CN112883999A - Pedometer system and method for detecting abnormal movement of dairy cow - Google Patents
Pedometer system and method for detecting abnormal movement of dairy cow Download PDFInfo
- Publication number
- CN112883999A CN112883999A CN202110041210.5A CN202110041210A CN112883999A CN 112883999 A CN112883999 A CN 112883999A CN 202110041210 A CN202110041210 A CN 202110041210A CN 112883999 A CN112883999 A CN 112883999A
- Authority
- CN
- China
- Prior art keywords
- layer
- pedometer
- cow
- neural network
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 17
- 235000013365 dairy product Nutrition 0.000 title claims abstract description 16
- 208000012661 Dyskinesia Diseases 0.000 title claims description 6
- 238000001514 detection method Methods 0.000 claims abstract description 31
- 230000033001 locomotion Effects 0.000 claims abstract description 30
- QSHDDOUJBYECFT-UHFFFAOYSA-N mercury Chemical compound [Hg] QSHDDOUJBYECFT-UHFFFAOYSA-N 0.000 claims abstract description 30
- 229910052753 mercury Inorganic materials 0.000 claims abstract description 30
- 238000004891 communication Methods 0.000 claims abstract description 24
- 238000013527 convolutional neural network Methods 0.000 claims description 37
- 238000013528 artificial neural network Methods 0.000 claims description 25
- 239000011159 matrix material Substances 0.000 claims description 21
- 238000011176 pooling Methods 0.000 claims description 21
- 239000013598 vector Substances 0.000 claims description 21
- 230000005856 abnormality Effects 0.000 claims description 12
- 238000012544 monitoring process Methods 0.000 claims description 12
- 210000002569 neuron Anatomy 0.000 claims description 6
- 239000008267 milk Substances 0.000 claims description 5
- 235000013336 milk Nutrition 0.000 claims description 5
- 210000004080 milk Anatomy 0.000 claims description 5
- WHXSMMKQMYFTQS-UHFFFAOYSA-N Lithium Chemical compound [Li] WHXSMMKQMYFTQS-UHFFFAOYSA-N 0.000 claims description 4
- 230000012173 estrus Effects 0.000 claims description 4
- 230000006870 function Effects 0.000 claims description 4
- 229910052744 lithium Inorganic materials 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 4
- 238000012549 training Methods 0.000 claims description 4
- 206010012735 Diarrhoea Diseases 0.000 claims description 3
- 229910000831 Steel Inorganic materials 0.000 claims description 3
- 230000004913 activation Effects 0.000 claims description 3
- 230000005540 biological transmission Effects 0.000 claims description 3
- 210000004027 cell Anatomy 0.000 claims description 3
- 208000030175 lameness Diseases 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 230000035945 sensitivity Effects 0.000 claims description 3
- 239000010959 steel Substances 0.000 claims description 3
- 230000002159 abnormal effect Effects 0.000 abstract description 5
- 230000001133 acceleration Effects 0.000 abstract description 3
- 238000013461 design Methods 0.000 abstract description 2
- 244000144972 livestock Species 0.000 abstract description 2
- 241000283690 Bos taurus Species 0.000 description 26
- 238000010586 diagram Methods 0.000 description 5
- 201000010099 disease Diseases 0.000 description 2
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 2
- 230000002354 daily effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 210000000003 hoof Anatomy 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000004575 stone Substances 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C22/00—Measuring distance traversed on the ground by vehicles, persons, animals or other moving solid bodies, e.g. using odometers, using pedometers
- G01C22/006—Pedometers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Arrangements For Transmission Of Measured Signals (AREA)
Abstract
The invention discloses a pedometer system and a method for detecting the abnormal motion of a dairy cow, and the system comprises a hardware module and a software algorithm, wherein the hardware module comprises a mercury switch step-counting sensor, a wireless communication module, a power supply, a gateway receiver and a server, the mercury switch step-counting sensor, the wireless communication module, the power supply, the gateway receiver and the server form a pedometer detection algorithm, the mercury switch step-counting sensor and the wireless communication module are electrically connected with the power supply, the gateway receiver is wirelessly connected with the wireless communication module, the software algorithm comprises a positioning algorithm and an abnormal detection algorithm, and the technical field of the livestock pedometer is particularly related. According to the invention, through setting the pedometer detection algorithm, the power consumption of the mercury switch in the pedometer detection algorithm is lower than that of the three-dimensional acceleration sensor, the CC integrates the CPU and the wireless radio frequency, the data can be processed and transmitted through wireless communication, and the overall design of the pedometer is lower than that of the traditional method.
Description
Technical Field
The invention relates to the technical field of livestock pedometers, in particular to a pedometer system and a method for detecting abnormal motion of a dairy cow.
Background
In a large-scale dairy farm, a pedometer is a very important management tool. The method is a basic stone for pasture informatization and accurate cultivation, and has the functions of identity recognition, exercise step number statistics and abnormality diagnosis based on exercise amount. Statistically, 98% of the cows are abnormal and can be detected by pedometer hair-off, 90% of the oestrus and 30% of hoof disease dependent pedometers.
An existing pedometer, such as an afielden (Afimilk) pedometer in israel, mainly comprises a sensor ADXL362, a wireless communication module, a main control CPU and a lithium battery. ADXL362 is a three-dimensional acceleration sensor, with high power consumption; the wireless communication distance is short, only when the dairy cow passes through a specific position, the card reader reads pedometer data worn by the dairy cow in a short distance, namely the data is limited by space and time, real-time or timing reading cannot be achieved, and the data can be read at any position in a dairy cow shed. On the other hand, after the traditional pedometer transmits the acquired cow motion amount information to the server, an experienced medical specialist needs to observe the motion curve rule and characteristics by virtue of personal experience so as to find out the individual abnormality of the cow, and the method is low in efficiency.
Disclosure of Invention
In order to overcome the above defects in the prior art, embodiments of the present invention provide a pedometer system for detecting abnormal cow movement, and the problems to be solved by the present invention are: the prior pedometer transmits the acquired milk cow motion amount information to the server, and an experienced medical specialist needs to observe the motion curve rule and characteristics by virtue of personal experience, so as to find out the abnormality of the milk cow individual, and the method has low efficiency.
In order to achieve the purpose, the invention provides the following technical scheme: the utility model provides a pedometer system for milk cow motion anomaly detects, includes hardware module and software algorithm, hardware module includes mercury switch meter step sensor, wireless communication module, power, gateway receiver and server, and mercury switch meter step sensor, wireless communication module, power, gateway receiver and server constitute pedometer detection algorithm, mercury switch meter step sensor and wireless communication module carry out electric connection with the power, the gateway receiver carries out wireless connection with wireless communication module, the software algorithm includes positioning algorithm and anomaly detection algorithm.
Preferably, the mercury switch step counting sensor forms an inclination angle of 35-45 degrees with the ground through two mercury switches, the current is conducted through the vibration of the motion of the cow, and the conducting circuit lasts for 100 Ms-1000 Ms to serve as a step counting detection device.
Preferably, the power supply adopts 1800mAh lithium subcell.
Preferably, the pedometer detection algorithm is implemented by two interrupt processes of the CC 2530.
Preferably, the positioning algorithm is based on the CC2530 wireless signal strength, and the positioning algorithm adopts the signal strength attenuation transmitted by the CC2530, and estimates the position information of the cow in the cowshed by deploying a grid-shaped receiving gateway and a weighted positioning algorithm in the cowshed.
Preferably, the anomaly detection algorithm comprises a two-dimensional convolutional neural network (2D CNN) and a deep neural network;
preferably, the pedometer detection algorithm comprises the following steps: 1) two mercury switches and the ground form an inclination angle between theta and (35-45 degrees), and when any one mercury switch is switched on due to sensing vibration, a timer is started to count;
2) only when two switches detect vibration at the same time, the circuit is conducted and the time lasts for Tmin=100MsAbove Tmax=1000MsWhen the number of the steps is less than 1, the mercury switch triggers the step counting by means of external interruption, so that the power consumption can be further saved;
3) when the step counter reaches 24 hours (48 bytes), a data packet is sent to the server once, the cache is cleared, and monitoring data can be sent to the gateway at regular time under the instruction of the server.
Preferably, the positioning algorithm comprises the steps of: 1) uniformly deploying the convergence gateways on a steel frame of the cowshed, and regularly arranging the convergence gateways into a queue grid form;
2) in the grid, all convergence gateways are located in the monitoring area, and the monitoring area D is decomposed into n2Or n × m cells;
3) the communication radius of CC2530 is set to be greater than 1 unit and less than or equal to 2 unit areas, i.e. a square area with a side length of 2 units, and the attenuation relation of the signal of CC2530 to the distance is:
los=32.44+20lg(d)+20lg(d)
wherein los is propagation loss, the unit is dB, d is the distance between the sending end and the receiving end, the unit is km, f is the working frequency of the electromagnetic wave, the unit is MHz, the transmission distance d can be estimated according to the signal intensity value of the receiving end, but d is greatly influenced by the environment, so that the d can only be used as a weighting factor, and the positioning formula of the target is obtained as follows:
wherein d isiRepresents the distance from the jth gateway to the ith target, (x)j yj zj)TRepresents the jth gateway location, and (x)CyC zC)TRepresenting the estimated target position.
Preferably, the anomaly detection algorithm comprises the steps of: 1) two-dimensional convolutional neural network (2D CNN): taking a one-day-long monitoring unit as an example, the one-dimensional vector is N ═ 48 numbers (each half hour occupies one byte of motion), that is, the vector input to the convolutional neural network is: n (n1, n2, n3... nN), each input vector is a sample, each sample is labeled by the data of a herdsman, and the data of the cow motion in four states of normal, lameness, diarrhea and oestrus are shown in fig. 4, but the data set is still continuously expanded;
2) deep neural network:
an input layer: after the data are preprocessed, each record seed comprises 48 data points (48 records are formed in each half hour of the motion data of the dairy cow in one day), so that a vector V of 48 multiplied by 1 is obtained, and the vector is input into a neural network seed;
first 1D CNN layer: setting a filter with the height of 6, namely the size of the Convolition Kernel, in the first layer, defining the filter to learn and extract features in the first layer, and defining 100 filters in the first layer to extract enough features, so that 100 different features are learned in the first layer of the network, the output of the first neural network is a 43 x 100 matrix, and each column of the output matrix contains a filter weight;
second 1D CNN layer: the output result of the first CNN is input into the second CNN layer, and we define 100 different filters in this layer for training, the size of the output matrix is 38 × 100, and each column of the output matrix contains a filter weight;
maximum pooling layer: this layer is to reduce the complexity of the output and to prevent overfitting of the data, the largest pooling layer is used after the CNN layer, in the present invention we select the pooling layer with the scale of 3, which means that the output layer matrix size is one third of the input layer, i.e. the output matrix size is 12 × 100;
third and fourth 1D CNN layers: the two layers of CNN are set for learning the characteristics of higher layers, and the output of the neural network after passing through the two layers is a 2 x 96 matrix;
average pooling layer: also here the purpose of adding pooling layers is to prevent overfitting, here average pooling instead of maximum pooling, and finally the output of the neural network is 1 x 96, each feature detector leaving only one weight in this layer of the neural network;
dropout layer: the layer randomly assigns 0 weight to each neuron in the neural network; by selecting a proportionality coefficient of 0.5, 50% of neurons will be cleared, and through the operation, the sensitivity of the network to small changes of data can be greatly reduced, so that the accuracy of processing invisible data can be improved, and the output of the layer is still 1 × 96;
full connection layer: finally, using the softmax activation function, classifying the vectors with the length of 96 according to the four categories in fig. 4 to obtain 4 vectors, which represent the probability of occurrence of each of the 4 categories.
The invention has the technical effects and advantages that:
1. according to the invention, through setting a pedometer detection algorithm, a mercury switch step-counting sensor in the pedometer detection algorithm has lower power consumption than a three-dimensional acceleration sensor, and the CC2530 integrates a CPU and a wireless radio frequency at the same time, so that data can be processed and can be sent through wireless communication, and the overall design of the pedometer is lower in power consumption than that of the traditional method;
2. according to the invention, the positioning algorithm is set, and the position estimation of the target is realized according to the CC2530 signal intensity and the grid positioning algorithm, so that the wireless signal is only used, a distance measuring system assisted by any hardware is not needed, the power consumption is low, and only when the pedometer sends data, the card reader arranged in the cowshed is used as a position reference point to estimate the position of the target, so that the low cost is realized.
3. According to the invention, by setting the anomaly detection algorithm, because various diseases or anomalies of the dairy cow almost show movement, various anomalies of the dairy cow can be detected by a deep learning algorithm by observing the movement data of the dairy cow in each time period every day. The method combines the daily movement data curve of the dairy cow with the clinical experience and data annotation of a herdsman expert, trains and learns by using an abnormality detection algorithm based on the one-dimensional convolutional neural network SSD, and finally can quickly diagnose the movement abnormality of the dairy cow.
Drawings
FIG. 1 is a schematic diagram of a circuit diagram according to the present invention;
FIG. 2 is a schematic flow chart of a pedometer control process according to the present invention;
FIG. 3 is a schematic diagram of a data annotation sample according to the present invention;
FIG. 4 is a pedometer dataset for training and testing;
FIG. 5 is a diagram of a neural network architecture of the present invention;
FIG. 6 is a diagram showing the result of the positioning algorithm of the present invention.
The reference signs are: 1. a mercury switch step-counting sensor; 2. a wireless communication module; 3. a power source; 4. a gateway receiver; 5. a server; 6. a positioning algorithm; 7. an anomaly detection algorithm; 8. a pedometer detection algorithm.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention discloses a pedometer system for detecting the movement abnormality of a cow, which comprises a hardware module and a software algorithm, wherein the hardware module comprises a mercury switch step-counting sensor 1, a wireless communication module 2, a power supply 3, a gateway receiver 4 and a server 5, the mercury switch step-counting sensor 1, the wireless communication module 2, the power supply 3, the gateway receiver 4 and the server 5 form a pedometer detection algorithm 8, the mercury switch step-counting sensor 1 and the wireless communication module 2 are electrically connected with the power supply 3, the gateway receiver 4 is wirelessly connected with the wireless communication module 2, and the software algorithm comprises a positioning algorithm 6 and an abnormality detection algorithm 7.
The mercury switch step counting sensor 1 forms an inclination angle of 35 degrees to 45 degrees with the ground through two mercury switches, current is conducted through vibration of milk cow movement, and the conducting circuit lasts for 100Ms to 1000Ms to serve as a step counting detection device.
The power supply 3 adopts 1800mAh lithium subcell.
The pedometer detection algorithm 8 is realized through two interrupt processes of the CC2530, a control flow chart is shown in figure 2, and the step identification algorithm is practical and simple and has low power consumption.
The positioning algorithm 6 is a positioning algorithm based on the CC2530 wireless signal strength, the positioning algorithm 6 adopts the signal strength attenuation transmitted by the CC2530, and estimates the position information of the cow in the cowshed by deploying a latticed receiving gateway and a weighted positioning algorithm in the cowshed.
The anomaly detection algorithm 7 includes a two-dimensional convolutional neural network (2D CNN) and a deep neural network.
The pedometer detection algorithm comprises the following steps: 1) two mercury switches and the ground form an inclination angle between theta and (35-45 degrees), and when any one mercury switch is switched on due to sensing vibration, a timer is started to count;
2) only when two switches detect vibration at the same time, the circuit is conducted and the time lasts for Tmin=100MsAbove Tmax=1000MsWhen the number of the steps is less than 1, the mercury switch triggers the step counting by means of external interruption, so that the power consumption can be further saved;
3) when the step counter reaches 24 hours (48 bytes), a data packet is sent to the server once, the cache is cleared, and monitoring data can be sent to the gateway at regular time under the instruction of the server.
The positioning algorithm comprises the following steps: 1) uniformly deploying the convergence gateways on a steel frame of the cowshed, and regularly arranging the convergence gateways into a queue grid form;
2) in the grid, all convergence gateways are located in the monitoring area, and the monitoring area D is decomposed into n2Or n × m cells;
3) the communication radius of CC2530 is set to be greater than 1 unit and less than or equal to 2 unit areas, i.e. a square area with a side length of 2 units, and the attenuation relation of the signal of CC2530 to the distance is:
los=32.44+20lg(d)+20lg(d)
wherein los is propagation loss, the unit is dB, d is the distance between the sending end and the receiving end, the unit is km, f is the working frequency of the electromagnetic wave, the unit is MHz, the transmission distance d can be estimated according to the signal intensity value of the receiving end, but d is greatly influenced by the environment, so that the d can only be used as a weighting factor, and the positioning formula of the target is obtained as follows:
wherein d isiRepresents the distance from the jth gateway to the ith target, (x)j yj zj)TRepresents the jth gateway location, and (x)CyC zC)TRepresenting the estimated target position.
The anomaly detection algorithm comprises the following steps: 1) two-dimensional convolutional neural network (2D CNN): taking a one-day-long monitoring unit as an example, the one-dimensional vector is N ═ 48 numbers (each half hour occupies one byte of motion), that is, the vector input to the convolutional neural network is: n (n1, n2, n3... nN), each input vector is a sample, each sample is labeled by the data of a herdsman, and the data of the cow motion in four states of normal, lameness, diarrhea and oestrus are shown in fig. 4, but the data set is still continuously expanded;
2) deep neural network:
an input layer: after the data are preprocessed, each record seed comprises 48 data points (48 records are formed in each half hour of the motion data of the dairy cow in one day), so that a vector V of 48 multiplied by 1 is obtained, and the vector is input into a neural network seed;
first 1D CNN layer: setting a filter with the height of 6, namely the size of the Convolition Kernel, in the first layer, defining the filter to learn and extract features in the first layer, and defining 100 filters in the first layer to extract enough features, so that 100 different features are learned in the first layer of the network, the output of the first neural network is a 43 x 100 matrix, and each column of the output matrix contains a filter weight;
second 1D CNN layer: the output result of the first CNN is input into the second CNN layer, and we define 100 different filters in this layer for training, the size of the output matrix is 38 × 100, and each column of the output matrix contains a filter weight;
maximum pooling layer: this layer is to reduce the complexity of the output and to prevent overfitting of the data, the largest pooling layer is used after the CNN layer, in the present invention we select the pooling layer with the scale of 3, which means that the output layer matrix size is one third of the input layer, i.e. the output matrix size is 12 × 100;
third and fourth 1D CNN layers: the two layers of CNN are set for learning the characteristics of higher layers, and the output of the neural network after passing through the two layers is a 2 x 96 matrix;
average pooling layer: also here the purpose of adding pooling layers is to prevent overfitting, here average pooling instead of maximum pooling, and finally the output of the neural network is 1 x 96, each feature detector leaving only one weight in this layer of the neural network;
dropout layer: the layer randomly assigns 0 weight to each neuron in the neural network; by selecting a proportionality coefficient of 0.5, 50% of neurons will be cleared, and through the operation, the sensitivity of the network to small changes of data can be greatly reduced, so that the accuracy of processing invisible data can be improved, and the output of the layer is still 1 × 96;
full connection layer: finally, using the softmax activation function, classifying the vectors with the length of 96 according to the four categories in fig. 4 to obtain 4 vectors, which represent the probability of occurrence of each of the 4 categories.
And finally: the above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that are within the spirit and principle of the present invention are intended to be included in the scope of the present invention.
Claims (9)
1. A pedometer system for detecting the abnormal movement of a cow comprises a hardware module and a software algorithm, and is characterized in that: the hardware module comprises a mercury switch step counting sensor (1), a wireless communication module (2), a power supply (3), a gateway receiver (4) and a server (5), the mercury switch step counting sensor (1) and the wireless communication module (2) are electrically connected with the power supply (3), a pedometer detection algorithm (8) is formed by the mercury switch step counting sensor (1), the wireless communication module (2), the power supply (3), the gateway receiver (4) and the server (5), the gateway receiver (4) is wirelessly connected with the wireless communication module (2), and the software algorithm comprises a positioning algorithm (6) and an anomaly detection algorithm (7).
2. The pedometer system for detecting cow motion abnormalities according to claim 1, wherein: the mercury switch step counting sensor (1) forms an inclination angle of 35 degrees to 45 degrees with the ground through two mercury switches, current is conducted through vibration of milk cow movement, and the conducting circuit lasts for 100Ms to 1000Ms and serves as a step counting detection device.
3. The pedometer system for detecting cow motion abnormalities according to claim 1, wherein: the power supply (3) adopts 1800mAh lithium subcell.
4. The pedometer system for detecting cow motion abnormalities according to claim 1, wherein: the pedometer detection algorithm (8) is implemented by two interrupt processes of the CC 2530.
5. The pedometer system for detecting cow motion abnormalities according to claim 1, wherein: the positioning algorithm (6) is based on CC2530 wireless signal strength, the positioning algorithm (6) adopts signal strength attenuation transmitted by the CC2530, and the position information of the cow in the cowshed is estimated by deploying a latticed receiving gateway and a weighted positioning algorithm in the cowshed.
6. The pedometer system for detecting cow motion abnormalities according to claim 1, wherein: the anomaly detection algorithm (7) includes a two-dimensional convolutional neural network (2D CNN) and a deep neural network.
7. The method for detecting the abnormal movement of the cow according to claim 1, wherein: the pedometer detection algorithm (8) comprises the steps of: 1) two mercury switches and the ground form an inclination angle between theta and (35-45 degrees), and when any one mercury switch is switched on due to sensing vibration, a timer is started to count;
2) only when two switches detect vibration at the same time, the circuit is conducted and the time lasts for Tmin=100MsAbove Tmax=1000MsWhen the number of the steps is less than 1, the mercury switch triggers the step counting by means of external interruption, so that the power consumption can be further saved;
3) when the step counter reaches 24 hours (48 bytes), a data packet is sent to the server once, the cache is cleared, and monitoring data can be sent to the gateway at regular time under the instruction of the server.
8. The method for detecting the abnormal movement of the cow according to claim 1, wherein: the positioning algorithm (6) comprises the steps of: 1) uniformly deploying the convergence gateways on a steel frame of the cowshed, and regularly arranging the convergence gateways into a queue grid form;
2) in the grid, all convergence gateways are located in the monitoring area, and the monitoring area D is decomposed into n2Or n × m cells;
3) the communication radius of CC2530 is set to be greater than 1 unit and less than or equal to 2 unit areas, i.e. a square area with a side length of 2 units, and the attenuation relation of the signal of CC2530 to the distance is:
los=32.44+20lg(d)+20lg(d)
wherein los is propagation loss, the unit is dB, d is the distance between the sending end and the receiving end, the unit is km, f is the working frequency of the electromagnetic wave, the unit is MHz, the transmission distance d can be estimated according to the signal intensity value of the receiving end, but d is greatly influenced by the environment, so that the d can only be used as a weighting factor, and the positioning formula of the target is obtained as follows:
wherein d isiRepresents the distance from the jth gateway to the ith target, (x)j yj zj)TRepresents the jth gateway location, and (x)C yCzC)TRepresenting the estimated target position.
9. The method for detecting the abnormal movement of the cow according to claim 1, wherein: the anomaly detection algorithm (7) comprises the steps of: 1) two-dimensional convolutional neural network (2D CNN): taking a one-day-long monitoring unit as an example, the one-dimensional vector is N ═ 48 numbers (each half hour occupies one byte of motion), that is, the vector input to the convolutional neural network is: n (n1, n2, n3... nN), each input vector is a sample, each sample is labeled by the data of a herdsman, and the data of the cow motion in four states of normal, lameness, diarrhea and oestrus are shown in fig. 4, but the data set is still continuously expanded;
2) deep neural network:
an input layer: after the data are preprocessed, each record seed comprises 48 data points (48 records are formed in each half hour of the motion data of the dairy cow in one day), so that a vector V of 48 multiplied by 1 is obtained, and the vector is input into a neural network seed;
first 1D CNN layer: setting a filter with the height of 6, namely the size of the Convolition Kernel, in the first layer, defining the filter to learn and extract features in the first layer, and defining 100 filters in the first layer to extract enough features, so that 100 different features are learned in the first layer of the network, the output of the first neural network is a 43 x 100 matrix, and each column of the output matrix contains a filter weight;
second 1D CNN layer: the output result of the first CNN is input into the second CNN layer, and we define 100 different filters in this layer for training, the size of the output matrix is 38 × 100, and each column of the output matrix contains a filter weight;
maximum pooling layer: this layer is to reduce the complexity of the output and to prevent overfitting of the data, the largest pooling layer is used after the CNN layer, in the present invention we select the pooling layer with the scale of 3, which means that the output layer matrix size is one third of the input layer, i.e. the output matrix size is 12 × 100;
third and fourth 1D CNN layers: the two layers of CNN are set for learning the characteristics of higher layers, and the output of the neural network after passing through the two layers is a 2 x 96 matrix;
average pooling layer: also here the purpose of adding pooling layers is to prevent overfitting, here average pooling instead of maximum pooling, and finally the output of the neural network is 1 x 96, each feature detector leaving only one weight in this layer of the neural network;
dropout layer: the layer randomly assigns 0 weight to each neuron in the neural network; by selecting a proportionality coefficient of 0.5, 50% of neurons will be cleared, and through the operation, the sensitivity of the network to small changes of data can be greatly reduced, so that the accuracy of processing invisible data can be improved, and the output of the layer is still 1 × 96;
full connection layer: finally, using the softmax activation function, classifying the vectors with the length of 96 according to the four categories in fig. 4 to obtain 4 vectors, which represent the probability of occurrence of each of the 4 categories.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110041210.5A CN112883999A (en) | 2021-01-13 | 2021-01-13 | Pedometer system and method for detecting abnormal movement of dairy cow |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110041210.5A CN112883999A (en) | 2021-01-13 | 2021-01-13 | Pedometer system and method for detecting abnormal movement of dairy cow |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112883999A true CN112883999A (en) | 2021-06-01 |
Family
ID=76045195
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110041210.5A Pending CN112883999A (en) | 2021-01-13 | 2021-01-13 | Pedometer system and method for detecting abnormal movement of dairy cow |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112883999A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113973739A (en) * | 2021-09-13 | 2022-01-28 | 安徽大学 | Multi-functional electronic ear tag for ox |
CN114881187A (en) * | 2022-05-23 | 2022-08-09 | 温州格洛博电子有限公司 | Passive radio frequency sensing tag and step counting method |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104768219A (en) * | 2014-01-06 | 2015-07-08 | 南京宜开数据分析技术有限公司 | Passive positioning and tracking method based on wireless sensor network |
CN106871923A (en) * | 2015-12-10 | 2017-06-20 | 哈尔滨黑石科技有限公司 | A kind of multi-functional dairy cow pedometer |
CN107124696A (en) * | 2017-03-20 | 2017-09-01 | 东华大学 | A kind of keyless systems localization method based on bluetooth RSSI |
CN107533140A (en) * | 2015-03-06 | 2018-01-02 | 看门人系统公司 | The low energy consumption positioning of loose impediment |
CN107801168A (en) * | 2017-08-17 | 2018-03-13 | 龙岩学院 | A kind of localization method of the adaptive passive type target in outdoor |
CN110476879A (en) * | 2019-08-26 | 2019-11-22 | 重庆邮电大学 | Milk cow behavior discriminant classification method and device based on multi-tag chain type ecological environment |
CN111339838A (en) * | 2020-02-10 | 2020-06-26 | 中国农业科学院农业信息研究所 | Pig behavior identification method and system based on information fusion |
CN112067015A (en) * | 2020-09-03 | 2020-12-11 | 青岛歌尔智能传感器有限公司 | Step counting method and device based on convolutional neural network and readable storage medium |
-
2021
- 2021-01-13 CN CN202110041210.5A patent/CN112883999A/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104768219A (en) * | 2014-01-06 | 2015-07-08 | 南京宜开数据分析技术有限公司 | Passive positioning and tracking method based on wireless sensor network |
CN107533140A (en) * | 2015-03-06 | 2018-01-02 | 看门人系统公司 | The low energy consumption positioning of loose impediment |
CN106871923A (en) * | 2015-12-10 | 2017-06-20 | 哈尔滨黑石科技有限公司 | A kind of multi-functional dairy cow pedometer |
CN107124696A (en) * | 2017-03-20 | 2017-09-01 | 东华大学 | A kind of keyless systems localization method based on bluetooth RSSI |
CN107801168A (en) * | 2017-08-17 | 2018-03-13 | 龙岩学院 | A kind of localization method of the adaptive passive type target in outdoor |
CN110476879A (en) * | 2019-08-26 | 2019-11-22 | 重庆邮电大学 | Milk cow behavior discriminant classification method and device based on multi-tag chain type ecological environment |
CN111339838A (en) * | 2020-02-10 | 2020-06-26 | 中国农业科学院农业信息研究所 | Pig behavior identification method and system based on information fusion |
CN112067015A (en) * | 2020-09-03 | 2020-12-11 | 青岛歌尔智能传感器有限公司 | Step counting method and device based on convolutional neural network and readable storage medium |
Non-Patent Citations (1)
Title |
---|
黄小平: "基于多传感器的奶牛个体信息感知与体况评分方法研究", 《万方数据库》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113973739A (en) * | 2021-09-13 | 2022-01-28 | 安徽大学 | Multi-functional electronic ear tag for ox |
CN114881187A (en) * | 2022-05-23 | 2022-08-09 | 温州格洛博电子有限公司 | Passive radio frequency sensing tag and step counting method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109145032B (en) | Intelligent monitoring method and system for bee breeding | |
WO2019218602A1 (en) | Internet of things system for cow behavior and physiological characteristics, and operation method therefor | |
CN107157458A (en) | The sensor-based system and method for a kind of animal individual feed intake and health monitoring | |
CN107027650A (en) | A kind of boar abnormal state detection method and device based on PSO SVM | |
CN112883999A (en) | Pedometer system and method for detecting abnormal movement of dairy cow | |
Arablouei et al. | Animal behavior classification via deep learning on embedded systems | |
CN105091938A (en) | Poultry health monitoring method and system | |
WO2018109725A1 (en) | Process and device for the detection of estrus in a ruminant animal | |
US20210338165A1 (en) | Method of managing disease, and apparatuses operating the same | |
Feiyang et al. | Monitoring behavior of poultry based on RFID radio frequency network | |
JP2021035355A (en) | Health condition management system for livestock, wearable device for livestock, health condition management method for livestock, and program | |
CN112568141A (en) | Supervision system for preventing diseases of pigs | |
CN112544491A (en) | Method and system for identifying diarrhea of captive animals | |
Shen et al. | Automatic recognition method of cow ruminating behaviour based on edge computing | |
CN107019501B (en) | Remote tumble detection method and system based on genetic algorithm and probabilistic neural network | |
CN110532926A (en) | Pig neurogenic disease intelligence Forecasting Method based on deep learning | |
CN107714000B (en) | Milk cow health condition detection method and device | |
Dulyala et al. | Classification system for estrus behavior of cow using an accelerometer | |
CN116935439A (en) | Automatic monitoring and early warning method and automatic monitoring and early warning system for delivery of pregnant sheep | |
CN113973739B (en) | Multifunctional electronic ear tag for cattle | |
CN116563758A (en) | Lion head goose monitoring method, device, equipment and storage medium | |
KR102443610B1 (en) | Method of managing disease, and apparatuses operating the same | |
CN212990135U (en) | Plant disease and insect pest image recognition device based on MobileNet neural network model | |
CN114557292A (en) | Method for detecting cow behaviors and electronic collar | |
Wang | Research of the cow body evaluation system based on Internet of things |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210601 |