CN112734731B - A livestock temperature detection method, device, equipment and storage medium - Google Patents

Abstract

The application discloses a livestock temperature detection method, which comprises the steps of acquiring an infrared photographing instruction, and acquiring infrared images of livestock by using an infrared photographing device; processing the infrared image by using a super-resolution model to obtain a super-resolution image; detecting target livestock in the super-resolution image, and detecting coordinates of temperature detection key points of the target livestock; and detecting the temperature of the temperature detection key point, and alarming when the temperature is higher than a preset threshold value. The method can realize all-weather automatic patrol on the premise of ensuring the precision, provides important guarantee for monitoring the body temperature of livestock, reduces the calculated amount, reduces the hardware pressure, saves the cost and prevents the spread of livestock diseases. The invention also discloses a livestock temperature detection device, equipment and a storage medium.

Description

A livestock temperature detection method, device, equipment and storage medium

technical field

The invention belongs to the technical field of animal husbandry, and in particular relates to a livestock temperature detection method, device, equipment and storage medium.

Background technique

The livestock breeding industry is a labor-intensive industry. In the existing technology, when it is necessary to detect the temperature of livestock, one is the manual method, which is time-consuming and laborious, and depends on the personal quality of the operator, and is affected by various factors. It is also not easy to find all livestock with abnormal temperature. Another method is to rely on automatic technology to detect the temperature of livestock. It uses two target detection methods. One is to obtain the target outline through traditional opencv edge detection, but This method is difficult to adapt to the needs of target detection in complex scenes. The other method is deep learning, but it is mainly aimed at photos of pedestrians and vehicles, and its accuracy is low. It is not suitable for application in the field of livestock temperature detection.

Contents of the invention

In order to solve the above problems, the present invention provides a livestock temperature detection method, device, equipment and storage medium, which can realize all-weather automatic fence patrol under the premise of ensuring accuracy, provide an important guarantee for livestock body temperature monitoring, and can reduce the amount of calculation, Reduce hardware pressure, save costs, and prevent the spread of livestock diseases.

A livestock temperature detection method provided by the invention comprises:

Obtain infrared camera instructions, and use infrared camera devices to collect infrared images of livestock;

Using a super-resolution model to process the infrared image to obtain a super-resolution image;

Detecting the target livestock in the super-resolution image, and detecting the coordinates of the temperature detection key points of the target livestock;

Detect the temperature of the temperature detection key point, and alarm when the temperature is higher than a preset threshold.

Preferably, in the above livestock temperature detection method, before the detection of the target livestock in the super-resolution image, it also includes training the detection model, including:

Using a data labeling tool to label the infrared image to obtain an initial training set;

Use data augmentation methods to obtain diverse data sets;

Use the deep learning target detection model to train to get the initial model;

Perform reasoning on the initial model, use the temperature matrix to perform data cleaning on the diverse data set, determine whether there is an object within the temperature range to be detected in the single infrared image, and use the deep learning target detection method if it exists The model performs inference and obtains the prediction result BBox in the infrared image;

By setting the threshold of BBox to filter out some targets, the final livestock target detection result is obtained.

Preferably, in the above livestock temperature detection method, the infrared image collected by the infrared camera device is:

Infrared images of livestock are collected using a thermal imaging camera mounted on a vehicle on the track above the aisle at the farm.

Preferably, in the above livestock temperature detection method, the data enhancement methods include mixup, flipping, translation, random cropping, adding random noise, and GAN methods.

Preferably, in the above livestock temperature detection method, the prediction result BBox in the infrared image is obtained through the yolov4 model.

A livestock temperature detection device provided by the invention comprises:

An infrared photographing component is used to obtain infrared photographing instructions and collect infrared images of livestock;

An image processing component, configured to process the infrared image using a super-resolution model to obtain a super-resolution image;

The temperature detection key point coordinate determining component is used to detect the target livestock in the super-resolution image, and detect the coordinates of the target livestock's temperature detection key point;

The alarm component is used to detect the temperature of the temperature detection key point, and alarm when the temperature is higher than the preset threshold.

Preferably, in the above-mentioned livestock temperature detection device, a detection model training component is also included, which is used to use a data labeling tool to label the infrared image to obtain an initial training set; use data enhancement to obtain a diverse data set; use deep learning The target detection model is trained to obtain an initial model; the initial model is reasoned, and the temperature matrix is used to perform data cleaning on the diverse data set, and it is judged whether there is an object within the temperature range to be detected in the single infrared image, if If it exists, use the deep learning target detection model to perform inference to obtain the prediction result BBox in the infrared image; filter out some targets by setting the threshold of the BBox to obtain the final livestock target detection result.

Preferably, in the above livestock temperature detection device, the infrared photographing component is specifically used to collect infrared images of livestock by means of an infrared thermal imager installed on the vehicle on the track above the aisle of the farm.

A kind of computer equipment provided by the invention comprises:

memory for storing computer programs;

A processor, configured to implement the steps of any one of the above livestock temperature detection methods when executing the computer program.

The present invention provides a computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of any one of the methods for detecting the temperature of livestock above are implemented.

It can be seen from the above description that the above-mentioned livestock temperature detection method provided by the present invention includes first obtaining an infrared photographing instruction, using an infrared photographing device to collect an infrared image of a livestock; then using a super-resolution model to process the infrared image to obtain a super-resolution image; Then detect the target livestock in the super-resolution image, and detect the coordinates of the temperature detection key points of the target livestock; finally detect the temperature of the temperature detection key points, when the temperature is higher than the preset threshold It can be seen that this method does not require manual participation, and can realize all-weather automatic fence patrol under the premise of ensuring accuracy, providing an important guarantee for livestock body temperature monitoring, and can reduce the amount of calculation, reduce hardware pressure, save costs, and prevent the spread of livestock diseases. The livestock temperature detection device, equipment and storage medium provided by the present invention have the same advantages as the above method.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only It is an embodiment of the present invention, and those skilled in the art can also obtain other drawings according to the provided drawings without creative work.

Fig. 1 is the schematic diagram of the embodiment of a kind of livestock temperature detection method provided by the present invention;

Fig. 2 is the schematic diagram of the embodiment of a kind of livestock temperature detection device provided by the present invention;

Fig. 3 is a schematic diagram of an embodiment of a computer device provided by the present invention.

Detailed ways

The core of the present invention is to provide a livestock temperature detection method, device, equipment and storage medium, which can realize all-weather automatic fence patrol under the premise of ensuring accuracy, provide an important guarantee for livestock body temperature monitoring, and can reduce the amount of calculation and hardware pressure , save costs and prevent the spread of livestock diseases.

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

An embodiment of a livestock temperature detection method provided by the present invention is shown in Figure 1, and Figure 1 is a schematic diagram of an embodiment of a livestock temperature detection method provided by the present invention, the method includes the following steps:

S1: Obtain an infrared camera instruction, and use an infrared camera device to collect infrared images of livestock;

Specifically, the infrared camera device can be, but not limited to, an infrared thermal imager. An infrared thermal imager converts invisible infrared energy emitted by an object into a visible thermal image, so any object with a temperature higher than 0° can Imaging, and the temperature of livestock is significantly higher than the ambient temperature, so it is not affected by light, and the status of livestock can be observed around the clock. Combining target detection technology to capture livestock targets can facilitate body temperature monitoring. Moreover, the infrared thermal imaging camera installed on the vehicle on the track above the aisle of the farm can be used to collect infrared images of livestock, instead of using the existing hand-held temperature measuring gun and human eyes to observe the temperature through the thermal imager screen. In this method, the RFID card can be pasted at the center of each column for positioning, and the patrol car can be used to stop at fixed points, combined with the infrared thermal imager on the car, to collect the infrared heat map data of livestock, and to collect Preliminary cleaning of the received data, selecting the data with normal angles and livestock in the pen as the initial sample set, and judging whether the maximum temperature in the temperature matrix exceeds the set threshold, for example, the current threshold is 45 degrees Celsius, and the data exceeding 45 degrees Celsius is It is judged as abnormal data and does not participate in the reasoning of subsequent models.

S2: Using the super-resolution model to process the infrared image to obtain a super-resolution image;

It should be noted that the single-picture super-resolution (SISR) algorithm used in this step is integrated into a recurrent neural network through the cross-scale non-local (CS-NL) attention module, by combining the local CS-NL prior and The non-local prior is connected to a powerful recurrent aggregation unit to obtain the final super-resolution image, and the use of super-resolution enhancement technology to capture imaging details can greatly improve the accuracy of model reasoning while reducing the cost of equipment. Specifically, the infrared livestock picture needs to extract image features through a convolutional neural network, and then enter a series of recurrent neural units, which form a recurrent network, and finally fuse the information of the recurrent units through the convolutional network to generate a high-resolution picture. For each recursive unit, the features of the previous layer go through three branches, which are the cross-scale non-local attention branch, the full-scale non-local attention branch and the local branch, and the last three branches extract all features through interactive mapping fusion. As the input of the next layer, the specific operation can be as follows: extract the features of the infrared livestock picture through the convolutional neural network; use the features of the previous step as input, and perform feature mining through the SEM unit (Self-Examplers Minig), including the features through the cross-scale The non-local attention network extracts features, extracts features through the inline scale non-local attention network, extracts features through the local information extraction network, performs interactive mapping and fusion of the features obtained in these three steps, and passes the fused information through Stride The convolutional layer and the Deconv layer network are processed; several SEM units are repeated to form a recurrent neural network; the information mined by the above several SEM units is combined through the network Concat, and the final high-resolution infrared infrared image is obtained through the convolutional neural network. Livestock pictures. Take the picture of livestock collected by an infrared camera with a resolution of 256*192 in a farm as an example. This picture is too blurry, especially the key parts of temperature measurement such as the edge of the livestock and the head, etc. are difficult to distinguish clearly Through this super-resolution model processing, a high-definition infrared livestock picture with a resolution of 512*384 is obtained. It can be seen that this step can provide great help for subsequent accurate target detection and instance segmentation based on infrared livestock pictures. This facilitates the detection of sick and dead livestock.

S3: Detect the target livestock in the super-resolution image, and detect the coordinates of the temperature detection key points of the target livestock;

In the above steps, the data collected by the infrared thermal imager needs to be processed, and the grayscale image data directly corresponding to the temperature matrix is encoded into an intuitive infrared heat map, and then enhanced to obtain a large number of infrared images. Using these data as a training data set, train a neural network model for livestock instance segmentation. Finally, through the trained model, the livestock in the infrared image can be segmented, and all livestock can be monitored in combination with the temperature. Specifically, select the grayscale image data corresponding to the temperature matrix in the collected infrared data, and perform gamma transformation and local enhancement of the histogram according to its contrast. The texture is stronger, and the local enhancement of the histogram can darken the high-light areas with low contrast, making the high-light areas of the image softer and the details clearer. Then use hot metal coding to encode the processed grayscale image into a visually compliant image. Infrared heat map; build a training data set; collect the collected infrared data, process the data in the previous step, and then select enough images with different distributions of livestock, mark them, and mark each head in the infrared image The outline of livestock, connect the outlines, all the pixels corresponding to the same livestock area in the image are marked with the same ID, different IDs correspond to different livestock; train the neural network model of livestock instance segmentation, build a neural network model suitable for instance segmentation, Set an appropriate loss function, input the training data set, and iteratively update the weights in the network model through backpropagation, and finally obtain the neural network model for livestock instance segmentation. The model after training can be used to detect the coordinates of the key points of temperature detection of the target livestock, and the coordinates can be mapped to the infrared image to obtain the body temperature data of the livestock. The detection of key points in the groin reduces the demand for hardware resources, reduces the complexity of the processing process, and improves efficiency.

S4: Detect the temperature of the key point of temperature detection, and alarm when the temperature is higher than the preset threshold.

It should be noted that the preset threshold can be but not limited to 40 degrees Celsius. When 40 degrees Celsius is used as the basis for judging fever, when the temperature value at the key point of the detection exceeds 40 degrees Celsius, it is judged as a livestock with a fever, or in summer In hot conditions, it can also be judged whether the temperature at the key detection point of one of the livestock is higher than that of other livestock by a value, for example, 1.8 degrees Celsius. The influence of factors leads to the inaccurate absolute temperature of the key point and misjudgment, and the livestock whose temperature value is lower than the set threshold is judged as dead livestock. When this happens, an alarm is required, and the predicted fever livestock It is pushed to each unit, and the breeder makes timely diagnosis and treatment based on the pushed results.

It can be seen that the above embodiment encodes and enhances the image data directly related to temperature to obtain a more intuitive infrared image, and then performs instance segmentation on livestock to obtain the temperature area of each livestock, thereby monitoring the body temperature of multiple livestock. The livestock in the above scheme can be but not limited to pigs, cattle and sheep, and any similar livestock can be treated by the above method.

It can be seen from the above description that in the embodiment of the above-mentioned livestock temperature detection method provided by the present invention, because it includes first obtaining an infrared photographing instruction, using an infrared photographing device to collect an infrared image of a livestock; then using a super-resolution model to process the infrared image to obtain a super-resolution image; then detect the target livestock in the super-resolution image, and detect the coordinates of the key point of temperature detection of the target livestock; finally detect the temperature of the key point of temperature detection, and alarm when the temperature is higher than the preset threshold, it can be seen that this method does not need Manual participation can realize all-weather automatic fence patrol on the premise of ensuring accuracy, providing an important guarantee for livestock temperature monitoring, and can reduce the amount of calculation, reduce hardware pressure, save costs, and prevent the spread of livestock diseases.

It should also be explained that, for the infrared images of livestock collected by the infrared thermal imager, the convolutional neural network deep learning method is used for model training to obtain a deep learning model that can directly extract key points such as the groin and ear roots of livestock from infrared images of livestock. Therefore, in a specific embodiment of the above livestock temperature detection method, before detecting the target livestock in the super-resolution image, it also includes training the detection model, which may also include:

Use the data labeling tool to label the infrared image to obtain the initial training set. Specifically, this is carried out by manual labeling. The key point detection model training based on deep learning requires a certain amount of diverse samples. Key points such as the frame of the livestock and the groin and ear roots are marked on it, and used as "knowledge" for deep neural network model learning;

Use data enhancement methods to obtain diverse data sets, including but not limited to using mixup methods to expand the data set;

Use the deep learning target detection model to train to get the initial model;

Inference the initial model, use the temperature matrix to clean the diverse data sets, and judge whether there is an object in the temperature range to be detected in a single infrared image. If it exists, use the deep learning target detection model to reason, and get the The predicted result BBox;

By setting the threshold of the BBox to filter out some targets, the final livestock target detection result is obtained, that is, the coordinate values of the key points of the ears and groin of the livestock are obtained by reasoning.

Specifically, it is necessary to establish a data set for deep learning model training in advance, provide a variety of samples for the deep neural network, and conduct deep neural network model training. All the livestock in the picture are extracted separately and set to a fixed size; the marked coordinates of the ears and groin key points of the livestock are converted into a real heatmap of 64x64 size through a Gaussian kernel; feature extraction is performed through a simple baseline neural network to obtain a predicted heatmap; Use the real heatmap and predicted heatmap to calculate the value of the L2 loss loss function; continuously update the model weight through error backpropagation until the model converges or meets the iteration termination condition; model application effect evaluation and optimization, through practical application, it is found that livestock is standing Sometimes the back edge is misjudged as the groin, and the loss function can also be improved by adding the loss value generated by the detection of invisible key points to the original loss function, and retraining the model to improve the model detection effect. Enhance detection accuracy. After the training of the key point detection model of livestock based on infrared heat map based on deep learning is completed, a new infrared heat map of livestock is provided as the input of the model. After the target detection and key point detection model are processed, the ear root, Coordinate information of key points such as the groin.

In another specific embodiment of the above livestock temperature detection method, the data enhancement methods may include mixup, flipping, translation, random cropping, adding random noise, and GAN methods, that is to say, at least one of them may be used to expand data Set to get the sample set for model training.

Further, the prediction result BBox in the infrared image can be obtained through the yolov4 model. Specifically, during model training, resize the pictures in the training set to the same size and obtain their real BBox position information, obtain the predicted BBox information through the yolov4 model, and compare the real BBox position information with the predicted BBox information In contrast, the sum of classification loss, confidence loss, location loss, and iou loss is used as the final loss, and the weight is continuously updated using the backpropagation algorithm until the model converges or the iteration termination condition is satisfied, and then model reasoning and post-processing are performed. Specifically, In the online inference stage, first combine the TTA method for multiple inferences to take the result with the highest confidence as the output result of the target detection, and then set the minimum threshold of the BBox. When the threshold of the BBox is less than the given threshold, the inference results are filtered out. When the threshold of the BBox is When it is greater than the given threshold, output the infrared heat map livestock target detection result.

An embodiment of a livestock temperature detection device provided by the present invention is shown in Figure 2, and Figure 2 is a schematic diagram of an embodiment of a livestock temperature detection device provided by the present invention, the device includes:

The infrared photographing part 201 is used to obtain infrared photographing instructions and collect infrared images of livestock. Specifically, the infrared photographing part can be, but not limited to, an infrared thermal imager. An infrared thermal imager is an invisible infrared energy emitted by an object. It is transformed into a visible thermal image, so any object with a temperature higher than 0° can be imaged, and the temperature of the livestock is significantly higher than the ambient temperature, so it is not affected by light, and the status of the livestock can be observed around the clock. Combining with the target detection technology to capture the livestock target, it can be convenient Perform temperature monitoring. Moreover, the infrared thermal imaging camera installed on the vehicle on the track above the aisle of the farm can be used to collect infrared images of livestock, instead of using the existing hand-held temperature measuring gun and human eyes to observe the temperature through the thermal imager screen. In this method, the RFID card can be pasted at the center of each column for positioning, and the patrol car can be used to stop at fixed points, combined with the infrared thermal imager on the car, to collect the infrared heat map data of livestock, and to collect Preliminary cleaning of the received data, selecting the data with normal angles and livestock in the pen as the initial sample set, and judging whether the maximum temperature in the temperature matrix exceeds the set threshold, for example, the current threshold is 45 degrees Celsius, and the data exceeding 45 degrees Celsius is It is judged as abnormal data and does not participate in the reasoning of subsequent models;

The image processing component 202 is used to process the infrared image using the super-resolution model to obtain a super-resolution image, and the single-picture super-resolution (SISR) algorithm used is integrated into a loop through the cross-scale non-local (CS-NL) attention module In the neural network, by combining the local CS-NL prior and the non-local prior into a powerful recurrent aggregation unit, the final super-resolution image is obtained, and the super-resolution enhancement technology is used to capture the imaging details, which can greatly improve the performance of model reasoning. accuracy while reducing equipment cost. Specifically, the infrared livestock picture needs to extract image features through a convolutional neural network, and then enter a series of recurrent neural units, which form a recurrent network, and finally fuse the information of the recurrent units through the convolutional network to generate a high-resolution picture. For each recursive unit, the features of the previous layer go through three branches, which are the cross-scale non-local attention branch, the full-scale non-local attention branch and the local branch, and the last three branches extract all features through interactive mapping fusion. As the input of the next layer, the specific operation can be as follows: extract the features of the infrared livestock picture through the convolutional neural network; use the features of the previous step as input, and perform feature mining through the SEM unit (Self-Examplers Minig), including the features through the cross-scale The non-local attention network extracts features, extracts features through the inline scale non-local attention network, extracts features through the local information extraction network, performs interactive mapping and fusion of the features obtained in these three steps, and passes the fused information through Stride The convolutional layer and the Deconv layer network are processed; several SEM units are repeated to form a recurrent neural network; the information mined by the above several SEM units is combined through the network Concat, and the final high-resolution infrared infrared image is obtained through the convolutional neural network. Livestock pictures. Take the picture of livestock collected by an infrared camera with a resolution of 256*192 in a farm as an example. This picture is too blurry, especially the key parts of temperature measurement such as the edge of the livestock and the head, etc. are difficult to distinguish clearly , through this super-resolution model processing, a high-definition infrared livestock picture with a resolution of 512*384 is obtained. It can be seen that it can provide a great help for subsequent accurate target detection based on infrared livestock pictures, and instance segmentation, which is Facilitate the detection of sick and dead livestock;

The temperature detection key point coordinate determining component 203 is used to detect the target livestock in the super-resolution image, and detect the coordinates of the temperature detection key point of the target livestock. The data collected by the infrared thermal imager needs to be processed. The grayscale image data directly corresponding to the temperature matrix is encoded into an intuitive infrared heat map, and then enhanced to obtain a large number of infrared images. These data are used as a training data set to train a neural network model for livestock instance segmentation. Finally, through the trained model, the livestock in the infrared image can be segmented, and all livestock can be monitored in combination with the temperature. Specifically, select the grayscale image data corresponding to the temperature matrix in the collected infrared data, and perform gamma transformation and local enhancement of the histogram according to its contrast. The texture is stronger, and the local enhancement of the histogram can darken the high-light areas with low contrast, making the high-light areas of the image softer and the details clearer. Then use hot metal coding to encode the processed grayscale image into a visually compliant image. Infrared heat map; build a training data set; collect the collected infrared data, process the data in the previous step, and then select enough images with different distributions of livestock, mark them, and mark each head in the infrared image The outline of livestock, connect the outlines, all the pixels corresponding to the same livestock area in the image are marked with the same ID, different IDs correspond to different livestock; train the neural network model of livestock instance segmentation, build a neural network model suitable for instance segmentation, Set an appropriate loss function, input the training data set, and iteratively update the weights in the network model through backpropagation, and finally obtain the neural network model for livestock instance segmentation. The model after training can be used to detect the coordinates of the key points of temperature detection of the target livestock, and the coordinates can be mapped to the infrared image to obtain the body temperature data of the livestock. The detection of key points in the groin reduces the demand for hardware resources, reduces the complexity of the processing process, and improves efficiency;

The alarm component 204 is used to detect the temperature of the key point of temperature detection. When the temperature is higher than the preset threshold, it will alarm. It should be noted that the preset threshold can be but not limited to 40 degrees Celsius. When 40 degrees Celsius is used as the basis for judging fever When the temperature value at the detection key point exceeds 40 degrees Celsius, it is determined to be a livestock with a fever, or in the case of hot summer, it can also be judged whether the temperature at the detection key point of one of the livestock is higher than that of other livestock. A higher value, such as 1.8 degrees Celsius, when it is determined that the livestock meeting the above requirements is a febrile livestock, it alleviates the misjudgment of the absolute temperature of the key point due to the influence of environmental temperature factors, and the temperature value is lower than the set threshold. The livestock is judged as dead livestock. When this happens, an alarm is required, and the predicted fever livestock can be pushed to each unit. The breeder can diagnose and deal with it in a timely manner based on the pushed results.

In a specific embodiment of the above-mentioned livestock temperature detection device, it may also include a detection model training component, which is used to mark the infrared image with a data labeling tool to obtain an initial training set; use data enhancement to obtain a diverse data set; use depth Learn the target detection model and train to obtain the initial model; reason the initial model, use the temperature matrix to clean the diverse data sets, and judge whether there is an object in the temperature range to be detected in a single infrared image, and if so, use deep learning The target detection model performs inference to obtain the prediction result BBox in the infrared image; by setting the threshold of the BBox to filter out some targets, the final livestock target detection result is obtained.

Specifically, it is necessary to establish a data set for deep learning model training in advance, provide a variety of samples for the deep neural network, and conduct deep neural network model training. All the livestock in the picture are extracted separately and set to a fixed size; the marked coordinates of the ears and groin key points of the livestock are converted into a real heatmap of 64x64 size through a Gaussian kernel; feature extraction is performed through a simple baseline neural network to obtain a predicted heatmap; Use the real heatmap and predicted heatmap to calculate the value of the L2 loss loss function; continuously update the model weight through error backpropagation until the model converges or meets the iteration termination condition; model application effect evaluation and optimization, through practical application, it is found that livestock is standing Sometimes the back edge is misjudged as the groin, and the loss function can also be improved by adding the loss value generated by the detection of invisible key points to the original loss function, and retraining the model to improve the model detection effect. Enhance detection accuracy. After the training of the key point detection model of livestock based on infrared heat map based on deep learning is completed, a new infrared heat map of livestock is provided as the input of the model. After the target detection and key point detection model are processed, the ear root, Coordinate information of key points such as the groin.

An embodiment of a computer device provided by the present invention is shown in FIG. 3, and FIG. 3 is a schematic diagram of an embodiment of a computer device provided by the present invention. The device includes:

Memory 301, used to store computer programs;

The processor 302 is configured to implement the steps of any one of the above livestock temperature detection methods when executing the computer program.

In an embodiment of a computer-readable storage medium provided by the present invention, a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of any one of the methods for detecting the temperature of livestock above are implemented.

The above-mentioned livestock temperature detection device, equipment and storage medium provided by the present invention can realize all-weather automatic fence patrol on the premise of ensuring accuracy, provide an important guarantee for livestock body temperature monitoring, and can reduce the amount of calculation, reduce hardware pressure, save costs, and prevent Spread of livestock disease.

The above description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention will not be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (8)

1. A method for detecting temperature of livestock, comprising:

acquiring an infrared photographing instruction, and acquiring an infrared image of livestock by using an infrared photographing device;

processing the infrared image by using a super-resolution model to obtain a super-resolution image;

detecting target livestock in the super-resolution image, and detecting coordinates of temperature detection key points of the target livestock;

detecting the temperature of the temperature detection key point, and alarming when the temperature is higher than a preset threshold value;

the infrared image is processed by utilizing a super-resolution model, the super-resolution image is obtained, the characteristics of the infrared image of livestock are extracted through a convolutional neural network, the characteristics are taken as input, the characteristics are mined through SEM units, the characteristics are extracted through a cross-scale non-local attention network, the characteristics are extracted through an inline scale non-local attention network, the characteristics are extracted through a local information extraction network, the characteristics obtained in the three steps are subjected to interactive mapping fusion, the fused information is processed through a Stride convolutional layer and a Dev layer network, and a plurality of SEM units are repeated to form a cyclic neural network; combining the information mined by the plurality of SEM units through a network Concat, and obtaining a final super-resolution image through a convolutional neural network;

before the target livestock in the super-resolution image is detected, training a detection model is further included, and the method comprises the following steps:

marking the infrared image by using a data marking tool to obtain an initial training set;

obtaining a diversified data set by utilizing a data enhancement mode;

training by using a deep learning target detection model to obtain an initial model;

reasoning the initial model, cleaning the data of the diversified data sets by utilizing a temperature matrix, judging Shan Zhangsuo whether an object in a temperature interval to be detected exists in the infrared image, and if so, reasoning by using the deep learning target detection model to obtain a prediction result BBox in the infrared image;

filtering out a part of targets by setting a threshold value of BBox to obtain a final livestock target detection result;

and filtering out a part of targets by setting a threshold value of BBox, wherein the final livestock target detection result is obtained by the following steps: and obtaining the coordinate information of the auricles and groin of each livestock in the picture.

2. The method for detecting the temperature of livestock according to claim 1, wherein the step of acquiring the infrared image of the livestock by using the infrared photographing device is:

infrared images of livestock are acquired by using a thermal infrared imager mounted on a car of a rail above a farm aisle.

3. The method of claim 1, wherein the data enhancement mode includes a mix up, flip, pan, random cut, add random noise, GAN mode.

4. The livestock temperature detection method of claim 1, wherein the prediction result BBox in the infrared image is obtained by a yolov4 model.

5. A livestock temperature detection device, comprising:

the infrared photographing component is used for acquiring an infrared photographing instruction and acquiring an infrared image of livestock;

the image processing component is used for processing the infrared image by utilizing the super-resolution model to obtain a super-resolution image;

a temperature detection key point coordinate determining unit for detecting a target livestock in the super-resolution image and detecting coordinates of a temperature detection key point of the target livestock;

the alarm component is used for detecting the temperature of the temperature detection key point, and alarming when the temperature is higher than a preset threshold value;

the image processing component is specifically used for extracting the characteristics of an infrared image of livestock through a convolutional neural network, taking the characteristics as input, carrying out characteristic mining through SEM units, and comprises the steps of extracting the characteristics through a cross-scale non-local attention network, extracting the characteristics through an inline scale non-local attention network, extracting the characteristics through a local information extraction network, carrying out interactive mapping fusion on the characteristics obtained in the three steps, processing the fused information through a Stride convolutional layer and a Deconv layer network, and repeating a plurality of SEM units to form the cyclic neural network; combining the information mined by the plurality of SEM units through a network Concat, and obtaining a final super-resolution image through a convolutional neural network;

the detection model training component is used for marking the infrared image by using a data marking tool to obtain an initial training set; obtaining a diversified data set by utilizing a data enhancement mode; training by using a deep learning target detection model to obtain an initial model; reasoning the initial model, cleaning the data of the diversified data sets by utilizing a temperature matrix, judging Shan Zhangsuo whether an object in a temperature interval to be detected exists in the infrared image, and if so, reasoning by using the deep learning target detection model to obtain a prediction result BBox in the infrared image; filtering out a part of targets by setting a threshold value of BBox to obtain a final livestock target detection result;

the detection model training component is used for obtaining the coordinate information of the auricles and the groins of the livestock in the pictures.

6. The livestock temperature detection device of claim 5, wherein the infrared photographing means is specifically configured to collect infrared images of the livestock using a thermal infrared imager mounted on a car of a rail above the farm aisle.

7. A computer device, comprising:

a memory for storing a computer program;

a processor for carrying out the steps of the method for detecting the temperature of livestock as claimed in any of claims 1 to 4 when executing said computer program.

8. A computer readable storage medium, characterized in that the computer readable storage medium has stored thereon a computer program which, when executed by a processor, implements the steps of the livestock temperature detection method according to any of claims 1 to 4.