. The temperature was maintained at 26±0.5℃, the dissolved oxygen was ≥7.5mg / L, the total ammonia nitrogen was ≤1mg / L, and the fish were fed twice a day (8:00 in the morning and 4:30 in the afternoon). The siphon method was used to clean up the leftover food and feces in time. The whole breeding cycle lasted for three months.

[0077] II. Growth index determination

[0078] After the infection breeding experiment, all experimental fish growth data were recorded after mild anesthesia with 200mg / L of MS-222. After the body surface was wiped dry with filter paper, the final body weight was measured with an electronic balance (accurate to 0.01g), and the final body length, body height and body thickness were measured with a scale (accurate to 0.01cm).

[0079] Growth index determination results of the virus infection breeding experiment

[0080] After the breeding experiment, the number of surviving individuals in the cement pool was 400, and the presence or absence of lesions such as spot ulceration, red bleeding spots, and cheek swelling on the body surface was used as the criterion for distinguishing standards. A total of 239 individuals with body surface lesions were selected, and 161 individuals with no body surface lesions were selected. The 161 individuals of Micropterus salmoides were determined to be healthy individuals, and the 239 individuals were determined to be diseased individuals. The growth data of the two groups are shown in Figure 2 . The body weight (124.93±33.27g), body length (16.79±1.37cm) and body height (5.04±0.50cm) of the diseased individuals were significantly lower than those of the healthy individuals, and the body thickness showed no significant change (see Figure 1 ). This indicates that the iridovirus can significantly reduce the growth performance of Micropterus salmoides and affect the normal development of individuals.

[0081] III. Damage data collection with visual machine assistance

[0082] After measuring the growth data of all experimental fish, the presence or absence of lesions such as spot ulceration, red bleeding spots, and cheek swelling on the body surface was used as the criterion for distinguishing standards. All individuals with body surface lesions were selected, the body surface was wiped dry, and the SONY ILCE-7RM3 camera was used to collect the lesion phenotype images of all individuals with body surface lesions in the small soft light shooting box (GODOX-LST60) with the parameters of wide circle F / 11, 75mm focal length and 1 / 200 seconds. To obtain complete individual phenotype data, four complete body surface phenotype images of Micropterus salmoides were taken, including the front of the body surface, the back of the body surface, the back of the body surface, and the abdomen of the body surface. A total of about 1000 lesion image data were collected.

[0083] IV. Damage Data Analysis.

[0084] SPSS 22.0 was used to statistically analyze growth data. After checking for homogeneity of variance, a t-test was used to compare the significance of differences between groups. P < 0.05 was considered significant. All data are expressed as mean ± standard deviation (Mean ± SD). ImageJ software was used to evaluate the image characteristics of the body surface injury. After converting the images to 8-bit, the injury areas were segmented into the front, back, dorsal, and ventral sides of the body surface. The frontal surface of the body is divided into three regions: region A extends from the front of the lower jaw to the posterior edge of the gill cover; region B extends from the posterior edge of the gill cover to the tip of the anal fin; and region C extends from the tip of the anal fin to the tip of the caudal fin. The underside of the body is also divided into three regions: region D extends from the tip of the caudal fin to the tip of the anal fin; region E extends from the tip of the anal fin to the posterior edge of the gill cover; and region F extends from the posterior edge of the gill cover to the front of the lower jaw. The dorsal surface is divided into three regions: region G extends from the tip of the caudal fin to the front of the second dorsal fin; region H extends from the front of the dorsal fin to the posterior end of the skull; and region I extends from the posterior end of the skull to the front of the lower jaw. The ventral surface is divided into three regions: region J extends from the tip of the caudal fin to the cloaca; region K extends from the cloaca to the posterior edge of the gill cover; and region L extends from the posterior edge of the gill cover to the front of the jaw. (See [reference needed]) Figure 2 After defining the damage zones, ImageJ software was used to define the scale distance based on the ruler. Then, the lasso tool was used to finely mark the area of ​​the damage region. The number of surface injuries and the area of ​​each damage region were calculated in detail. The number of damage regions for each individual was also counted for the subsequent construction of the damage level evaluation system.

[0085] Results of data collection and analysis of surface injury images

[0086] Data on the injury phenotypes of 239 individuals with body surface injuries were collected. First, the frequency of injury occurrence in 12 different segmentation regions of the 239 individuals with body surface injuries was statistically analyzed. Figure 3 As shown in Figure A, the highest number of injured individuals (192) occurred in area E, with a frequency of 80.3%. The frequency was also high (77.41%) in area B, while the lowest number (29) occurred in area L, with a frequency of only 12.1%. The frequencies in areas I and K were also relatively low (<15%). These results indicate that largemouth bass infected with iridovirus are more prone to injury in the lateral muscle and trunk areas, while the probability of injury in areas such as the head and abdomen is lower. Further analysis of the number of lesion zones in each individual was conducted to evaluate the extent of injury spread. Figure 3As shown in Figure B, the largest number of individuals (55, or 23.0%) had four damaged areas on their body surface. Similarly, 50 individuals had three damaged areas. The fewest individuals (only 3) had 11 damaged areas, and only 5 had only one damaged area. The number of damaged areas in the 239 individuals showed a normal distribution (P = 0.062). This indicates that the spread of body surface damage was relatively low after the largemouth bass contracted iridovirus in the early stages, which was beneficial for timely disease control. Simultaneously, the frequency of damaged areas on the four body surfaces (front, back, dorsal, and ventral) was analyzed. The results showed that the highest frequency was when all four body surfaces had one damaged area, with the highest frequency on the dorsal side (63.6%) and the lowest on the ventral side (30.1%). The frequencies of having three or zero damaged areas on the four body surfaces were both low. The highest frequency was when all three damaged areas were on the front (16.7%), and the highest frequency was when no damage was found on the ventral side (57.3%). Figure 3 (C). These results further indicate that the extent of surface damage caused by iridovirus infection during the early seedling stage is low, but damage is more likely to occur in important muscle and trunk areas.

[0087] ImageJ software was used to count the number and size of injuries in 239 individuals. A one-to-one correspondence was established between individuals with different injury segmentation numbers and the number and area of ​​injuries, such as... Figure 3 As shown in Figure D, individuals with 11 lesion divisions had an average of 42.7 lesions on their body surface. As the number of lesion divisions decreased, the number of lesions also decreased; individuals with only one lesion division had an average of only 1.4 lesions on their body surface. The area of ​​the lesion on the body surface reflects the severity of injury within a single area. Similar to the number of lesions, individuals with 11 lesion divisions had an average lesion area of ​​3.93 cm². 2 Ten lesion zones were identified in the individuals, with an average surface lesion area of ​​3.69 cm². 2 Individuals with a maximum of four injury areas on their body surface had an average injury area of ​​only 1.41 cm². 2 ( Figure 3 (E). Simultaneously, the number of injuries on the four body surfaces—front, back, dorsal, and ventral—was counted. The results showed that the more injury divisions there were, the greater the number of injuries. The front of the body had the most injuries (10.3), while the ventral side had the fewest (only 1.3), consistent with the results regarding the area of ​​injury within each division. Figure 3 (F). Overall, in the early seedling cultivation stage, the surface damage mainly consisted of small-scale spot ulceration and a small number of red hemorrhages, which provided an opportunity for early disease control measures.

[0088] Analysis of the characteristics of body surface injuries

[0089] The typical symptoms of the 12 segmentation areas of the four body surfaces, including the front, back, dorsal and ventral, were analyzed. The B and E areas, which had the highest frequency of injury, had a high density of red bleeding spots on the body surface, but few ulcerative lesions. In the C, D, G and J areas, muscle necrosis, scale shedding and large areas of ulcerative lesions were often observed, accompanied by a large area of redness and bleeding. In the H area, the dorsal fin also had large areas of tissue necrosis, red or white lesion nodules, and extensive viral infection. In the A, E, I and L areas, the head had a lower frequency of injury, but was more prone to redness and ulceration on the gill cover. In the K area, the abdominal region of the largemouth bass had a lower degree of injury, and the injury was only in the form of small and scattered red bleeding spots, which may be related to the fewer scales on the abdomen and the presence of more fat in the fish belly.

[0090] Five, establishment of injury grade discrimination criteria

[0091] To eliminate the dimensional differences between the three variables, the number of segmentation areas, the number of injuries and the area of injuries were normalized. The Min-Max normalization method was used for the number of segmentation areas. Considering that the number of injuries and the area of injuries are continuous variables, the z-score method was used for normalization. To avoid negative values after normalization, the linear translation method was used to map negative values to the non-negative interval. The area of injuries was normalized twice to maintain the data distribution pattern. To address the possible redundancy between the three variables, i.e., the more damaged areas, the more likely the number and area of injuries are larger, the Pearson correlation coefficient was used to quantify the correlation between the three variables. If the correlation coefficient is greater than 0.8, it is considered redundant. To avoid redundancy between variables, principal component analysis (PCA) was used to compress the three redundant variables into a single comprehensive index to quantify the degree of body surface injury of each fish and establish the injury grade discrimination criteria.

[0092] Establishment of injury grade discrimination criteria

[0093] The results showed that there was a high correlation between the injury segmentation area and the number and area of injuries. The Pearson correlation coefficient was used to quantify the correlation between the three variables. The results are shown in Table 2. Figure 5The correlation coefficients of the number of lesions and the lesion area were 0.9462 and 0.9453, respectively, as the independent variables of the lesion segmentation area, showing a high positive correlation. Similarly, the correlation coefficient of the number of lesions and the lesion area was also high, which was 0.7901, indicating that there was redundancy among the three variables. Therefore, PCA analysis was used to compress the three redundant variables into a single comprehensive index, quantify the degree of body surface injury of each fish, and construct the lesion grade discrimination standard. After standardizing the lesion segmentation area, the number of lesions and the lesion area, the PCA dimension reduction results are shown in Table 1.

[0094] Table 1 PCA dimension reduction analysis of three variable data after standardization analysis

[0095] Principal component Variance contribution rate Cumulative variance contribution rate PC1 72.30％ 72.30％ PC2 21.50％ 93.80％ PC3 6.20％ 100％ Variable PC1 load Explanation Damage area 0.59 Reflecting the damage spread range Damage number 0.63 Reflecting the damage density Damage area 0.68 Reflecting the single area damage severity

[0096] PC1 explained 72.3% of the variance, meeting the dimension reduction requirement. By calculating the covariance matrix and eigenvalue decomposition, the principal component load of PC1 was obtained, and PC1 was defined as the comprehensive injury degree. The results showed that the lesion area had the highest weight, with a load of 0.68, indicating that the single area lesion area had the most significant impact on the comprehensive score. This result was consistent with the high mortality rate of individuals with large area ulcer lesions on the body surface after suffering from iridovirus.

[0097] To eliminate the dimensional differences between the number of body surface injury areas, the number of body surface injuries and the lesion area, the three variables were normalized. The Min-Max normalization method was used for the segmentation area, considering that the number of lesions and the lesion area were continuous variables, the z-score method was used for normalization, and linear translation method was used to map negative values to non-negative interval. The lesion area was normalized twice to maintain the data distribution pattern.

[0098] The Min-Max normalization method calculated the lesion segmentation area according to formula I:

[0099] Lesion segmentation area = Lesion segmentation area / 10 Formula I;

[0100] The z-score normalization method calculated the z-score normalized value of the number of body surface injuries according to formula II, and the z-score normalization method calculated the z-score normalized value of the body surface injury area according to formula III:

[0101] z-score normalized value of the number of body surface injuries = (Number of body surface injuries - μ) / σ Formula II;

[0102] z-score normalized value of the body surface injury area = (Body surface injury area - μ) / σ Formula III;

[0103] Linear translation method calculates the body surface injury area according to formula IV:

[0104] Body surface injury number = z-score normalized value of body surface injury number + |z-score minimum value| formula IV; body surface injury area = (z-score normalized value of body surface injury number + |z-score minimum value|) / 4 formula V;

[0105] In formula II, μ is the average value of the number or area of injuries; σ is the standard deviation of the number or area of injuries.

[0106] In formula IV, |z-score minimum value| is 0.882; in formula V, |z-score minimum value| is 1.261.

[0107] According to the variance contribution rate of the above three indicators in Table 1, the comprehensive injury score is calculated according to formula I:

[0108] Comprehensive injury score = 0.59 x injury segmentation area + 0.63 x body surface injury number + 0.68 x body surface injury area formula VI.

[0109] Example 2

[0110] Calculation of body surface injury grade of Micropterus salmoides iridovirus

[0111] Taking randomly selected Micropterus salmoides infected with iridovirus in a culture pond as an example, the method of Example 1 was used to evaluate the body surface injury grade, and the results showed that the number of injury segmentation area was 7, the Min-Max normalization method of injury segmentation area was 0.7; the number of body surface injury was 8, the linear translation z-score normalization was 0.7; the injury area was 0.689 cm 2 , the linear translation z-score secondary normalization was 0.11.

[0112] The number of injury segmentation area calculation result: 7 / 10 = 0.7;

[0113] The number of body surface injury calculation result: [(8-9.775) / 9.952]+0.882 = 0.7;

[0114] The body surface injury area calculation result: {[(0.689-1.535) / 1.020]+1.261} / 4 = 0.11;

[0115] The standardized comprehensive injury score: 0.59 x 0.7 + 0.63 x 0.7 + 0.68 x 0.11 = 0.93.

[0116] According to the above method, the comprehensive injury scores of 239 Micropterus salmoides were calculated, and the values were sorted from high to low, and the quantile method was used to divide the comprehensive injury score into three grades:

[0117] i.e. 0 < comprehensive injury score ≤ 0.53 is mild injury;

[0118] 0.53 < comprehensive injury score ≤ 1.13 is moderate injury;

[0119] 1.13 < comprehensive injury score ≤ 3.65 is severe injury.

[0120] A method for establishing a large-mouth bass iridovirus injury grade discrimination standard is established according to the comprehensive injury score standard, and a grade injury discrimination schematic diagram is drawn. Figure 6 As shown in the schematic diagram, an individual with no damage on the body surface after infection with iridovirus is determined as non-injury grade; a small part of red bleeding points appear on the front and back of the body surface, and there is no obvious ulcer-like injury, which is determined as mild injury grade; a large area of body surface appears ulcer, and obvious red bleeding points appear on the gill cover and tail handle, which is determined as moderate injury grade; when a large range of red bleeding points appear on the body surface, accompanied by ulcer lesions, body surface muscle necrosis is determined as severe injury grade.

[0121] Based on the visual machine-assisted large-mouth bass infected with iridovirus injury phenotype data, a large-mouth bass iridovirus body surface injury grade system is constructed, which not only provides a quantitative tool for disease research, but also provides reference data for early disease prevention and control of large-mouth bass.

[0122] The above is only the preferred embodiment of the present application, it should be noted that for ordinary skilled in the art, without departing from the principles of the present application, can make a number of improvements and refinements, these improvements and refinements should also be considered as the protection scope of the present application.

Claims

1. A method for assisting in identifying the grade of Micropterus salmoides iridovirus damage based on machine vision, characterized in that, The method comprises the following steps: Collecting the data of the body surface lesions of the Micropterus salmoides infected with iridovirus under the assistance of visual machine, counting the number of lesion segmentation areas, the number of body surface lesions and the area of body surface lesions on the front surface of the body surface, the back surface of the body surface, the dorsal surface of the body surface and the ventral surface of the body surface of the Micropterus salmoides to be tested; The lesion segmentation areas are divided into 12 lesion segmentation areas by taking the gill cover and the end of the anal fin of the Micropterus salmoides as the boundary line to divide the front surface of the body surface, the back surface of the body surface, the dorsal surface of the body surface and the ventral surface of the body surface; The data of the number of lesion segmentation areas, the number of body surface lesions and the area of body surface lesions on the front surface of the body surface, the back surface of the body surface, the dorsal surface of the body surface and the ventral surface of the body surface of the Micropterus salmoides to be tested are normalized and substituted into the formula VI to calculate the comprehensive lesion score; The higher the comprehensive lesion score of the Micropterus salmoides to be tested is, the higher the lesion degree is. The number of lesion segmentation areas is processed by using the Min-Max normalization method, and the number of lesion segmentation areas after the Min-Max normalization method is calculated according to the formula I:

2. The method of claim 1, wherein, The number of lesion segmentation areas after the Min-Max normalization method = the number of lesion segmentation areas / 10 formula I; The number of body surface lesions or the area of body surface lesions is processed by using the z-score normalization method and the linear translation method; The z-score normalized value of the number of body surface lesions is calculated according to the formula II or the z-score normalized value of the area of body surface lesions is calculated according to the formula III; The z-score normalized value of the number of body surface lesions = (the number of body surface lesions - μ) / σ formula II; The z-score normalized value of the area of body surface lesions = (the area of body surface lesions - μ) / σ formula III; In the formula II or the formula III, μ is the average value of the number or area of lesions; σ is the standard deviation of the number or area of lesions; The number of body surface lesions is calculated according to the formula IV, and the area of body surface lesions is calculated according to the formula V; The number of body surface lesions = the z-score normalized value of the number of body surface lesions + |z-score minimum value| formula IV; The area of body surface lesions = (the z-score normalized value of the area of body surface lesions + |z-score minimum value|) / 4 formula V; In the formula IV, |z-score minimum value| is 0.882; in the formula V, |z-score minimum value| is 1.

261. The lesion segmentation areas are divided into 12 areas according to the following method:

3. The method of claim 1, wherein, The front surface of the body surface is divided into three areas, which are defined as A area from the front end of the lower jaw to the rear edge of the gill cover, B area from the rear edge of the gill cover to the end of the anal fin, and C area from the end of the anal fin to the end of the tail fin; The back surface of the body surface is divided into three areas, which are defined as D area from the end of the tail fin to the end of the anal fin, E area from the end of the anal fin to the rear edge of the gill cover, and F area from the rear edge of the gill cover to the front end of the lower jaw; The dorsal surface of the body surface is divided into three areas, which are defined as G area from the end of the tail fin to the front end of the second dorsal fin, H area from the front end of the dorsal fin to the rear end of the skull, and I area from the rear end of the skull to the front end of the lower jaw; ​ The body surface is divided into three regions, J region from the end of the tail fin to the cloaca, K region from the cloaca to the posterior edge of the operculum, and L region from the posterior edge of the operculum to the front end of the jaw.

4. The method of claim 1, wherein, The screening criteria of the largemouth bass infected with the iridovirus include the presence of spot ulcer, red bleeding spots and cheek redness lesions on the body surface.

5. The method of claim 1, wherein, The growth environment of the largemouth bass infected with the iridovirus is: temperature 25.5-26.5℃, dissolved oxygen ≥7.5mg / L, and total ammonia nitrogen ≤1mg / L.

6. The method of claim 1, wherein, The method for collecting the body surface damage data of the largemouth bass infected with the iridovirus by visual machine assistance is to take the complete body surface phenotype images of the front surface, back surface, dorsal surface and ventral surface of the largemouth bass to be tested, finely mark the damage area, and calculate the number and area of the body surface damage of each damage area.

7. The method of claim 6, wherein, The device for shooting includes a camera.

8. The method of claim 6, wherein, The software for finely marking the damage area includes ImageJ software.

9. The method of any one of claims 1 to 8, wherein, After obtaining the comprehensive damage score of the largemouth bass to be tested, the damage degree of the largemouth bass to be tested is determined according to the scoring standard. The scoring standard is divided into the following three levels: 0<comprehensive damage score ≤0.53 is mild damage; 0.53<comprehensive damage score ≤1.13 is moderate damage; 1.13<comprehensive damage score ≤3.65 is severe damage.

10. The method of claim 9, wherein, The body surface morphology of the largemouth bass with mild damage is that part of the red bleeding spots appear on the front surface and back surface of the body surface, and there is no obvious ulcer-like damage; The body surface morphology of the largemouth bass with moderate damage is that obvious red bleeding spots appear on the whole body surface, operculum and caudal fin, and the number of ulceration increases; The body surface morphology of the largemouth bass with severe damage is that the number of red bleeding spots on the whole body surface increases compared with the body surface morphology of the largemouth bass with moderate damage, and is accompanied by ulcer lesions and muscle necrosis.

Images