Intelligent method for separating broad-skinworms

Abstract

The present application belongs to the technical field of broad earthworm sorting, and particularly relates to an intelligent broad earthworm sorting method, which comprises the following steps: step one: data collection of the broad earthworms bred in the earthworm factory is performed through a 4K camera; step two: the density of the broad earthworms in the current range is automatically obtained through an image processing target detection algorithm, the area range of the broad earthworms is located, and the accurate positioning of the broad earthworms is realized; step three: the length of the broad earthworms is calculated; step four: comparison and analysis of the length of the broad earthworms obtained in step three and the broad earthworm growth standard set set in advance by industry experts are performed, and the current broad earthworms are judged to belong to a certain stage of growth, so as to obtain comprehensive processing opinions and sort the broad earthworms. The present application can solve the problems of low accuracy of the traditional naked eye judgment method and low work efficiency due to the need of a large amount of manpower for measuring with a ruler, and has a good market application prospect.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of wide earthworm sorting, and particularly relates to an intelligent wide earthworm sorting method. BACKGROUND

[0002] Wide earthworm, commonly known as earthworm, is an important biological body with extremely high medical and medicinal value. Although the wild wide earthworm has high medical and medicinal value, the quantity is small and cannot meet the demand. Although the production quantity is improved by using the field cultivation production mode, the annual production is still low, and the field required for production occupies a large area. Only by taking the feed raw material as the center and establishing the earthworm factory, can the problem of insufficient production be solved by intensive cultivation. Through the standardized cultivation mode, the wide earthworm processing and quality standard system is constructed to solve the problem of mixed quality of wide earthworm and the problem of using inferior goods.

[0003] With the rapid development of artificial intelligence technology, through the combination with the artificial intelligence technology, many industries are more automated and intelligent. Many tasks in the biological field can be well completed by the artificial intelligence algorithm, which greatly improves the work efficiency and improves the production of wide earthworm.

[0004] In order to determine the size of the wide earthworm by the intelligent artificial intelligence algorithm, the traditional method is usually determined by the naked eye, which has low accuracy and great subjectivity. The accurate measurement by the ruler measurement often needs to consume a large amount of manpower, and the work efficiency is low. In the face of large-scale cultivation of wide earthworm, this method is not desirable. The present application provides an intelligent method for accurately positioning the position information of the wide earthworm, obtaining the coordinate information of the rectangular detection frame in the image, and calculating the size of the wide earthworm in the image, which can greatly improve the work efficiency.

[0005] The information disclosed in this BACKGROUND section is only for the purpose of increasing the understanding of the background of the present application and should not be taken as an acknowledgment or any form of suggestion that this information forms prior art that is publicly known. SUMMARY

[0006] The purpose of the present application is to provide an intelligent wide earthworm sorting method to solve the problems of low accuracy of the traditional naked eye determination method and consumption of a large amount of manpower by the ruler measurement and low work efficiency.

[0007] In order to achieve the above-mentioned purpose, the present application provides the following technical scheme:

[0008] An intelligent wide earthworm sorting method comprises the following steps:

[0009] Step one: data acquisition module, through the 4k camera to the data of the earthworm factory to cultivate the broad earthworm, to obtain a certain video segment P = {p1, p2..., p n} in the cultivation of broad earthworm growth standard set A = {a1, a2,..., a n} is set in advance by industry experts;

[0010] Step two: broad earthworm detection module: through the target detection algorithm of image processing to automatically obtain the density of broad earthworm in the current range, locate the area range of broad earthworm, realize the accurate positioning of broad earthworm;

[0011] The video segment P is input to the target detection algorithm, and the identification process is as follows:

[0012] Q = Transformer(P, θ)

[0013] Wherein, θ is the learnable weight parameter of Transformer model, and Transformer model represents that P obtains the image set Q = {q1, q2,..., q n} of each frame of picture of broad earthworm in video segment;

[0014] The pictures of picture set Q are transformed one by one by convolution operator, and new feature set

[0015] Q * = W q Q+b q

[0016] Wherein, Q represents picture set, W q represents the weight value of picture set Q, and b q represents the bias value of picture set;

[0017] The target position of broad earthworm in each picture in picture set Q = {q1, q2,..., q n} is obtained by the target detection algorithm, and the rectangular detection box box of broad earthworm is accurately obtained;

[0018] Step three: calculate the length of broad earthworm, according to the rectangular detection box box of broad earthworm in image obtained in step two, the coordinates of four points of rectangular detection box are (x1, y1), (x2, y2), (x3, y3) (x4, y4); the diagonal length of rectangular detection box box can be calculated through coordinate information The length of rectangle is l = |x1-x2| or |y1-y2|, the size r of broad earthworm can be obtained by calculating the coordinate information; the value range of r is l≤r≤m, that is, the size length of broad earthworm;

[0019] Step 4: Comprehensive analysis. The length r of *Geoceratops guangdi* obtained in Step 3 is compared with the pre-defined growth standard set A = {a1, a2, ..., a...} set by industry experts. n By conducting comparative analysis, we can determine which stage of growth the Guangdi dragon is currently in, and thus obtain comprehensive treatment recommendations, sort the Guangdi dragons, and carry out scientific breeding.

[0020] As a preferred option, the specific operation for obtaining the rectangular detection box of the ground dragon in step two is as follows:

[0021] (1) Use the self-Transformer model to find the key feature information S of the giant earthworm in the image; S = self-Transformer(Q * C, W), where C = (c1, c2, ..., c n ) represents the location feature information of Guangdilong, W is the weight of Guangdilong's location, and S represents the key feature information of Guangdilong;

[0022] (2) Feature fusion The fusion features T of the *Geoceratops guangensis* are obtained, where f(.) represents the target detection function, W is the weight of the *Geoceratops guangensis*'s position, S represents the key feature information of the *Geoceratops guangensis*, and C = (c1, c2, ..., c...). n ) represents the location feature information of the Globedra chinensis, and bias represents the function bias value;

[0023] (3) Predictive analysis obtains the location information of *Geoceratops guangdi*, and based on the fusion feature T and feature set Q... * This yields a rectangular detection box. in This represents the weight value of each image in image Q. This represents the bias value for each image.

[0024] Traditional methods of sorting earthworms rely on subjective human judgment of size for selection and rearing. This makes it difficult to reasonably evaluate each earthworm, and the subjective nature of the judgment can easily lead to misjudgments and omissions. When raising large numbers of earthworms, using manual sorting methods results in a large workload and the possibility of omissions and incorrect selections.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] The intelligent broad earthworm sorting method of the application can objectively analyze the broad earthworm according to the sorting standard set by industry experts in advance, and give objective judgment and processing opinions, thereby playing an auxiliary role in manual judgment and reducing the disadvantages of individual experience bias caused by individual judgment of the size of the broad earthworm. Through rapid collection of data of the broad earthworm, intelligent sorting of the broad earthworm, and reasonable breeding of the broad earthworm, the workload can be greatly reduced. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 is a general flowchart of the application;

[0028] Figure 2 is a schematic diagram of a broad earthworm detection module of the application;

[0029] Figure 3 is a schematic diagram of a rectangular detection frame coordinate of the broad earthworm. DETAILED DESCRIPTION

[0030] The technical solutions of the application will be described below in a clear and complete manner. Obviously, the described embodiments are part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application.

[0031] Example 1

[0032] Refer to the accompanying Figures 1-3 When a large area of broad earthworms are bred, the size and length of the broad earthworms can be automatically measured by using the application scheme for analysis and research, and the whole process is intelligent and automatic, which mainly includes the following steps

[0033] Step 1: The data acquisition module collects the data of the broad earthworms bred in the earthworm factory through a 4k camera, and obtains a certain video segment P={p1, p2,..., p n} of the broad earthworms in the breeding according to the pre-set broad earthworm growth standard set A={a1, a2,..., a n};

[0034] Step 2: The broad earthworm detection module automatically obtains the density of the broad earthworms in the current range through the target detection algorithm of image processing, locates the range of the broad earthworms, and realizes accurate positioning of the broad earthworms;

[0035] The video segment P is input into the target detection algorithm, and the identification process is as follows:

[0036] Q=Transformer(P, θ)

[0037] Where θ represents the learnable weight parameters of the Transformer model, and the Transformer model represents the set of images Q = {q1, q2, ..., q} of the earth dragon in each frame of the video segment P. n};

[0038] By performing feature transformations on each image in image set Q using convolution operators, a new feature set is obtained.

[0039] Q * =W q Q+b q

[0040] Where Q represents the image set, W q b represents the weight value of the image set Q. q Indicates the bias value of the image set;

[0041] The image set Q = {q1, q2, ..., q3} was obtained using an object detection algorithm. n In each image of the image, the target location of the *Geoceratops fasciatus* is accurately determined, and its rectangular detection box is obtained. The specific operation is as follows:

[0042] (1) Use the self-Transformer model to find the key feature information S of the giant earthworm in the image; S = self-Transformer(Q * C, W), where C = (c1, c2, ..., c n ) represents the location feature information of Guangdilong, W is the weight of Guangdilong's location, and S represents the key feature information of Guangdilong;

[0043] (2) Feature fusion The fusion features T of the *Geoceratops guangensis* are obtained, where f(.) represents the target detection function, W is the weight of the *Geoceratops guangensis*'s position, S represents the key feature information of the *Geoceratops guangensis*, and C = (c1, c2, ..., c...). n ) represents the location feature information of the Globedra chinensis, and bias represents the function bias value;

[0044] (3) Predictive analysis obtains the location information of *Geoceratops guangdi*, and based on the fusion feature T and feature set Q... * This yields a rectangular detection box. in This represents the weight value of each image in image Q. This represents the bias value for each image;

[0045] Step three: calculate the length of the broad-skinworm, according to the rectangular detection box of the broad-skinworm in the image obtained in step two, the coordinates of the four points of the rectangular detection box are (x1, y1), (x2, y2), (x3, y3) and (x4, y4), the diagonal length of the rectangular detection box can be calculated through the coordinate information The length of the rectangle is l=|x1-x2| or |y1-y2|, and the size r of the broad-skinworm can be obtained by calculating the coordinate information; the value range of r is l<=r<=m, that is, the size length of the broad-skinworm;

[0046] Step four: comprehensive research and judgment, compare the length r of the broad-skinworm obtained in step three with the broad-skinworm growth standard set A={a1, a2,..., a n} set by industry experts in advance, judge and analyze the current broad-skinworm belonging to a certain stage of growth, so as to obtain comprehensive processing opinions and sort the broad-skinworms.

[0047] The intelligent broad-skinworm sorting method of the application is a strategic method, which collects the video data of the broad-skinworm in the breeding growth process, converts it into a frame by frame image by using an algorithm, determines the position information of the broad-skinworm in the image, and intelligently sorts the broad-skinworm by calculating the length of the broad-skinworm, solving the problems of low accuracy of traditional naked eye judgment method and large amount of manpower and low work efficiency required by ruler measurement.

[0048] The application relates to image recognition technology and deep neural network technology, and the position information of the broad-skinworm is obtained by using an artificial intelligence algorithm to obtain a rectangular detection box; the value range of the length of the broad-skinworm is calculated according to the coordinate information of the rectangular detection box; and the broad-skinworm is intelligently comprehensively judged and analyzed according to the value range.

[0049] The method can quickly collect data of the broad-skinworm, intelligently sort the broad-skinworm, reasonably breed the broad-skinworm, greatly reduce the workload, reduce the misjudgment caused by personal empiricism bias, and is very practical and has good market application prospect.

[0050] The foregoing description of specific exemplary embodiments of the application is for the purpose of illustration and example. These descriptions are not intended to limit the application to the precise forms disclosed, and it will be apparent that many modifications and variations are possible in light of the above teachings. The exemplary embodiments are chosen and described in order to explain the principles of the application and its practical application, thereby enabling others skilled in the art to implement and utilize the application in various embodiments and various modifications as are suited to the particular use contemplated. It is intended that the scope of the application be defined by the claims and their equivalents.

Claims

1. An intelligent method for separating broadloper, characterized in that, Comprising the following steps: Step one: through the 4K camera to the earthworm factory to breed the data collection of the earthworm, to get the video segment P = {p1, p2... p n} of the earthworm in the breeding; the earthworm growth standard set set by industry experts in advance is A = {a1, a2,..., a n} Step two: automatically obtain the density of Guangdilong in the current range through the target detection algorithm of image processing, locate the area range of Guangdilong, and realize the accurate positioning of Guangdilong; Input the video segment P into the target detection algorithm, and the identification process is as follows: Q=Transformer(P, theta) wherein θ is a learnable weight parameter of the Transformer model, and the Transformer model represents a picture collection Q = {q1, q2,..., qT} of each frame of the image in which P obtains the Tianjiang Lizard in the video clip. n} The pictures of the picture set Q are transformed by a convolution operator one by one to obtain a new feature set Q * = W q Q + b q wherein Q represents a picture set, W q represents a weight value of the picture set Q, b q represents a bias value of the picture set; The target position of Guangdilong in each picture in a picture set Q = {q1, q2,..., q n} is obtained through a target detection algorithm, and a rectangular detection box box of Guangdilong is accurately obtained. Step three: calculate the length of the broad earthworm, according to the rectangle detection box box of the broad earthworm obtained in step two, the coordinates of the four points of the rectangle detection box are (x1, y1), (x2, y2), (x3, y3) (x4, y4); the diagonal length of the rectangle detection box box is calculated through the coordinate information The length of the rectangle is l=|x1-x2| or |y1-y2|; through the calculation of the coordinate information, the length r of the broad earthworm can be obtained, and the value range of r is l≤r≤m; Step four: comprehensive research and judgment, compare the length r of Guangdilong obtained in step three with the Guangdilong growth standard set A = {a1, a2,..., an} set by industry experts in advance, and judge and analyze the current Guangdilong belonging to a certain stage of growth, so as to obtain comprehensive processing opinions and sort Guangdilong. n} are compared and analyzed, and the current Guangdilong is judged and analyzed to belong to a certain stage of growth, so as to obtain comprehensive processing opinions and sort Guangdilong.

2. The intelligent separation method of broadloper according to claim 1, characterized in that, The specific operation of obtaining the rectangular detection box box of Guangdilong in step two is as follows: (1) Finding the key feature information S of the broad-skinworm in the image through the self-Transformer model; S=self-Transformer(Q * , C, W); wherein C=(c1, c2, …, c n ) represents the position feature information of the broad-skinworm; W is the weight of the position of the broad-skinworm, and S represents the key feature information of the broad-skinworm; (2) Feature fusion The fusion feature T of the D. japonica is obtained. wherein f(.) represents a target detection function, W is a weight of the location of Guangdilong, S represents key feature information of Guangdilong; C=(c1, c2,..., c n ) represents location feature information of Guangdilong; bias represents a function bias value; (3) The position information of the broad earthworm is obtained by prediction analysis, and the fusion feature T and the feature set Q are obtained according to the position information * , a rectangular detection box box is obtained, wherein represents the weight value of each image Q, represents the bias value of each image.

Images