WO2020030052A1

WO2020030052A1 - Animal count identification method, device, medium, and electronic apparatus

Info

Publication number: WO2020030052A1
Application number: PCT/CN2019/099817
Authority: WO
Inventors: 王怀庆
Original assignee: 京东数字科技控股有限公司
Priority date: 2018-08-10
Filing date: 2019-08-08
Publication date: 2020-02-13
Also published as: CN110826581A; CN110826581B

Abstract

Embodiments of the present disclosure provide an animal count identification method, a device, a medium, and an electronic apparatus. The method comprises: acquiring a pre-processed animal image, wherein the pre-processed animal image comprises animals and animal marker points; generating, on the basis of the animal marker points, a digital matrix having the same size as the pre-processed animal image; comparing the digital matrix against a pre-determined training matrix, and generating an animal count identification model on the basis of a comparison result acquired from the comparison; and performing animal count identification on an image on the basis of the animal count identification model. The technical solution of the embodiments in the present disclosure enables accurate and real-time understanding of a total count of farm animals, and achieves accurate calculation of an animal count, thereby ensuring data accuracy, providing directly visible data, and saving labor costs. (FIG. 1)

Description

Animal quantity identification method, device, medium and electronic equipment

Technical field

The present disclosure relates to the field of image-based animal recognition technology, and in particular, to a method for identifying animal numbers and an animal recognition device.

Background technique

In recent years, artificial intelligence technology has been slowly introduced into the pig breeding process, using big data technology to analyze the data correlation analysis in the breeding process, including quantitative identification, herd behavior analysis, disease identification and early warning, and no-weighing. Target the value of big data pigs.

For a complete pig house monitoring system, pig identification and points are the most basic core part. Pig identification refers to identifying the positions of all pigs in the image in the input image. Pig point artificial intelligence technology can greatly Improve pig raising efficiency and save a lot of labor costs. It is difficult to realize intelligent technology, and related technology research is relatively small.

The basic methods for identifying the number of pigs in a pig farm are as follows:

(1) By selecting empty pens and applying green paint around them, and then identifying non-green areas as pig areas.

(2) Detecting moving pigs through an improved inter-frame difference method, mainly detecting targets based on motion information.

(3) Pig color recognition based on color space clustering model pig identification algorithm.

The above prior art solutions have the following disadvantages:

(1) By coating the empty pens with pigsty, this method is feasible in the experimental test stage. In the actual scenario, the number of pigs is huge, it is difficult to operate, and the site cost is large.

(2) The interframe difference method, taking into account the habit of the pig, may not move for a long time, and the application of motion information monitoring will fail.

(3) Pig color recognition algorithm based on color space clustering model. Due to the complex texture of the pigs and the relatively simple color of the pigs, the similarity between black pigs and pig pens will cause inaccurate recognition and low recognition rate. .

It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute the prior art known to those of ordinary skill in the art.

Summary of the invention

The purpose of the embodiments of the present disclosure is to provide a method for identifying animal numbers and an animal identification device, thereby at least to a certain extent overcoming the limitations and disadvantages of related technologies to determine the number of animals, which is difficult to operate and the animals are not It is difficult to accurately identify the number of animals and the color of the animals. The method of identifying the number of animals is easily affected by the environment and one or more problems of inaccurate identification.

Other features and advantages of the embodiments of the present disclosure will become apparent from the following detailed description, or may be learned in part through the practice of the present disclosure.

According to a first aspect of the embodiments of the present disclosure, a method for identifying an animal quantity includes:

Obtaining a pre-processed animal image, wherein the pre-processed animal image includes an animal and a marker point of the animal;

Generating a digital matrix having the same size as the pre-processed animal image based on the marked points of the animal;

Comparing the number matrix with a preset training matrix, and generating an animal number recognition model based on the comparison result obtained by the comparison;

The identification of the number of animals in the image is realized based on the animal number recognition model.

In an embodiment of the present disclosure, the above-mentioned generating a digital matrix having the same size as the pre-processed animal image based on the marked points of the animal includes:

Determining a position coordinate of the marked point in the pre-processed animal image according to the marked point of the animal;

Generating a digital matrix of the same size as the pre-processed animal image, wherein the digital matrix is composed of numbers in the range of 0 to 1;

A corresponding position is determined in the digital matrix based on the position coordinates, and a marked point corresponding to the position coordinate is represented by the number 1, the remaining positions in the digital matrix are represented by 0, and a preset Gaussian kernel is used The function performs Gaussian blur processing to obtain a digital matrix corresponding to the pre-processed animal image.

In an embodiment of the present disclosure, the above-mentioned comparing the digital matrix with a preset training matrix, and generating an animal number recognition model based on the comparison result obtained by the comparison, include:

Operate the digital matrix according to a preset convolution rule to obtain a convolution matrix;

Comparing the convolution matrix with the preset training matrix to obtain a comparison result;

Correct the preset recognition model according to the comparison result;

The recognition accuracy rate of the modified recognition model is verified. When the recognition accuracy rate of the recognition model is greater than or equal to a preset threshold, an animal number recognition model is generated.

In an embodiment of the present disclosure, the above-mentioned identifying the number of animals based on the animal number recognition model includes:

Input the acquired image of the number of animals to be identified into the animal number recognition model, and obtain a digital matrix corresponding to the image of the number of animals to be identified;

Sum the digital matrix to obtain the number of animals in the image of the number of animals to be identified.

According to a second aspect of the embodiments of the present disclosure, an animal identification device is provided, including: an acquisition module, a first generation module, a second generation module, and an identification module; wherein,

An acquisition module for acquiring a pre-processed animal image, wherein the pre-processed animal image includes an animal and a marker point of the animal;

A first generating module, configured to generate a digital matrix having the same size as the pre-processed animal image based on the marked points of the animal;

A second generating module, configured to compare the digital matrix with a preset training matrix, and generate an animal number recognition model based on the comparison result obtained by the comparison;

A recognition module is configured to recognize the number of animals in an image based on the animal number recognition model.

In an embodiment of the present disclosure, the first generating module is specifically configured to:

Generating a digital matrix of the same size as the pre-processed animal image, wherein the digital matrix is composed of numbers in the range of 0 and 1;

In an embodiment of the present disclosure, the second identification module is specifically configured to:

Correct the preset recognition model according to the comparison result;

In an embodiment of the present disclosure, the second identification module is further configured to:

According to a third aspect of the embodiments of the present disclosure, there is provided a computer-readable medium having stored thereon a computer program that, when executed by a processor, implements the method for identifying an animal quantity as described in the first aspect of the above embodiment .

According to a fourth aspect of the embodiments of the present disclosure, there is provided an electronic device including: one or more processors; a storage device for storing one or more programs, and when the one or more programs are used by the one When executed by one or more processors, the one or more processors are caused to implement the animal quantity identification method according to the first aspect in the foregoing embodiment.

The technical solutions provided by the embodiments of the present disclosure may include the following beneficial effects:

In the technical solution provided by some embodiments of the present disclosure, a pre-processed animal image is obtained by obtaining the pre-processed animal image, and the pre-processed animal image includes an animal and a marker point of the animal; Pre-processing a digital matrix of the same size as the animal image; comparing the digital matrix with a preset training matrix, and generating an animal number recognition model based on the comparison result obtained by the comparison; implementing the animal in the image based on the animal number recognition model Identification of quantity. The technical solution of the embodiment of the present disclosure can accurately grasp the total number of animals on the farm in real time, accurately calculate the number of animals, ensure the accuracy of the data, and be intuitive and visible, and save a lot of labor costs.

It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, and should not limit the present disclosure.

BRIEF DESCRIPTION OF THE DRAWINGS

The drawings herein are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure, and together with the description serve to explain the principles of the present disclosure. Obviously, the drawings in the following description are only some embodiments of the present disclosure. For a person of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. In the drawings:

FIG. 1 schematically illustrates a flowchart of an animal number recognition method according to an embodiment of the present disclosure;

FIG. 2 schematically illustrates a pre-processed image according to an embodiment of the present disclosure;

FIG. 3 schematically illustrates a flow chart of an animal number recognition method applied to identifying individual pigs according to an embodiment of the present disclosure; FIG.

4 schematically illustrates a block diagram of an animal identification device according to an embodiment of the present disclosure;

FIG. 5 is a schematic structural diagram of a computer system suitable for implementing an electronic device according to an embodiment of the present disclosure.

detailed description

Example embodiments will now be described more fully with reference to the accompanying drawings. However, the example embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, the embodiments are provided so that this disclosure will be more comprehensive and complete, and the concepts of the example embodiments will be fully conveyed To those skilled in the art.

Furthermore, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, many specific details are provided to give a full understanding of the embodiments of the present disclosure. However, those skilled in the art will realize that the technical solutions of the present disclosure may be practiced without one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other instances, well-known methods, devices, implementations or operations have not been shown or described in detail to avoid obscuring aspects of the present disclosure.

The block diagrams shown in the drawings are merely functional entities and do not necessarily correspond to physically separate entities. That is, these functional entities may be implemented in the form of software, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor devices and / or microcontroller devices. entity.

The flowcharts shown in the accompanying drawings are only exemplary descriptions, and it is not necessary to include all contents and operations / steps, nor are they necessarily performed in the order described. For example, some operations / steps can also be decomposed, and some operations / steps can be merged or partially merged, so the order of actual execution may be changed according to the actual situation.

FIG. 1 schematically illustrates a flowchart of an animal number recognition method according to an embodiment of the present disclosure.

Referring to FIG. 1, a method for identifying an animal quantity according to an embodiment of the present disclosure includes the following steps:

Step S110, obtaining a pre-processed animal image;

Step S120: Generate a digital matrix with the same size as the pre-processed animal image based on the marked points of the animal;

In step S130, the digital matrix is compared with a preset training matrix, and an animal number recognition model is generated based on the comparison result obtained by the comparison;

Step S140: Recognize the number of animals in the image based on the animal number recognition model.

The technical solution of the embodiment shown in FIG. 1 can calculate artificially marked images processed by artificial intelligence algorithms, sum the obtained matrix to calculate the number of animals, and accurately calculate the number of animals to ensure the accuracy of the data. It is intuitive and visible, which saves a lot of labor costs.

The implementation details of each step shown in FIG. 1 are described in detail below:

In step S110, a pre-processed animal image is acquired.

In one embodiment of the present disclosure, the pre-processed animal image includes at least the animal and the marked points of the animal.

In an embodiment of the present disclosure, the labeled points of the animals in the pre-processed animal image may be artificially labeled points of the animals in the animal picture, and a pre-processed animal image labeled with the animal is obtained and provided to the subsequent machine recognition process. Therefore, it is necessary to pre-process the marked points in the animal image in a one-to-one correspondence with the animals, and the situation of missing animals or heavy animals cannot occur.

FIG. 2 schematically illustrates a pre-processed image according to an embodiment of the present disclosure.

As shown in FIG. 2, the corresponding labeled points are marked on each animal in the pre-processed animal image, and there are no missing marks and weight-weighted animals. That is, the pre-processed animal image is an accuracy image, which can be used for subsequent Improving the accuracy of machine recognition algorithms provides a good basis.

In step S120, a digital matrix having the same size as the pre-processed animal image is generated based on the marked points of the animal.

In an embodiment of the present disclosure, the position coordinates of the marker point in the pre-processed animal image are determined according to the marker points of the processed animal image, and a digital matrix of the same size as the pre-processed animal image is generated, where the number matrix ranges from 0 to The number structure within the range of 1; the corresponding position is determined in the number matrix based on the position coordinates, and the marked point corresponding to the position coordinates is represented by the number 1, the remaining positions in the number matrix are represented by 0, and a preset Gaussian kernel is used The function performs Gaussian blur processing to obtain a digital matrix corresponding to the pre-processed animal image.

In an embodiment of the present disclosure, the marker points in the pre-processed animal image are extracted and extracted to the coordinate positions of the marker points. For subsequent training of the machine recognition model, the pre-processed animal images are converted into a matrix to form a A digital matrix composed of numbers in the range of 0 to 1 with the same size of the animal image is processed. The coordinate position of the marker point is set to 1 in the digital matrix, and the remaining elements are set to 0.

In step S130, the digital matrix is compared with a preset training matrix, and an animal number recognition model is generated based on the comparison result obtained by the comparison.

In an embodiment of the present disclosure, before the digital matrix is compared with a preset training matrix, the digital matrix needs to be convolved, and the digital matrix is operated according to a preset convolution rule to obtain a convolution matrix; The convolution matrix is compared with a preset training matrix to obtain a comparison result; the preset recognition model is modified according to the comparison result; the recognition accuracy of the modified recognition model is verified, and when the recognition model is recognized When the accuracy is greater than or equal to a preset threshold, an animal number recognition model is generated.

In an embodiment of the present disclosure, the convolution rule may be a convolution operation of a digital matrix and a Gaussian kernel function through a convolution kernel function to generate a new convolution matrix to enhance the signal in the preprocessed image corresponding to the digital matrix. Characteristics, and reduce noise and other interference factors.

In an embodiment of the present disclosure, the comparison result obtained by comparing the convolution matrix with a preset training matrix can train a preset recognition model, and the comparison needs to be continuously used in the process of constructing the recognition model. Results The parameters in the recognition model were modified, and the animal number recognition model was obtained after the modification.

In an embodiment of the present disclosure, each parameter in the above recognition model is in an initial state, and a comparison result obtained after the first training of the recognition model is performed to check the recognition model once to ensure the subsequent recognition model. The iteration result is correct. For example, the first pass calculation of the loss value (comparison result) can be checked once. First, taking CIFAR-10 as an example, if you use the Softmax classifier, the prediction can get an initial loss of about 2.302 (because For the 10 categories, the initial probability should not be 0.1, and the Softmax loss is: -log (probability of the correct category): -ln (0.1) = 2.302); Second, set the regularization coefficient to a normal small value, and add it back to regularization Term, and then calculate the loss / loss; finally, before training the large data set, you can train a small data set (such as 20 pictures), and then see if your neural network can achieve 0 loss, because If the implementation of the neural network is correct, in the case of no regularization term, it is completely able to overfit this small part of the data; after starting the training of the recognition model, you can monitor some The indicator understands the state of training. For example, it can be determined by the change of the loss value after each full round of iteration. The appropriate learning rate can ensure that after each round of full training, the loss value is reduced, and it can be reduced to a relatively high level after a period of time. Small degree; then you need to track the accuracy status on the training set and the validation set to determine the state of the classifier. As time progresses, the accuracy on the training set and the validation set will increase. After the accuracy reaches a certain level, the difference between the two is relatively large, then it should be noted that it may be an over-fitting phenomenon. If the difference is not large, it indicates that the recognition model is in good condition. The last thing to pay attention to is the weight. The ratio of the update magnitude to the current weight magnitude is preferably checked independently for each set of parameters.

In one embodiment of the present disclosure, based on the foregoing scheme, after determining that the training gradient of the recognition model is implemented correctly, it uses its update weight parameter in the backward propagation algorithm. The most common weight update method is SGD + Momentum, or RMSProp adaptive The learning rate update algorithm, etc .; then, the learning rate is attenuated in different ways. Mind the attenuation of the learning rate by the following common methods: (1) Step attenuation: every full round of training cycles (all pictures pass through once) ), The learning rate drops a little, (2) exponential level decay: need to customize the hyperparameters and iteration rounds, (3) 1 / t decay: need to customize the hyperparameters and iteration rounds; then use cross-validation to search And find the most suitable hyperparameters. Among them, for large deep neural networks, it takes a lot of time to train, so before this, you need to do a hyperparameter search to determine the best setting. The most direct way is to During the implementation of the framework, design a continuous transformation of hyperparameters for optimization, and record the validation set under each full training iteration for each hyperparameter. And effect. In practical applications, these hyperparameters are determined in neural networks. Generally, n-fold cross-validation is rarely used, and a fixed cross-validation set is generally used. Finally, the obtained recognition model is model-fused and generated. An identification model for the number of animals whose recognition accuracy is greater than or equal to a preset threshold. The model fusion may be to retain several intermediate model weights and the last model weight, average them, and then test the results on the cross-validation set. It is usually one or two percentage points higher than the results of directly trained models. The intuitive understanding is that for the bowl-shaped structure, there are many times when our weights are jumping around the lowest point, and we can't really reach the lowest point, and averaging the positions near the two lowest points, there will be more A high probability falls closer to the lowest point.

In an embodiment of the present disclosure, the acquired image of the number of animals to be identified is input to an animal number recognition model, and a digital matrix corresponding to the image of the number of animals to be identified is obtained; the digital matrix is summed to obtain the animals in the image of the number of animals to be identified. quantity.

An embodiment of the present disclosure provides a method for identifying animal numbers by acquiring a pre-processed animal image, where the pre-processed animal image includes labeled points of the animal and the animal; and based on the labeled points of the animal, a digital matrix having the same size as the pre-processed animal image is generated. ; Compare the digital matrix with a preset training matrix, and generate an animal number recognition model based on the comparison result obtained by the comparison; based on the animal number recognition model, the number of animals in the image is recognized. The technical solution of the embodiment of the present disclosure can accurately grasp the total number of animals on the farm in real time, accurately calculate the number of animals, ensure the accuracy of the data, be intuitive and visible, and save a lot of labor costs.

It should be noted that the foregoing is merely a preferred embodiment of the present disclosure, and is not intended to limit the protection scope of the present disclosure.

The following describes an embodiment of the method for identifying the number of animals provided by the present disclosure for identifying the number of individual pigs.

FIG. 3 schematically illustrates a flowchart of an animal number recognition method applied to identification of individual pig numbers according to an embodiment of the present disclosure.

As shown in FIG. 3, the method for identifying the number of animals according to an embodiment of the present disclosure is applied to the process of identifying individual pigs, and includes the following steps:

Step S301, the identification process starts;

Step S302, obtaining a manually marked picture;

In one embodiment of the present disclosure, the manual marking picture refers to the artificial marking of the pig picture by the pig points, and the corresponding picture is provided for the subsequent algorithm. The main function of this step is to provide the picture for the subsequent machine recognition of pig quantity training. Therefore, the accuracy of the pictures required to be marked should be high, and there will be no missing or heavy marks.

Step S303: converting the manually marked picture into a digital matrix;

In an embodiment of the present disclosure, the position coordinates of the marker point in the pre-processed animal image are determined according to the marker points of the processed animal image, and a digital matrix of the same size as the pre-processed animal image is generated, where the number matrix ranges from 0 to Numbers in the range of 1; the corresponding position is determined in the digital matrix based on the position coordinates, and the corresponding point is marked with the number 1; the remaining positions in the number matrix are represented by 0, and a preset Gaussian kernel function is used Gaussian blurring is performed to obtain a digital matrix corresponding to the pre-processed animal images.

In an embodiment of the present disclosure, the marker points in the pre-processed animal image are extracted and extracted to the coordinate positions of the marker points. For subsequent training of the machine recognition model, the pre-processed animal images are converted into a matrix to form a A digital matrix composed of digital structures in the range of 0 to 1 with the same size of the animal image is processed. The coordinate position of the marker point is set to 1 in the digital matrix, and the remaining elements are set to 0.

Step S304, calculating a Gaussian kernel function;

In one embodiment of the present disclosure, the Gaussian blur algorithm uses a normal distribution for image processing. The principle of Gaussian blur can be understood as that each pixel takes the average value of surrounding pixels and calculates it according to the set σ value and the blur radius. The weight matrix is used as a Gaussian kernel function for subsequent convolution operations.

Step S305, performing a convolution operation on the digital matrix according to the Gaussian kernel function;

In an embodiment of the present disclosure, a convolution operation refers to using a convolution kernel to perform a series of operations on each pixel in an image, converting a manually marked image into a matrix, and performing a convolution operation with a Gaussian kernel to generate A new matrix, in which the convolution operation is an operation often used in image processing, which has the effect of enhancing the characteristics of the original signal and reducing noise.

Step S306, obtaining a matrix after convolution;

Step S307, comparing the convolved matrix with the deconvolution matrix of step S312;

In an embodiment of the present disclosure, the most common loss of the degree of deviation between the predicted value and the true value. The deviation between the graph matrix after deconvolution and the label matrix calculated by the convolution operation is the loss. As a result, in the process of constructing the algorithm model, it is necessary to continuously use the loss value to modify the parameters, iterate the algorithm model, and the accuracy rate is constantly improved. When the accuracy rate reaches the standard, you can take the pig picture through the network to calculate a new Matrix, sum the matrix to get the number of pigs.

Step S308, obtaining a training map matrix;

In one embodiment of the present disclosure, the original image of the manually labeled pig quantity picture is called a training image, the training image is changed to a grayscale image, and the image matrix is used as training input data.

Step S309: Perform operation on the training graph matrix through a deep learning convolutional neural network;

In one embodiment of the present disclosure, the training graph matrix can be calculated by a VGG deep learning convolutional neural network. VGG is a deep learning convolutional neural network, showing that the network depth is a key part of the excellent performance of the algorithm. The network contains 16 convolutional layers, and the network structure is very consistent. Specifically, VGG is extracted through convolutional features in several stages. The number of convolutions in each layer starts from the number of the first stage and increases by one in each stage. Times until it reaches the highest value, and then keep it. Although there are many VGG parameters and deep levels, it takes a few iterations to converge.

Step S310: Obtain a feature matrix of a training graph matrix;

In one embodiment of the present disclosure, the training graph matrix is obtained through a VGG network to obtain a feature matrix.

Step S311, performing a deconvolution operation on the feature matrix;

In one embodiment of the present disclosure, the deconvolution operation is very successful in the application of neural network visualization. This algorithm uses a deconvolution operation on the feature matrix to obtain some low-level features and form a graph matrix that is equivalent in size to the label matrix. For displaying predicted images.

Step S312, obtaining a deconvolution matrix;

Step S313, obtaining a loss result, and generating a pig number recognition model;

In an embodiment of the present disclosure, the degree of deviation between the predicted value and the true value is the loss value. The comparison result between the graph matrix after deconvolution and the label matrix calculated by the convolution operation is the loss result. In the process of algorithm model construction, it is necessary to continuously use the loss value to modify the parameters therein, iterate the algorithm model, and the accuracy rate of the algorithm model is continuously improved. When the accuracy rate reaches the standard, you can take the picture of the pig The network calculates a new matrix and sums the matrix to get the number of pigs.

In an embodiment of the present disclosure, each parameter in the above recognition model is in an initial state, and a comparison result obtained after the first training of the recognition model is performed to check the recognition model once to ensure the subsequent recognition model. The iteration result is correct. For example, the first pass calculation of the loss value (comparison result) can be checked once. First, taking CIFAR-10 as an example, if you use the Softmax classifier, the prediction can get an initial loss of about 2.302 (because For the 10 categories, the initial probability should not be 0.1, and the Softmax loss is: -log (probability of the correct category): -ln (0.1) = 2.302); Second, set the regularization coefficient to a normal small value, and add it back to regularization. Term, and then calculate the loss / loss; finally, before training the large data set, you can train a small data set (such as 20 pictures), and then see if your neural network can achieve 0 loss, because If the implementation of the neural network is correct, in the case of no regularization term, it is completely able to overfit this small part of the data; after starting the training of the recognition model, you can monitor some The standard can understand the training status, for example, it can be determined by the change of the loss value after each full round of iteration. The appropriate learning rate can ensure that after each round of full training, the loss value is reduced, and it can be reduced to a relatively high level after a period of time Small degree; then you need to track the accuracy status on the training set and the validation set to determine the state of the classifier. As time progresses, the accuracy on the training set and the validation set will increase. After the accuracy reaches a certain level, the difference between the two is relatively large, then it should be noted that it may be an over-fitting phenomenon. If the difference is not large, it indicates that the recognition model is in good condition. The last thing to pay attention to is the weight. The ratio of the update magnitude to the current weight magnitude is preferably checked independently for each set of parameters.

In step S314, it is determined whether the accuracy rate of the pig quantity identification model meets the standard. If so, step S315 is performed; if not, the process returns to step S308.

In step S315, a pig number recognition model is output and the recognition process is ended.

The following describes the device embodiments of the present disclosure, which can be used to implement the above-mentioned animal number recognition method of the present disclosure.

FIG. 4 schematically illustrates a block diagram of an animal identification device according to an embodiment of the present disclosure.

Referring to FIG. 4, an animal identification device 400 according to an embodiment of the present disclosure includes: an acquisition module 401, a first generation module 402, a second generation module 403, and an identification module 404;

The obtaining module 401 is configured to obtain a pre-processed animal image, where the pre-processed animal image includes an animal and a marked point of the animal;

A first generating module 402, configured to generate a digital matrix having the same size as the pre-processed animal image based on the marked points of the animal;

A second generating module 403, configured to compare a digital matrix with a preset training matrix, and generate an animal number recognition model based on the comparison result obtained by the comparison;

The recognition module 404 is configured to recognize the number of animals in the image based on the number of animals recognition model.

In an embodiment of the present disclosure, the first generating module 402 is specifically configured to:

Determine the position coordinates of the marked point in the pre-processed animal image according to the marked point of the animal;

Generate a digital matrix of the same size as the pre-processed animal image, where the digital matrix is composed of digital structures in the range of 0 to 1;

The corresponding position is determined in the digital matrix based on the position coordinates, and the marked point corresponding to the position coordinates is represented by the number 1, the remaining positions in the digital matrix are represented by 0, and the preset Gaussian kernel function is used to perform Gaussian blurring. Digital matrix corresponding to pre-processed animal images.

In an embodiment of the present disclosure, the second identification module 403 is specifically configured to:

Operate on a digital matrix according to a preset convolution rule to obtain a convolution matrix;

Compare the convolution matrix with a preset training matrix to obtain the comparison result;

Correct the preset recognition model according to the comparison result;

The recognition accuracy rate of the revised recognition model is verified. When the recognition accuracy rate of the recognition model is greater than or equal to a preset threshold, an animal number recognition model is generated.

In an embodiment of the present disclosure, the second identification module 403 is further configured to:

Input the acquired image of the number of animals to be identified into an animal number recognition model, and obtain a digital matrix corresponding to the image of the number of animals to be identified;

Sum the number matrix to obtain the number of animals in the image of the number of animals to be identified.

Since each functional module of the animal identification device of the exemplary embodiment of the present disclosure corresponds to the steps of the above-mentioned example embodiment of the animal number identification method, for details not disclosed in the device embodiments of the present disclosure, please refer to the above-mentioned animal number of the present disclosure. Examples of identification methods.

Reference is now made to FIG. 5, which illustrates a schematic structural diagram of a computer system 500 suitable for implementing an electronic device according to an embodiment of the present disclosure. The computer system 500 of the electronic device shown in FIG. 5 is only an example, and should not impose any limitation on the functions and scope of use of the embodiments of the present disclosure.

As shown in FIG. 5, the computer system 500 includes a central processing unit (CPU) 501 which can be loaded into a random access memory (RAM) 503 from a program stored in a read-only memory (ROM) 502 or from a storage portion 508 Instead, perform various appropriate actions and processes. In the RAM 503, various programs and data required for system operation are also stored. The CPU 501, the ROM 502, and the RAM 503 are connected to each other through a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

The following components are connected to the I / O interface 505: an input portion 1206 including a keyboard, a mouse, etc .; an output portion 507 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc .; and a speaker; a storage portion 508 including a hard disk, etc. ; And a communication section 509 including a network interface card such as a LAN card, a modem, and the like. The communication section 509 performs communication processing via a network such as the Internet. The driver 510 is also connected to the I / O interface 505 as needed. A removable medium 511, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 510 as needed, so that a computer program read therefrom is installed into the storage section 508 as needed.

In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart may be implemented as a computer software program. For example, embodiments of the present disclosure include a computer program product including a computer program carried on a computer-readable medium, the computer program containing program code for performing a method shown in a flowchart. In such an embodiment, the computer program may be downloaded and installed from a network through the communication section 509, and / or installed from a removable medium 511. When this computer program is executed by a central processing unit (CPU) 501, the above-mentioned functions defined in the system of the present application are executed.

It should be noted that the computer-readable medium shown in the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the foregoing. The computer-readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable Programming read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the foregoing. In this disclosure, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in combination with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium may include a data signal that is included in baseband or propagated as part of a carrier wave, and which carries computer-readable program code. Such a propagated data signal may take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. The computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, and the computer-readable medium may send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device . Program code embodied on a computer-readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, which contains one or more of the logic functions used to implement the specified logic. Executable instructions. It should also be noted that in some alternative implementations, the functions noted in the blocks may also occur in a different order than those marked in the drawings. For example, two successively represented boxes may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram or flowchart, and combinations of blocks in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with A combination of dedicated hardware and computer instructions.

The units described in the embodiments of the present disclosure may be implemented by software or hardware. The described units may also be provided in a processor. The names of these units do not, in some cases, define the unit itself.

As another aspect, the present application also provides a computer-readable medium, which may be included in the electronic device described in the foregoing embodiments; or may exist alone without being assembled into the electronic device in. The computer-readable medium carries one or more programs, and when the one or more programs are executed by one electronic device, the electronic device implements the method for identifying an animal quantity as in the foregoing embodiment.

For example, the above electronic device may implement as shown in FIG. 1: step S110, obtaining a pre-processed animal image; step S120, generating a digital matrix having the same size as the pre-processed animal image based on the marked points of the animal; The number matrix is compared with a preset training matrix, and an animal number recognition model is generated based on the comparison result obtained by the comparison; step S140, the number of animals in the image is recognized based on the animal number recognition model.

As another example, the above electronic device can implement each step shown in FIG. 3.

It should be noted that although several modules or units of the device for action execution are mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above may be embodied in one module or unit. Conversely, the features and functions of a module or unit described above can be further divided into multiple modules or units to be embodied.

Through the description of the above embodiments, those skilled in the art can easily understand that the example embodiments described herein can be implemented by software, or by software in combination with necessary hardware. Therefore, the technical solution according to the embodiment of the present disclosure may be embodied in the form of a software product, which may be stored in a non-volatile storage medium (which may be a CD-ROM, a U disk, a mobile hard disk, etc.) or on a network It includes several instructions to enable a computing device (which may be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the embodiment of the present disclosure.

Those skilled in the art will readily contemplate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that conform to the general principles of this disclosure and include the common general knowledge or conventional technical means in the technical field not disclosed by this disclosure. . It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.

It should be understood that the present disclosure is not limited to the precise structure that has been described above and illustrated in the accompanying drawings, and various modifications and changes may be made without departing from the scope thereof. The scope of the disclosure is limited only by the following claims.

Claims

An animal number identification method, comprising:

Obtaining a pre-processed animal image, wherein the pre-processed animal image includes an animal and a marker point of the animal;

Generating a digital matrix having the same size as the pre-processed animal image based on the marked points of the animal;

Comparing the number matrix with a preset training matrix, and generating an animal number recognition model based on the comparison result obtained by the comparison;

The identification of the number of animals in the image is realized based on the animal number recognition model.
The method for identifying the number of animals according to claim 1, wherein the generating a digital matrix of the same size as the pre-processed animal image based on the marked points of the animal, comprising:

Determining a position coordinate of the marked point in the pre-processed animal image according to the marked point of the animal;

Generating a digital matrix of the same size as the pre-processed animal image, wherein the digital matrix is composed of numbers in the range of 0 to 1;

A corresponding position is determined in the digital matrix based on the position coordinates, and a marked point corresponding to the position coordinate is represented by the number 1, the remaining positions in the digital matrix are represented by 0, and a preset Gaussian kernel is used The function performs Gaussian blur processing to obtain a digital matrix corresponding to the pre-processed animal image.
The animal number recognition method according to claim 1, wherein the comparing the number matrix with a preset training matrix, and generating an animal number recognition model based on the comparison result obtained by the comparison, comprises:

Operate the digital matrix according to a preset convolution rule to obtain a convolution matrix;

Comparing the convolution matrix with the preset training matrix to obtain a comparison result;

Correct the preset recognition model according to the comparison result;

The recognition accuracy rate of the modified recognition model is verified. When the recognition accuracy rate of the recognition model is greater than or equal to a preset threshold, an animal number recognition model is generated.
The method for identifying the number of animals according to claim 3, wherein the identifying the number of animals based on the animal number recognition model comprises:

Input the acquired image of the number of animals to be identified into the animal number recognition model, and obtain a digital matrix corresponding to the image of the number of animals to be identified;

Sum the digital matrix to obtain the number of animals in the image of the number of animals to be identified.
An animal quantity identification device, characterized in that the device includes: an acquisition module, a first generation module, a second generation module, and an identification module; wherein,

An acquisition module for acquiring a pre-processed animal image, wherein the pre-processed animal image includes an animal and a marker point of the animal;

A first generating module, configured to generate a digital matrix having the same size as the pre-processed animal image based on the marked points of the animal;

A second generating module, configured to compare the digital matrix with a preset training matrix, and generate an animal number recognition model based on the comparison result obtained by the comparison;

A recognition module is configured to recognize the number of animals in an image based on the animal number recognition model.
The apparatus for identifying an animal quantity according to claim 5, wherein the first generating module is specifically configured to:

Determining a position coordinate of the marked point in the pre-processed animal image according to the marked point of the animal;

Generating a digital matrix of the same size as the pre-processed animal image, wherein the digital matrix is composed of numbers in the range of 0 to 1;

A corresponding position is determined in the digital matrix based on the position coordinates, and a marked point corresponding to the position coordinate is represented by the number 1, the remaining positions in the digital matrix are represented by 0, and a preset Gaussian kernel is used The function performs Gaussian blur processing to obtain a digital matrix corresponding to the pre-processed animal image.
The animal number recognition device according to claim 5, wherein the second recognition module is specifically configured to:

Operate the digital matrix according to a preset convolution rule to obtain a convolution matrix;

Comparing the convolution matrix with the preset training matrix to obtain a comparison result;

Correct the preset recognition model according to the comparison result;

The recognition accuracy rate of the modified recognition model is verified. When the recognition accuracy rate of the recognition model is greater than or equal to a preset threshold, an animal number recognition model is generated.
The animal number recognition device according to claim 7, wherein the second recognition module is further configured to:

Input the acquired image of the number of animals to be identified into the animal number recognition model, and obtain a digital matrix corresponding to the image of the number of animals to be identified;

Sum the digital matrix to obtain the number of animals in the image of the number of animals to be identified.
A computer-readable medium having stored thereon a computer program, characterized in that when the program is executed by a processor, the method for identifying an animal quantity according to any one of claims 1 to 7 is implemented.
An electronic device, comprising:

One or more processors;

A storage device, configured to store one or more programs, and when the one or more programs are executed by the one or more processors, cause the one or more processors to implement any one of claims 1 to 7 The animal number identification method.