CN112861872A

CN112861872A - Penaeus vannamei phenotype data determination method, device, computer equipment and storage medium

Info

Publication number: CN112861872A
Application number: CN202011638959.XA
Authority: CN
Inventors: 高广春; 郑一琨; 栾生; 代平; 孔杰
Original assignee: Hangzhou Feirui Technology Co ltd; Yellow Sea Fisheries Research Institute Chinese Academy of Fishery Sciences; Zhejiang University City College ZUCC
Current assignee: Hangzhou Feirui Technology Co ltd; Yellow Sea Fisheries Research Institute Chinese Academy of Fishery Sciences; Zhejiang University City College ZUCC
Priority date: 2020-12-31
Filing date: 2020-12-31
Publication date: 2021-05-28

Abstract

The invention provides a method, a device, computer equipment and a storage medium for measuring phenotypic data of penaeus vannamei boone, wherein the method comprises the following steps: collecting side and overlook image samples of the penaeus vannamei boone; performing target detection on the top view image and the side view image, determining the specific position of the penaeus vannamei boone, obtaining bounding box information of the penaeus vannamei boone, and cutting out the image of the penaeus vannamei boone as input of the next step; detecting key points of the Penaeus vannamei Boone based on a Hourglassnet deep learning network model, and acquiring coordinate information of the key points; constructing a posture framework structure of the penaeus vannamei boone in water by using the extracted coordinate information of the key points, and performing posture estimation on the penaeus vannamei boone; calculating phenotype data of the penaeus vannamei: and calculating by combining three-dimensional space transformation and proportion parameters according to the obtained coordinate information of the key points to obtain the phenotypic data of the penaeus vannamei boone. The phenotypic data of the penaeus vannamei boone in the invention has no strict constraint in the measurement process, thereby enhancing the robustness of the algorithm.

Description

Penaeus vannamei phenotype data determination method, device, computer equipment and storage medium

Technical Field

The invention belongs to the field of computer vision, realizes measurement based on computer vision, and particularly relates to a method and a device for measuring phenotype data of penaeus vannamei boone, computer equipment and a storage medium.

Background

For a long time, the traditional method for measuring phenotypic data of penaeus vannamei boone mainly utilizes manual and manual contact measurement to measure relevant parameters such as external dimension, weight and the like. However, this method is very susceptible to subjective factors such as manual experience, habits, and professional skill level, and external environment interference, that is, there are usually not small errors between the results measured by different measuring personnel in different environments, and the system error of the measurement result cannot be controlled, so that the measurement result is unbalanced, the standard is inconsistent, and the error rate is high. On the other hand, the method of manually measuring phenotypic data requires a lot of manpower and time, and the detection process is time-consuming and labor-consuming. Meanwhile, the aquatic organisms are highly sensitive and easily damaged in the manual contact type measurement process, and even disease transmission and water environment pollution can be caused to influence the normal growth of the aquatic organisms. In the market, a few methods for measuring fish phenotype data based on computer vision are available, most methods are water-leaving measurement, and measurement indexes are few, so that not only can a large stress response be still brought to the fish, but also more limiting conditions are provided.

Since twenty-first century, with the rapid development of computer information technology, optical imaging technology, image processing algorithm and other technologies, computer vision algorithms based on machine learning have played an extremely important role in various fields, and have been widely used in the field of automatic animal identification research due to their advantages such as non-contact, high accuracy and quantifiability. Therefore, in the face of the needs of the aquaculture industry, the traditional manual contact measurement method is gradually replaced by an automatic detection method based on computer vision technology. Compared with the traditional manual contact type measurement method, the computer vision processing method is faster, more objective and more accurate. In the field of measuring visually related attributes of aquatic organisms using computer vision techniques, a number of researchers have conducted a series of studies involving aquatic organisms such as fish, shrimp, crab, shellfish.

However, due to the characteristics of the algorithm, the operation speed of the traditional computer vision algorithm is greatly improved compared with that of manual measurement, but the requirement of the aquaculture industry for developing more rapidly still cannot be met. Meanwhile, the aquatic organism measuring system based on the traditional image processing algorithm and machine learning is poor in robustness and poor in adaptability to the transformed environment and the form transformation of aquatic organisms. On the other hand, although the computer vision algorithm can eliminate errors caused by artificial subjectivity during artificial measurement and ensure that the measurement results have consistent standards, the computer vision algorithm is limited by algorithm performance and often cannot achieve higher precision.

The shrimp is one of the most important parts in aquatic products, and in shrimp culture, information such as the shape, size, color, texture and the like of a shrimp body is very important basic information, so that the growth condition of the shrimp can be visually expressed, culture personnel can be helped to feed, screen, classify and the like more conveniently, and meanwhile, the shrimp scientific research personnel can be helped to perform relevant basic research. For a long time, the traditional method for measuring the phenotypic data of the water product mainly utilizes manual and manual contact measurement to measure the relevant parameters of the size, weight and the like of the water product. However, this method is very susceptible to subjective factors such as manual experience, habits, professional skill level, and external environmental interference, resulting in unbalanced measurement results, inconsistent standards, and high error rate. On the other hand, the method of manually measuring phenotypic data requires a lot of manpower and time, and the detection process is time-consuming and labor-consuming. Meanwhile, the aquatic organisms are highly sensitive and easily damaged in the manual contact type measurement process, and even disease transmission and water environment pollution can be caused to influence the normal growth of the aquatic organisms.

Therefore, the research on the rapid, objective, accurate and non-contact automatic measurement method plays a crucial role in the continuous healthy and efficient development of the aquaculture industry. The problems of the traditional method for measuring phenotypic data of the penaeus vannamei boone can be summarized as follows:

1. the detection method needs a large amount of manpower and material resources, and the efficiency is low as the measurement index is increased.

2. The method is easily influenced by subjective factors such as experience, habit, preference and the like and external environment interference, so that the detection process is time-consuming and labor-consuming, the detection result is strong in subjectivity, poor in consistency and high in error rate, and a unified standard cannot be formed.

3. Because certain stress reaction exists after the prawn individuals are dehydrated and the prawn individuals are dehydrated for a long time, the contact type measuring method can cause certain influence on the activity of the prawn individuals and cause certain interference on subsequent breeding research.

Due to the strong correlation of key point water production morphological parameters and its easy detectability in image processing algorithms, still image based key point detection algorithms have begun to be applied in the research of morphological measurements of aquatic organisms since the 80 s of the 20 th century. Irving, etc. takes images of the side of a fish as the fish swims through a particular channel, detects key points at the head and tail of the fish by means of image processing, calculates the length of the fish, and predicts the weight of the fish based on the length of the fish. Lines and the like subtract two adjacent pictures at a time interval, a crescent shape of a fish head part in an obtained image is generated due to the movement of the fish, a binary mode classifier with high robustness is used for identifying key points of the fish head, a Point Distribution Model (PDM) is used for identifying key points of the boundary of the fish according to the approximate position and direction of the fish body, phenotype data of the fish is calculated, and the weight of the fish is predicted according to the length of the fish. The method comprises the steps of collecting images of shrimps through a camera, performing smoothing processing through a threshold segmentation algorithm and morphological on-off operation, extracting key points of the shrimps, constructing a contour curve, calculating the length of the shrimps, performing posture recognition on the shrimps by utilizing wavelet transform low-pass filtering, determining the head and tail orientation of the shrimps, performing a recognition algorithm on the head and body connection points of the shrimps, and calculating the head length and the body length. These preliminary studies demonstrate the feasibility of computer vision techniques in the production and cultivation of aquatic organisms and the advantages over traditional manual measurements. However, most of the algorithms have the characteristics of poor robustness, unstable precision, strict environmental requirements and the like, and are difficult to apply to practical research.

In summary, the research of the method for measuring phenotype data of penaeus vannamei boone is mainly manual measurement, and is realized by adopting a computer vision algorithm in the field of prawn processing, but no corresponding research result exists in water-carrying unconstrained measurement, and the measurement of the phenotype data of the penaeus vannamei boone in the prior art has certain problems, such as the following:

1. in the traditional image processing, the method has a certain research on the measurement of prawn phenotype data by acquiring an image contour, searching a key point on the contour by constructing a specific characteristic and acquiring final phenotype data through the key point, but due to the particularity of a prawn body, the traditional characteristic extraction has difficulty, and the traditional method is not easy to realize living body measurement.

2. Most of the manual phenotypic data measuring methods at the present stage are dehydration measurement, have certain damage to the activity of the prawns, and do not have a general and specific measuring scheme.

3. The parameters measured by adopting the manual measuring method are not many, and the significance of subsequent research guidance is less.

Therefore, how to improve the determination precision and the detection speed of the phenotypic data of the penaeus vannamei boone is a problem to be solved in the field.

Disclosure of Invention

The invention aims to provide a method for measuring phenotype data of penaeus vannamei boone aiming at the defects of the prior art, which takes the penaeus vannamei boone as a research object and realizes the measurement of the non-contact phenotype data of the penaeus vannamei boone in an unconfined state in water.

In order to achieve the purpose, the invention adopts the following technical scheme:

a method for measuring phenotypic data of Penaeus vannamei Boone is characterized by comprising the following steps:

s1, collecting side and overhead image samples of the penaeus vannamei boone;

s2, performing target detection on the top view image and the side view image by adopting a yolo3 algorithm, determining the specific position of the penaeus vannamei boone, obtaining bounding box information of the penaeus vannamei boone, and cutting out the image of the penaeus vannamei boone as the input of the next step;

s3, detecting key points of the Penaeus vannamei Boone based on the Hourglassnet deep learning network model, and acquiring coordinate information of the key points;

s4, constructing a posture framework structure of the penaeus vannamei boone in water by using the extracted coordinate information of the key points, and performing posture estimation on the penaeus vannamei boone;

s5, calculating phenotype data of the penaeus vannamei: and calculating by combining three-dimensional space transformation and proportion parameters according to the obtained coordinate information of the key points to obtain the phenotypic data of the penaeus vannamei boone.

Further, the step S1 includes: after the penaeus vannamei boone is placed in the water box filled with 3/4 water, the water box is placed in an image sample collecting device, industrial cameras are arranged above and on the left side of the device, the industrial cameras are started, and side faces and overlooking image samples of the penaeus vannamei boone placed in the water box are collected.

Further, the step S3 includes: after detecting bounding box frame information of the penaeus vannamei boone, inputting the cut penaeus vannamei boone image into a Hourglassnet network to obtain different key point positioning coordinates of a top view and a side view; wherein, 19 key points need to be detected in the side view and are arranged according to the position sequence of the points; in top view, 23 key points are detected.

Further, in step S3, the information of the key points of the top view is obtained through a top view detection module, the top view detection module includes two cascaded 4-order hourglass modules and a relay supervision module, and each hourglass module retains original information through residual addition by a residual module, so that the hourglass modules can learn features at different scales and retain original features as much as possible; the relay supervision module conducts intermediate prediction on each hourglass module, namely calculating the training loss after each hourglass module to help the training of the hourglass modules in the later stage.

Further, in step S3, the key points of the side view are obtained by the side view detection module, which replaces the relay monitoring module and the hourglass module in the top view detection module with the subnets connecting high resolution to low resolution in parallel, so as to maintain the accuracy of the key point detection by performing multi-scale repeated fusion by repeatedly exchanging information on the parallel multi-resolution subnets while greatly reducing the model parameters.

Further, the step S5 includes: extracting the length of the point by combining three-dimensional transformation according to the information of the key point obtained from the side view and the top view; on the top view, the key point coordinate information of the top view is segmented, the segmentation curve angle of the top view of the penaeus vannamei boone is extracted, on the side view, the obtained penaeus vannamei boone curve segmentation angle is obtained by combining the top view, the pixel value of the penaeus vannamei boone phenotype data is calculated, then the scale information of a scale in water is obtained by combining an experiment, and the final penaeus vannamei boone phenotype data is calculated.

Further, the phenotypic data of the penaeus vannamei boone comprise body length, full length, head length, body width and body height.

The invention also provides a penaeus vannamei phenotype data measuring device, which is characterized by comprising the following components:

the data acquisition unit is used for acquiring side and overhead image samples of the penaeus vannamei boone;

the target detection unit is used for determining the specific position of the penaeus vannamei boone, obtaining bounding box information of the penaeus vannamei boone and cutting out an image of the penaeus vannamei boone as the input of the next step;

the key point positioning unit is used for detecting key points of the Penaeus vannamei Boone based on the Hourglassnet deep learning network model and acquiring coordinate information of the key points;

the posture skeleton construction unit is used for constructing a posture skeleton structure of the penaeus vannamei in water by utilizing the extracted coordinate information of the key points, and carrying out posture estimation on the penaeus vannamei;

and the phenotype data calculation unit is used for calculating phenotype data of the penaeus vannamei boone by combining three-dimensional space transformation and proportion parameters according to the acquired coordinate information of the key points.

The invention also provides a computer device, which is characterized by comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to realize the method.

The invention also provides a storage medium, characterized in that the storage medium stores a computer program which, when executed by a processor, implements the method as described above.

Compared with the prior art, the method and the device for measuring the phenotypic data of the penaeus vannamei boone disclosed by the invention have the following advantages:

1. the body of the penaeus vannamei boone is completely and naturally unfolded in water, the position of the key point of the phenotype data is more accurate, no matter the penaeus vannamei boone is manually measured, or the penaeus vannamei boone is separated from water and is subjected to computer vision measurement, the penaeus vannamei boone body can be extruded to different degrees, and the coordinate of the initial position of the phenotype data is influenced.

2. In the invention, the living body acquires 5 types of phenotype data in a non-contact way, so that the stress response of the penaeus vannamei boone is reduced as much as possible; when the method is used for measuring the phenotype data of the prawns, the prawns are in a natural swimming state in water, the whole measuring process of the phenotype data is completed within 1 second, and the stress response caused by contact is reduced.

3. The measuring method does not restrict the posture of the shrimp body, allows the shrimp body to freely move in the water box, can accurately acquire 5 types of phenotypic data of the penaeus vannamei at one time even if the shrimp body is bent, and has strong algorithm robustness. Therefore, the phenotypic data of the penaeus vannamei boone in the invention has no strict constraint in the measurement process, and the robustness of the algorithm is enhanced.

4. The invention expands a deep learning network model for detecting key points of human posture to a phenotype determination system of the penaeus vannamei boone, provides 42 penaeus vannamei boone key point detections, provides a phenotype detection algorithm of the penaeus vannamei boone combining a HourglassNet model and an HR net model, and improves the detection precision and the detection speed.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

FIG. 1 is a diagram of a prawn phenotype data research index provided by an embodiment of the present invention;

fig. 2 is a schematic structural diagram of a device for collecting images of a top view and a side view of penaeus vannamei boone according to an embodiment of the present invention;

fig. 3 is a top view key point mark of penaeus vannamei provided by an embodiment of the present invention;

fig. 4 is a side view key point mark of penaeus vannamei provided by the embodiment of the present invention;

FIG. 5 is a schematic diagram of an overall network structure for detecting key points according to an embodiment of the present invention, wherein (a) an image is read; (b) a target detection module; (c) (ii) a Obtaining accurate target positioning through target detection; (d) a key point detection module; (e) the final output result comprises the key points of the shrimps and the movement postures of the shrimps;

FIG. 6 is a schematic diagram illustrating an overall network structure of a top view detection module according to an embodiment of the present invention;

FIG. 7 is an improved version of an hourglass module of a top view inspection module provided in accordance with embodiments of the present invention;

fig. 8 is a relay supervision module in the top view detection module according to an embodiment of the present invention;

FIG. 9 is a side view detection module provided in accordance with an embodiment of the present invention;

fig. 10 is a probability heatmap generated according to the shrimp keyword correspondence provided in the embodiment of the present invention, wherein: (a) the key points correspond to the original images; (b) a generated probability heatmap;

fig. 11 is a key point detection model training parameter provided in the embodiment of the present invention, in which: (a) a top view detection model training process; (b) a side view detection model training process;

fig. 12 is a key point detection result provided in the embodiment of the present invention, in which: (a) and (b) detecting the image to be detected.

FIG. 13 is a three-dimensional transformation diagram in an embodiment of the invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

It will be understood that the terms "comprises" and/or "comprising," when used in this specification and the appended claims, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.

The invention is further described with reference to the following drawings and specific examples, which are not intended to be limiting.

It is also to be understood that the terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in the specification of the present invention and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

It should be further understood that the term "and/or" as used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

Interpretation of terms:

the key points are as follows: some positioning points required by measuring the phenotype data of the penaeus vannamei are called key points in the patent of the invention.

Phenotypic data: and in the growth process of the penaeus vannamei boone, data indexes related to the form of the penaeus vannamei boone need to be measured.

High flux: and a plurality of index measurements are rapidly completed at one time.

The present embodiment is a method for measuring phenotypic data of penaeus vannamei boone, the research object of the present invention is penaeus vannamei boone, and the indexes of the research are shown in fig. 1, including body length (AL), full length (BL), head length (CL), body width (CW), and body height (CH). The method specifically comprises the following steps:

s1, collecting side and overhead image samples of the penaeus vannamei boone;

because the posture of the living penaeus vannamei boone in water is not kept in a linear state at any time, most states are bent, and the phenotype data of fish cannot be accurately and effectively measured through one camera, therefore, the top view image and the side view image of the penaeus vannamei boone individual in the closed box are respectively extracted through the two cameras. Specifically, after the penaeus vannamei boone is placed in a water box filled with 3/4 water, the water box is placed in an image sample collection device, industrial cameras are arranged on the upper side and the left side of the device, the industrial cameras are started, and side faces and overlooking image samples of the penaeus vannamei boone placed in the water box are collected.

In order to obtain the phenotypic data of the shrimps efficiently and precisely, the shrimps in the top view and the side view are respectively calibrated at key points in the embodiment, which is shown in fig. 3 and 4.

The respective point indications are shown in tables 1 and 2 below.

TABLE 1 prawn Top-view Key Point notation

Serial number	Name of point	Description of the location points
			1	head	Edge of shrimp head
2-3	b1_up(down)	Upper (lower) edge point of crossing of shrimp head and shrimp body
			4-13	bn_up(down)	The upper (lower) edge point of the front shrimp and the back shrimp
14-15	b7_up(down)	Upper (lower) edge point of crossing shrimp body and shrimp tail
			16-17	tail_up(down)	Upper (lower) edge point of shrimp tail
18-19	eye_up(down)	Upper edge point of shrimp eye and shrimp head intersection
			20-21	bw_up(down)	Upper (lower) edge point of widest part of shrimp head
22	tail	Marginal point of central axis of shrimp body far away from shrimp head
			23	mid	Midpoint of curve of first section shrimp

TABLE 2 prawn side elevation Key Point Mark Specification

Serial number	Name of point	Description of the location points
			1	head	Edge of shrimp head
2-3	b1_up(down)	Upper (lower) edge point of crossing of shrimp head and shrimp body
			4-13	bn_up(down)	The upper (lower) edge point of the front shrimp and the back shrimp
14-15	b7_up(down)	Upper (lower) edge point of crossing shrimp body and shrimp tail
			16	tail)	Marginal point of shrimp tail
17	eye	The rightmost end of the round shape of the shrimp eye
			18-19	bw_up(down)	Upper (lower) edge point of widest part of shrimp head

Therefore, the integrated image acquisition system is built in the embodiment, 10000 shrimp images at the overlooking angle and the side viewing angle are respectively acquired, and the shrimp images are completely marked based on the marking mode and are used as the training data set of the embodiment.

in this embodiment, after bounding box frame information of a penaeus vannamei boone is detected, inputting a cut image of the penaeus vannamei boone into a Hourglassnet network to obtain different key point location coordinates of a top view and a side view; wherein, 19 key points need to be detected in the side view and are arranged according to the position sequence of the points; in top view, 23 key points are detected.

In the embodiment, based on the characteristics of the top view and the side view of the penaeus vannamei boone in the image acquisition hardware, the balance of the operation precision and the operation efficiency is considered, a novel shrimp detection algorithm is designed, the whole model architecture is shown in fig. 5, and the whole model architecture is divided into a target detection submodule, a top view detection submodule and a side view detection submodule.

Fig. 6 shows that the top view detection module includes two cascaded 4-step hourglass modules and a relay supervision module. Each hourglass module is composed of residual error modules through the structure shown in fig. 7, original information is reserved through residual error addition, so that the hourglass module can learn characteristics under different scales, and original characteristics are reserved as much as possible.

Meanwhile, in order to stack a plurality of hourglass modules in the whole network structure and prevent the gradient disappearance problem caused by the network depth, the network can continuously repeat the process of extracting the characteristics from bottom to top and from top to bottom so as to improve the reliability of the network. We use the relay supervision module to make an intermediate prediction of each hourglass module, i.e. to calculate the training loss after each hourglass module, to aid the training of the hourglass modules at the later stage, see fig. 8.

Different from the top view detection module, in this embodiment, the side view key points have a simple structure and a number greatly reduced compared to the top view, and in order to further speed up the overall efficiency of the algorithm, a lighter-weight network structure is used in the side view detection module, as shown in fig. 9, the side view detection module replaces the relay supervision module and the hourglass module in the top view detection module with subnets that connect high resolution to low resolution in parallel, and maintains the accuracy of key point detection by performing multi-scale repeated fusion by repeatedly exchanging information on parallel multi-resolution subnets while greatly reducing the model parameters.

In the embodiment, the length of the point is extracted by combining three-dimensional transformation according to the information of the key point obtained from the side view and the top view; on the top view, the key point coordinate information of the top view is segmented, the segmentation curve angle of the top view of the penaeus vannamei boone is extracted, on the side view, the obtained penaeus vannamei boone curve segmentation angle is obtained by combining the top view, the pixel value of the penaeus vannamei boone phenotype data is calculated, then the scale information of a scale in water is obtained by combining an experiment, and the final penaeus vannamei boone phenotype data is calculated.

Specifically, phenotype data are obtained by combining three-dimensional space transformation and a proportion parameter according to the obtained key points. Since the acquired points are obtained from a side view and a top view, it is necessary to extract the lengths of the points in combination with three-dimensional transformation, three-dimensional space transformation display, and three-dimensional space actual length calculation, see fig. 13.

In the three-dimensional transformation shown in fig. 13, OA is the real length of the object in the three-dimensional space, OA 'is the projection of the line OA on the plane YOZ, α is the angle between the line OA' and the OY direction axis, and β is the angle between the projection of OA on the plane XOY and the OY direction axis. From this, the actual length of OA can be calculated as:

from the information of the points obtained from the side and top views, in combination with the three-dimensional spatial variations, the following is calculated:

comprehensively processing the points obtained from the side view, on the top view, segmenting the key point coordinate information of the top view to extract the segmentation curve angle of the top view to the shrimp body, on the side view, combining the prawn body curve segmentation angle obtained from the top view to calculate the prawn phenotype data pixel value, then combining the experiment to obtain the scale information of the scale in the water to calculate the final prawn body phenotype data.

In addition, in order to train and obtain a high-robustness and high-precision key point detection model, 4000 pictures are used as a training set for training in the embodiment, wherein the resolution of the size of an input image is 512X512, and a 128X128 probability heatmap is generated by using corresponding true value coordinates. In this embodiment, the probability heatmap is used to represent the coordinates of the key points, and the probability value of each pixel point on the generated probability heatmap is calculated through the multivariate gaussian model, that is, the closer to the key point, the closer to 1 the corresponding probability value is, see fig. 10.

In the training and reasoning process, the embodiment uses a device configured as a 3.70GHz Intel i7-8700K CPU and NVIDIA RTX2080 SUPER graphics card as our training device. In the whole training process, the accuracy of the training set and the accuracy of the verification set change, as shown in fig. 11; the prediction results based on the trained model are shown in fig. 12. Due to the characteristics of a large number of top view keys, a large number of structural changes and the like, the top view detection model is slower in training speed than a side view, and the accuracy of a final verification set is slightly lower than that of the side view. Finally, when training is finished, the detection accuracy of the key points in the top view verification set reaches 97.57%, and the detection accuracy of the key points in the side view verification set reaches 98.61%.

In addition, in order to verify the algorithm stability of the embodiment, based on the result of the key point detection and the three-dimensional coordinate transformation equation, 20 penaeus vannamei data samples were measured in the embodiment, and the relative average error of the measurement is shown in table 3. The stability of the algorithm was then assessed by simultaneously measuring 10 times per shrimp and recording the error of 10 measurements. The present example only shows 10 shrimp measurements for 2 shrimps, as shown in tables 4 and 5, and the algorithm of the present example was found to be stable and reliable by the evaluation of the overall measured data.

TABLE 320 mean relative error of measurements

Number of times

AL

BL

CL

CW

CH

1

2.10％

0.83％

10.80％

1.40％

21.30％

2

4.90％

0.90％

5.10％

1.90％

14.40％

3

5％

3.30％

11.40％

0.90％

13..6％

4

2.90％

1.50％

10.20％

1.50％

25％

5

2.60％

0.50％

4.60％

5.50％

6.60％

6

1％

4.60％

6.90％

4.80％

7

1.30％

2.17％

6.70％

2.20％

2.60％

8

5.48％

1.30％

7.80％

0％

13.60％

9

3.50％

2.15％

12.70％

5.96％

6.70％

10

4.80％

1.70％

14.70％

15.40％

8.50％

11

0.90％

2.80％

19.30％

1.40％

8.80％

12

1.90％

2.25％

5.96％

1.60％

11.80％

13

3.07％

2.98％

20.30％

1.88％

10.18％

14

2.29％

1.23％

1.14％

2.83％

8.50％

15

6.17％

1.79％

16.40％

17.80％

9.07％

16

5.37％

1.40％

8.32％

5.79％

8.09％

17

6.78％

3.48％

17.28％

1.64％

4.13％

18

3.97％

1.61％

22.09％

0.96％

0.42％

19

3.20％

0.68％

4.43％

5.29％

4.44％

20

5.99％

1.77％

7.57％

10.24％

13.83％

Mean error

3.66％

1.77％

10.60％

4.55％

9.10％

TABLE 4 prawn phenotype 10 test results (1)

No	AL	BL	CL	CW	CH
							1	50.66	85.61	20.85	13.37	14.64
2	51.69	84.2	18.74	13.19	15.13
						3	51.18	84.71	18.74	13.33	14.66
4	49.72	84.57	19.64	13.35	16.42
						5	52.05	86.68	20.04	13.07	15.12
6	51.96	86.71	19.16	13.07	14.5
						7	52.04	85.75	18.83	13.07	14.75
8	52.03	85.94	18.42	13.07	14.63
						9	52.44	85.96	18.42	13.07	14.89
10	52.47	86.02	18.42	13.07	14.65
						Average of 10 measurements	51.624	85.615	19.126	13.166	14.939
Average error of 10 measurements	1.28％	0.78％	3.33％	0.88％	2.48％
						Measured value	53.5	87.5	21.9	14	14
Relative error	3.50％	2.15％	12.70％	5.96％	6.70％

TABLE 5 prawn phenotype 10 test results (2)

No	AL	BL	CL	CW	CH
							1	62.99	101.5	21.66	15.83	18.91
2	63.19	100.47	21.2	15.25	19.26
						3	63.18	100.93	20.9	14.7	19.39
4	63.07	101.26	21.33	14.7	19.27
						5	63.1	100.61	21.43	14.43	19.39
6	63.06	101.36	21.45	14.43	19.39
						7	63.22	101.29	21.22	14.76	19.51
8	64.02	101.53	20.66	15.03	19.39
						9	63.2	101.9	21.82	15	19.86
10	64.09	100.73	20.56	14.76	19.86
						Average of 10 measurements	63.312	101.158	21.223	14.889	19.423
Average error of 10 measurements	0.47％	0.37％	1.48％	2.09％	0.99％
						Measured value	62	102	23.8	15.1	16
Relative error	2.10％	0.83％	10.80％	1.40％	21.30％

In summary, according to the method for measuring phenotype data of penaeus vannamei boone based on deep learning, the method realizes phenotype data measurement based on a key point detection idea. The method can realize the measurement of the body length, the full length, the head length, the body height and the body width of the penaeus vannamei boone, and the error of the full length and the body length is strictly controlled within 5 percent of the error, thereby meeting the requirement of practical application. Meanwhile, the system can track the movement of the shrimps in real time in the measuring process, the phenotype data is measured within 1s, and the requirement of rapid and accurate acquisition in the production process is met.

This embodiment also provides a south america white shrimp phenotype data assay device, includes:

It should be noted that, as will be clearly understood by those skilled in the art, the device for estimating the posture of the penaeus vannamei and measuring the phenotype data of the penaeus vannamei and the specific implementation process of each unit block may refer to the corresponding description in the foregoing method embodiment, and for convenience and brevity of description, no further description is provided herein.

The penaeus vannamei phenotype data measuring device can be realized in the form of a computer program which can run on a computer device.

The computer device includes a processor, a memory, and a network interface connected by a system bus, where the memory may include a non-volatile storage medium and an internal memory.

The non-volatile storage medium may store an operating system and a computer program. The computer program includes program instructions that, when executed, cause a processor to perform a method for measuring phenotypic data of penaeus vannamei.

The processor is used to provide computational and control capabilities to support the operation of the overall computer device.

The internal memory provides an environment for running a computer program in the nonvolatile storage medium, and the computer program can enable the processor to execute a penaeus vannamei phenotype data measurement method when being executed by the processor.

The network interface is used for network communication with other devices. Those skilled in the art will appreciate that the above-described computer device configurations are merely part of the configurations associated with the present application and do not constitute a limitation on the computer devices to which the present application may be applied, and that a particular computer device may include more or less components than those shown in the figures, or may combine certain components, or have a different arrangement of components.

The processor is used for running a computer program stored in the memory, and the program realizes the method for measuring phenotypic data of the penaeus vannamei boone in the embodiment one.

It should be understood that in the embodiments of the present Application, the Processor may be a Central Processing Unit (CPU), and the Processor may also be other general purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components, and the like. Wherein a general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

It will be understood by those skilled in the art that all or part of the flow of the method implementing the above embodiments may be implemented by a computer program instructing associated hardware. The computer program includes program instructions, and the computer program may be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the flow steps of the embodiments of the method described above.

The invention also provides a storage medium. The storage medium may be a computer-readable storage medium. The storage medium stores a computer program, wherein the computer program, when executed by the processor, causes the processor to execute a method for measuring phenotypic data of penaeus vannamei.

The storage medium may be a usb disk, a removable hard disk, a Read-Only Memory (ROM), a magnetic disk, or an optical disk, which can store various computer readable storage media.

Those of ordinary skill in the art will appreciate that the elements and algorithm steps of the examples described in connection with the embodiments disclosed herein may be embodied in electronic hardware, computer software, or combinations of both, and that the components and steps of the examples have been described in a functional general in the foregoing description for the purpose of illustrating clearly the interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.

In the embodiments provided in the present invention, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative. For example, the division of each unit is only one logic function division, and there may be another division manner in actual implementation. For example, various elements or components may be combined or may be integrated into another system, or some features may be omitted, or not implemented.

The steps in the method of the embodiment of the invention can be sequentially adjusted, combined and deleted according to actual needs. The units in the device of the embodiment of the invention can be merged, divided and deleted according to actual needs. In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a storage medium. Based on such understanding, the technical solution of the present invention essentially or partially contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a terminal, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention.

It is to be noted that the foregoing is only illustrative of the preferred embodiments of the present invention and the technical principles employed. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, although the present invention has been described in greater detail by the above embodiments, the present invention is not limited to the above embodiments, and may include other equivalent embodiments without departing from the spirit of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A method for measuring phenotypic data of Penaeus vannamei Boone is characterized by comprising the following steps:

s1, collecting side and overhead image samples of the penaeus vannamei boone;

2. The method for determining phenotypic data of Penaeus vannamei Boone according to claim 1, wherein the step S1 includes: after the penaeus vannamei boone is placed in the water box filled with 3/4 water, the water box is placed in an image sample collecting device, industrial cameras are arranged above and on the left side of the device, the industrial cameras are started, and side faces and overlooking image samples of the penaeus vannamei boone placed in the water box are collected.

3. The method for determining phenotypic data of Penaeus vannamei Boone according to claim 1, wherein the step S3 includes: after detecting bounding box frame information of the penaeus vannamei boone, inputting the cut penaeus vannamei boone image into a Hourglassnet network to obtain different key point positioning coordinates of a top view and a side view; wherein, 19 key points need to be detected in the side view and are arranged according to the position sequence of the points; in top view, 23 key points are detected.

4. The method for determining phenotypic data of Penaeus vannamei Boone according to any one of claims 1-3, wherein in step S3, information of key points of a top view is obtained through a top view detection module, the top view detection module comprises two cascaded 4-order hourglass modules and a relay supervision module, and each hourglass module retains original information through residual error addition by a residual error module, so that the hourglass modules can learn characteristics at different scales and retain original characteristics as much as possible; the relay supervision module conducts intermediate prediction on each hourglass module, namely calculating the training loss after each hourglass module to help the training of the hourglass modules in the later stage.

5. The penaeus vannamei phenotype data assay method according to claim 4, wherein in step S3, the key points of the side view are obtained by a side view detection module, and the side view detection module replaces a relay supervision module and an hourglass module in the top view detection module with subnets connecting high resolution to low resolution in parallel, so that the accuracy of key point detection is maintained by repeated fusion of multiple scales through repeated exchange of information on parallel multi-resolution subnets while the model parameters are substantially reduced.

6. The method for determining phenotypic data of Penaeus vannamei Boone according to claim 1, wherein the step S5 includes: extracting the length of the point by combining three-dimensional transformation according to the information of the key point obtained from the side view and the top view; on the top view, the key point coordinate information of the top view is segmented, the segmentation curve angle of the top view of the penaeus vannamei boone is extracted, on the side view, the obtained penaeus vannamei boone curve segmentation angle is obtained by combining the top view, the pixel value of the penaeus vannamei boone phenotype data is calculated, then the scale information of a scale in water is obtained by combining an experiment, and the final penaeus vannamei boone phenotype data is calculated.

7. The method for determining phenotypic data of Penaeus vannamei Boone according to claim 1, wherein the phenotypic data of Penaeus vannamei Boone includes body length, full length, head length, body width and body height.

8. A south america white shrimp phenotype data assay device which characterized in that includes:

9. A computer device, characterized in that the device comprises a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to realize the penaeus vannamei phenotype data determination method according to any one of claims 1 to 7.

10. A storage medium storing a computer program which, when executed by a processor, implements the method for determining phenotypic data of penaeus vannamei according to any one of claims 1 to 7.