KR20200059445A - Method and apparatus for detecting behavior pattern of livestock using acceleration sensor - Google Patents

Abstract

Disclosed are a method and a device for detecting livestock behavior patterns by using a sensor. The method for detecting livestock behavior patterns comprises collecting and analyzing data of a sensor attached to livestock to predict the conditions of the livestock. In particular, the prediction method is effective for specific conditions such as estrus, disease and pregnancy of the livestock, thereby reducing labor, improving the productivity of dairy households and increasing income.

Description

Method and apparatus for detecting behavior pattern of livestock using acceleration sensor

The present invention relates to a method and apparatus for detecting a behavioral pattern of a livestock using a necklace type sensor including an acceleration sensor, and more specifically, by analyzing data acquired from the acceleration sensor using a machine learning method, the behavioral pattern of the livestock The result is to be able to accurately determine the condition of the cow.

Today, the dairy industry is gradually moving away from the traditional breeding method, and is gradually becoming full-time, corporate, and large-scale, but on the other hand, the population aging of dairy farmers and the lack of successors are prominent, and the social demand for the environment increases, thereby increasing the environment for dairy farmers. Regulations are increasingly tightened, and the productivity of the dairy industry is declining.

In particular, according to the 2013 statistics of the Ministry of Food, Agriculture, Forestry and Rural Affairs, the number of dairy households has decreased by half in the past 10 years due to the aging of dairy farmers, etc., while the number of heads raised per house has increased by 1.5 to 1.6 times. Problems that are difficult to properly manage are increasing.

As mentioned above, in order to increase productivity in the trend of decreasing and aging of the agricultural population, such as the livestock industry, there are increasing attempts to converge various scientific technologies such as ICT and BT to replace the labor force in the industry.

On the other hand, in the field of accurately grasping the actual health status and developmental status of domesticated livestock using ICT technology, the spread of technology development is still slow, and more research is needed in the future.

Registration No. 10-1657682, feed efficiency analysis system using acceleration sensor and GPS Registration No. 10-1318716, Animal movement pattern analysis system

An object of the present invention for solving the above problems is to provide a livestock behavioral pattern detection method and apparatus capable of accurately analyzing a livestock behavioral pattern by analyzing data acquired through a sensor worn on the livestock.

Specifically, in the present invention, the purpose of the present invention is to collect the value of the acceleration sensor worn on the livestock, to analyze the pattern of the behavior of the corresponding livestock by grafting machine learning techniques, and to clearly discern the specific state of the livestock. Is done.

A method of detecting livestock behavior patterns to achieve the above object is a data collection step (S100) of collecting data from the sensor, a feature point generation step (S200) of analyzing the collected data to generate feature points, and a classifier modeling the feature points It characterized in that it comprises a pattern analysis step (S300) for detecting the behavior pattern by passing through.

In addition, the sensor of the animal behavior pattern detection method of the present invention is an acceleration sensor, and the data is characterized in that the displacement amount of the X-axis, Y-axis, and Z-axis.

In addition, the feature point generation step (S200) of the livestock behavior pattern detection method of the present invention is calculated based on the displacement amount of the X-axis, Y-axis, and Z-axis, and a vector signal forming step (S210) to form a vector signal and the vector Characterized in that it further comprises a frequency conversion step of converting the signal into a frequency domain signal (S220) and the step of sampling the signal of the converted frequency domain (S230).

In addition, the modeling used in the pattern analysis step (S300) of the livestock behavior pattern detection method of the present invention is characterized by using a machine learning algorithm.

In addition, the frequency conversion used in the frequency conversion step (S220) of the method for detecting livestock behavior pattern of the present invention uses a fast Fourier transform, and the sampling used in the sampling step (S230) is It is characterized in that the peak value (Max Value) in a specific unit region of each signal is extracted and matrixed.

In the case of using the method and apparatus for detecting a behavior pattern of livestock according to the present invention as described above, a livestock manager remotely monitors the exact state of the livestock and inputs workers by creating an appropriate breeding environment according to the exact state of the livestock. It can reduce working hours and additional administrative costs of workers.

In addition, by using the method and apparatus for detecting livestock behavioral patterns of the present invention, it is possible to effectively predict or diagnose important state changes related to the health of livestock such as estrus, pregnancy, and disease of livestock, thereby maintaining well-being of livestock, and ultimately It can increase the productivity and profits of dairy furniture.

In addition, by using the method and apparatus for detecting livestock behavior patterns of the present invention, if it is possible to effectively predict the estrus of a specific livestock, it is possible to increase the fertility rate of the livestock and consequently increase the income and productivity of the dairy farm. have.

1 is a logical configuration diagram of a livestock behavior pattern detection device attached to a livestock.
2 logically illustrates a method for detecting livestock behavior patterns.
3 shows a method of extracting a feature point of a sensor signal through frequency analysis.
Figure 4 shows the detection method of the present invention by way of example.
5 is a view showing an implementation of the housing part and the sensing member of the livestock behavior pattern detection device of the present invention.
6 shows an example in which the livestock behavior pattern detecting apparatus of the present invention is mounted on a livestock.

The terms used in the present application are only used to describe specific embodiments, and are not intended to limit the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this application, terms such as “include” or “have” are intended to indicate that a feature, number, step, operation, component, part, or combination thereof described in the specification exists, one or more other features. It should be understood that the existence or addition possibilities of fields or numbers, steps, operations, components, parts or combinations thereof are not excluded in advance.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by a person skilled in the art to which the present invention pertains. Terms such as those defined in a commonly used dictionary should be interpreted as having meanings consistent with meanings in the context of related technologies, and should not be interpreted as ideal or excessively formal meanings unless explicitly defined in the present application. Does not.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

Figure 1 shows the configuration of a livestock behavior pattern detection device attached to the livestock. The livestock behavior pattern detection apparatus is basically a sensor 101 for detecting changes in livestock, a transmission unit 102 for transmitting data obtained from the sensor to a remote server, a battery for driving the sensor and the transmission unit, etc. In order to manage the unit 103, the sensor 101, the transmission unit 102, the battery unit 103, and the like, a sensing member 100 composed of a CPU 104 or a PCB 105, and the above-described components are stored. It may be composed of a housing portion 200 and a fastening member 300 to facilitate mounting of the housing portion to a livestock.

In addition, when the sensing member 100 and the housing part 200 are fastened to the livestock by using a fastening member 300, an additional weight adjusting member 400, etc. is used to accurately position the sensing member 100 and the like. This can be used further.

Specifically, the sensor 101 may use an acceleration sensor, and more specifically, a six-axis acceleration sensor may be used.

In addition, specifically, the transmission unit 102 mainly uses a wireless communication method, and the detailed wireless communication method used may be selectively used as needed, such as Wi-Fi, Bluetooth, BLE, ZigBee, NBIOT, LORA, LTE-M, etc. Can be.

In addition, although the weight control member 400 may be used in various forms, the weight control member 400 may be implemented in the form of a weight, so that the housing part 200 can provide a function that can be fixed to a certain position of the livestock.

Through this function of the weight adjusting member 400, the sensing member 100 can be fastened to the same position or direction of the livestock as much as possible, and as a result, the sensing data of the livestock obtained from the sensing member 100 is statistically significant. And more accurately.

In addition, the remote server as described above collects data from a plurality of sensors mounted on a plurality of livestock, performs various machine learning techniques or analytical methods as described below, and serves to analyze the behavioral pattern of the livestock. Can be.

5 illustrates that the sensing member 100 presented in FIG. 1 is implemented in the form of a PCB, and also shows an actual implementation of the housing 200 accommodating the sensing member 100. will be.

In addition, Figure 6 shows the shape of the sensing member 100, and the housing 200 seen in cattle among cattle are fastened to cattle through the fastening member 300, and the weight adjustment member 400 is added.

2 is a logical diagram of the method for detecting livestock behavior patterns of the present invention.

The method for detecting a livestock behavior pattern described above includes a data collection step (S100) of collecting data from the sensor; A feature point generating step of analyzing the collected data and generating a feature point (S200); And a pattern analysis step (S300) of detecting behavior patterns by passing the feature points through a modeled classifier.

In addition, the sensor is an acceleration sensor, the data is characterized in that the displacement amount of the X-axis, Y-axis, Z-axis.

The above-described feature point generating step (S200) includes a vector signal forming step (S210) of forming a vector signal as shown in Equation 1 below based on the displacement amount of the X-axis, Y-axis, and Z-axis; A frequency conversion step of converting the vector signal into a frequency domain signal (S220); And sampling the converted frequency domain signal (S230).

As a method of calculating a feature point that can represent the movement of a livestock from displacements of the X-axis, Y-axis, and Z-axis output from the acceleration sensor, other methods that can calculate the average value of the vector can be used, but the above formula (1 Since the vector signal such as) is the same concept as the sum of the average sizes of each axis, it has a characteristic that can indicate how much the displacement amount has changed from the reference point, and this characteristic is especially the behavior of the animal in the behavior pattern of the animal. In terms of the importance of how large or large it is, it can be said that it has great utility.

In addition, the above-described vector signal may be regarded as representing the magnitude of displacements with respect to a plurality of axes measured from the acceleration sensor, and the characteristics of the vector signal may be used as described below for the error state of the acceleration sensor.

That is, when the data measured from the acceleration sensor of the livestock for a sufficiently long period is analyzed by vector signaling, it is possible to determine that a value above a certain threshold is a value that cannot be obtained from the movement of the livestock. That is, after obtaining and analyzing sufficient data, if the threshold of this state is defined as a reference threshold vector signal value, then a vector signal value (hereinafter referred to as an abnormal vector signal) larger than the reference threshold vector signal value is a normal livestock. It can be easily seen from the movement of the data that it cannot be measured.

Furthermore, such an abnormal vector signal can be largely classified as a case in which noise is generated in data acquired from an acceleration sensor instantaneously or the acceleration sensor itself is a malfunction, and among these, a malfunction of the sensor is also described later. If the method is used, it can be accurately detected.

That is, by counting the number of occurrences of an abnormal vector signal from a specific sensor (vector_signal_error_counter), when the counter becomes larger than the reference count, the sensor is determined to be potentially abnormal (malfunction) and detected, including the sensor Replace the device.

In addition, the detection method of an abnormal sensor using such a counter is also reset by a counter at regular intervals (for example, one year), and the occurrence of abnormal vector signals due to instantaneous noise is accumulated and thus may be mistaken for sensor errors. You can also reduce the probability of being able to. As a result, by using the above-described abnormal vector signal or by utilizing the counter of the abnormal vector signal well, it is possible to predict to some extent about the abnormal operation of the sensor and replace the livestock behavior pattern detection device including the corresponding sensor at an appropriate time. .

The frequency transform used in the above-mentioned frequency transformation step (S220) uses a Fast Fourier Transform, and the sampling used in the sampling step (S230) is the peak value in a specific unit region of each signal. It is characterized by extracting (Max Value) and matrixing.

In the case of the fast Fourier transform described above, the average value of the signal for the data (displacement amount) measured from the sensor is calculated in the same manner as in Equation (1) above, and then used to convert it into the frequency domain. The reason for converting the processed signal from the sensor to the frequency domain is that it is possible to easily determine the data pattern and the like through various classifiers described later.

In addition, the sampling method described above extracts a peak value in a unit domain of a signal converted into a frequency domain and arranges it as a first peak value, a second peak value, a third peak value, and the like, thereby forming a matrix (FrequencyMaxValue Matrix). ). As a result, it is characterized by being able to give a representative for the moment with the greatest amount of change in motion.

In Table 1 below, the values of the vector signals obtained through the above-described conversion process for the X, Y, and Z displacements obtained from the sensor through the present invention are listed in a table.

x y z Vector signal 8584 11876 152 14654.27 9060 13332 1552 16193.66 5636 15308 1272 16362.07 3820 15936 324 16390.65 856 15944 1788 16066.76 440 15764 2160 15917.38 424 16016 2192 16170.87 424 16088 1032 16126.64 6308 14592 1576 15975.01 6608 15152 1128 16568.68 6480 14704 1292 16120.4 6400 14608 1216 15994.76 6720 15208 1128 16664.75 5400 15504 1196 16461 5480 15144 1208 16150.24 4416 15928 1124 16567 4532 15588 1184 16276.57 4228 15840 568 16404.4 3980 16012 720 16514.93 4692 9660 2664 11064.69 560 11640 2724 11967.6 4624 15988 1424 16704.05 4640 15400 2220 16236.32 4464 15848 1856 16568.98 5428 15568 1712 16575.79 5036 15520 1420 16378.28 4828 15652 960 16407.81 3820 15936 324 16390.65

Subsequently, as shown in the shape of a red star in FIG. 3, when the peak value in the unit region is sampled, the above-described FrequencyMaxValue Matrix can be obtained as the matrix disclosed in Equation 2 below.

In addition, the modeling used in the above-described pattern analysis step (S300) is characterized by using a machine learning algorithm.

In the case of the above-described machine learning algorithm, a deep neural network (DNN) may be used, or various types of machine learning algorithms such as a synthetic neural network (CNN) or a circular neural network (RNN) method may be used.

However, as in the present invention, it is preferable to use a support vector machine (SVM) technique, which will be described later, in a prediction technique that can accurately distinguish a specific behavior pattern of a cow.

SVM is one of supervised learning machine learning methods that are mainly used for classification, regression, and outliers detection. For example, among the various methods of classifying two groups of datasets, it can be said that the best way to increase classification accuracy is to be able to accurately distinguish the midpoint at the maximum distance of each group.

In particular, SVM is known as an optimized method for finding an optimal decision boundary that can distinguish a plurality of dimensions well for data having a plurality of dimensions.

In addition, when the SVM technique as described above is applied to the data obtained from the sensors, the number of trains of machine learning increases as the number of animals equipped with the sensor increases and the data transmitted from the sensors accumulates more and more. As a result, the accuracy of modeling acquired through training becomes increasingly high. This feature is that the analysis error through specific mathematical or statistical modeling always has a certain prediction error rate, but the prediction error rate can be improved more and more as the number of training increases. It is an advantage.

As a result, if the above-described matrix is analyzed and predicted using the above-described SVM technique, the behavioral pattern of livestock can be predicted. At this time, the predicted behavior of the above-mentioned livestock means, for example, an action such as standing, moving, and lying.

In addition, by obtaining long-term data on these behaviors, it is possible to predict specific conditions of livestock (eg, estrus, disease, pregnancy, etc., which are important for raising livestock).

In the case of FIG. 4, the behavior pattern analysis method as described above is applied to the cow, and the cow behavior is analyzed sequentially from the measured acceleration sensor data.

That is, it is shown that the acceleration data and this are converted into vector signals, and then the cow behavior pattern is predicted through fast Fourier transform and peak-to-peak sampling.

As a result, Figure 4 shows that the analysis of the two behavior patterns was formed and predicted differently.

In addition, through the analysis method as shown in FIG. 4, it was confirmed by comparing the actual state of the cow that the accuracy of the cow's behavior pattern was 82 to 92%.

This not only improves the accuracy of the observation of a cow's condition through the naked eye of a conventional expert about 50% or so, but also reduces the labor force and increases productivity by eliminating the need for visual observation.

In addition, the training set that is one of the characteristics of the machine learning method (learning) and the number of training increases, the accuracy is considered to be increased, and this increased accuracy is again modeled in the entire behavior pattern analysis method and machine learning model of the device Considering that it can be easily reflected in, it is expected that the accuracy of predicting the behavioral patterns of livestock can be increased.

In addition, although not shown in the drawings, in the present invention, in the case of a remote server that collects various sensor data and performs a role of machine learning, it may be in the form of one control server or may be provided in the form of a cloud service.

In addition, although not shown in the drawings, the result of the corresponding behavior pattern analysis method can be easily displayed in the form of a web on the manager's PC or installed in the form of an application on the manager's smartphone to easily check the situation. That is, the detection method of the present invention can easily provide information to the user, and the detection method provision type may be a form of an application installed on the user's smart phone, or it may be made in the form of a specific web page and suitable authentication means. It may be in the form of a web-based page that can be accessed via.

In addition, when the livestock is limited to cattle as an embodiment of the present invention, it is possible to effectively predict the estrus, pregnancy, disease, etc. of the cow through the behavior pattern detection method of the present invention.

Specifically, in the case of cattle, it is known that the amount of activity increases when the estrus occurs, and the amount of activity occurs when a disease occurs. In addition, in the conventional case, it is known that the accuracy of using a method of examining or predicting a change in the body of a cow such as horns through the human eye of a breeder is about 50%.

On the other hand, if the behavior pattern detection method of the present invention using frequency analysis and modeling is applied to the data of the acceleration sensor, it is possible to effectively predict the estrus time of cattle than the conventional visual examination method.

The method for detecting livestock behavior patterns according to an embodiment of the present invention may be implemented as a computer program stored in a storage medium for execution through combination with a computer.

In addition, the operation method of the livestock behavior pattern detection method according to an embodiment of the present invention can be recorded in a computer-readable medium by being implemented in the form of program instructions that can be executed through various computer means. The computer-readable medium may include program instructions, data files, data structures, or the like alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the present invention, or may be known and available to those skilled in computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs, DVDs, and magnetic media such as floptical disks. Includes hardware devices specifically configured to store and execute program instructions such as magneto-optical media, and ROM, RAM, flash memory, and solid state drives (SSDs). Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter, etc., as well as machine language codes produced by a compiler.

In addition, a configuration such as a computer or a computer program used in the present invention is modified in the form of a mobile communication terminal, such as a smartphone, and as computing power dramatically increases, in the same sense as an application executed in a smartphone or smartphone. Can be used.

Although described above with reference to preferred embodiments of the present invention, those skilled in the art variously modify and change the present invention without departing from the spirit and scope of the present invention as set forth in the claims below. You can understand that you can.

100: sensing member
101: sensor
102: transmission unit
103: battery unit
104: CPU
105: PCB
200: housing
300: fastening member
400: weight adjusting member

Claims (7)

In the method for detecting livestock behavior patterns using sensors,
The detection method
A data collection step of collecting data from the sensor (S100);
A feature point generating step of analyzing the collected data and generating a feature point (S200); And
And a pattern analysis step (S300) of detecting behavior patterns by passing the feature points through a modeled classifier.
According to claim 1,
The sensor is an acceleration sensor, and the data is a livestock behavior pattern detection method, characterized in that the displacement of the X-axis, Y-axis, and Z-axis.
According to claim 2,
The feature point generating step (S200) includes a vector signal forming step (S210) of forming a vector signal based on the displacement amount of the X-axis, Y-axis, and Z-axis;
A frequency conversion step of converting the vector signal into a frequency domain signal (S220);
Sampling the converted frequency domain signal (S230);
Livestock behavior pattern detection method further comprising a.
According to claim 3,
The modeling used in the pattern analysis step (S300) is a livestock behavior pattern detection method characterized in that it uses a machine learning algorithm.
According to claim 3,
The frequency transform used in the frequency transformation step (S220) uses a Fast Fourier Transform,
The sampling used in the sampling step (S230) is a livestock behavior pattern detection method characterized in that the peak value (Max Value) in a specific unit region of each signal is extracted and matrixed.
A computer program for performing the method for detecting livestock behavior patterns according to any one of claims 1 to 5, and storing them in a computer-readable recording medium.
A method for detecting livestock behavior patterns according to any one of claims 1 to 5, and an apparatus for detecting livestock behavior patterns attached to livestock.

Images