CN113456063A

CN113456063A - Artificial intelligence-based dead chicken disease detection system and detection method

Info

Publication number: CN113456063A
Application number: CN202110740241.XA
Authority: CN
Inventors: 包依勤; 许文斌; 卢鸿兵
Original assignee: Nanjing Xiaozhuang University
Current assignee: Xianyang Kangda Modern Agriculture Co.,Ltd.
Priority date: 2021-06-30
Filing date: 2021-06-30
Publication date: 2021-10-01
Anticipated expiration: 2041-06-30
Also published as: CN113456063B

Abstract

The invention discloses a dead chicken disease detection system and a detection method based on artificial intelligence, belonging to the technical field of dead chicken disease detection and comprising the following steps: the coop detection unit is used for acquiring the motion track data of running chickens and analyzing and identifying the state data of the chickens through the motion track data; the front-end processor is used for collecting data of the coop detection unit; the data server is connected with the front-end processor and processes the coop detection unit data collected by the front-end processor; the method is characterized in that a foot ring is buckled on each chicken, the maximum displacement and the three-dimensional total variance of the chicken are measured, the state of the chicken is identified by a machine learning method, so that dead chicken is judged, the activity state of each chicken is displayed by a detection terminal through wireless and foot ring ad hoc networks, the detection terminal is connected to a system through a network, so that the automatic detection of the dead chicken is realized, the identification accuracy rate is 95.6% proved by experiments, the practical effect is achieved, the labor is reduced, and the working efficiency is improved.

Description

Artificial intelligence-based dead chicken disease detection system and detection method

Technical Field

The invention belongs to the technical field of dead chicken disease detection, and particularly relates to a dead chicken disease detection system and a dead chicken disease detection method based on artificial intelligence.

Background

In modern chicken farms, the supply of feed and clean water, the collection of eggs and chicken manure, the control of temperature and wind speed and the like are automated, but the detection of dead chicken and sick chicken still adopts manual timing patrol and inspection, and the untimely inspection and removal of the dead chicken and sick chicken can cause great harm and loss, on one hand, according to medical knowledge, two hours after the chicken die normally, the muscle and blood begin to propagate the pathogenic bacteria, five hours later, the propagated pathogenic bacteria grow exponentially, which can generate great harm, on the other hand, if the chicken die abnormally due to infectious diseases, cross infection among chicken flocks can be caused, if the chicken die untimely removal can cause the chicken farm to suffer great economic loss, on the other hand, the epidemic diseases of the chicken, especially the avian influenza epidemic situation, still can be an important factor for inhibiting the development of the poultry industry, and at present, no effective method for treating and preventing the avian influenza situation exists, once epidemic situation occurs, the epidemic situation can be prevented from spreading only by killing a large number of chickens, huge loss is caused to farmers, whether the epidemic situation can be found in time or not can be found, and a countermeasure is taken as soon as possible, so that the method is an important way for reducing economic loss.

How to discover the sick chicken early is to analyze the characteristics of the sick and dead chicken firstly, and the prior artificial experience analysis method comprises the following steps:

1) touching body temperature, the thigh roots of sick chickens are hot at the top and cold at the bottom, and the cocks are scalded; the thigh roots of healthy chickens are cold at the upper part and hot at the lower part, and the cockscombs do not scald hands;

2) looking up the skin color, the muscles of the sick chicken are mostly thinned, stiff and inelastic, and the skin color is dark red; the skin color of the healthy chicken is reddish, and the muscles are full and elastic;

3) when the sick chicken see the anus, the feathers around the anus are wet and contain excrement or other dirt, the anus is dark red, the excrement is mostly white or yellow green, the fluff near the anus of the healthy chicken is clean and relatively dry, the anus is wet and reddish, and the excrement is semisolid;

4) when the eyes are in focus, the eyes of sick chickens have no spirit, the sick chickens tear, or cheese-like secretion exists around the eyes, and the eyes of individual chickens also lengthen; healthy chicken eyes are bright and have spirit, and the eyeballs are flexible;

5) listening to the sound, namely, the sound of a quiet chicken at 3-5 am, so that the chicken with a Michler's slight call or even no call can be produced, and the sound is similar to a frog call snore, screaming sound or strange call sound; healthy chicken can make a bright, crisp and scream sound;

6) the sick chicken is listless, unstable in pace, inappetence and reluctant to move to eat, and if the chicken dies, the chicken can move still; healthy chicken are lively and well-moving, legs and feet are strong and powerful, and the healthy chicken are willing to move to eat;

in the process of raising chickens, the chickens are subjected to characteristic analysis from the 6 aspects, so that the chickens which are ill or dead can be found as early as possible, however, the large-scale chickens are raised, the number of the chickens is large and the chickens are dense, and hundreds of thousands or even millions of chickens exist, if the chickens are comprehensively detected manually, huge manpower and financial resources are paid, and if the detection method from the 1) -5) is adopted in a small space of a coop , accurate information of the characteristics of the chickens, such as sound, images, temperature and the like, is not easy to obtain, and information noise exists; with the increasing maturity of information processing technology, artificial intelligence technology and internet of things technology, the traditional livestock breeding mode in the past is deeply influenced, but due to the fact that chickens live in a large-scale chicken farm in a small space of a hencoop , a breeding room for breeding 4 ten thousand chickens is provided, the size of each hencoop is about 60cm by 50cm, and 8 chickens need to be bred, as shown in fig. 1. In the space of , it is difficult to automatically find dead chicken, and it is not easy to obtain the characteristic information of feces, sound, image, etc. of chicken, so that there is a problem of great information noise, and finally the recognition rate is not high, so that it can only stay in the theoretical research of ideal environment, and there is great difficulty in practical application, and not only it needs to consider the problem of implementation cost, so it needs to develop a dead chicken detection system and detection method to solve the existing problems.

Disclosure of Invention

The invention aims to provide a dead chicken detection system and a detection method based on artificial intelligence to solve the problem of low efficiency of artificial detection of dead chicken.

In order to achieve the purpose, the invention provides the following technical scheme: a dead chicken disease detection system and a detection method based on artificial intelligence comprise:

the coop detection unit is used for acquiring the motion track data of running chickens and analyzing and identifying the state data of the chickens through the motion track data;

the front-end processor is used for collecting data of the coop detection unit;

the data server is connected with the front-end processor and processes the coop detection unit data collected by the front-end processor;

wherein, chicken coop detecting element includes:

the foot ring is worn on the chicken body and used for acquiring the motion track data of the running of the chicken;

the detection terminal and the foot rings form a wireless network and receive the motion track data transmitted by the foot rings, and the state data of the chickens are analyzed and identified through the motion track data.

Preferably, the foot ring comprises:

the sensor unit is used for acquiring chicken movement acceleration data;

the battery unit is used for supplying power to the foot ring;

and the first microprocessor unit is used for carrying out displacement calculation on the acceleration data acquired by the sensor unit, obtaining the maximum displacement and the three-dimensional total variance and forming the moving track data of the running chicken.

Preferably, the probe terminal includes:

the storage unit is used for storing the acquired acceleration data;

the wireless transmission unit establishes a transmission channel with the foot ring and the front-end processor;

the second microprocessor unit analyzes the state data of the chicken by the acquired foot ring data through an artificial intelligence learning method;

the indicating lamp unit displays the state of the chicken according to the state data of the second microprocessor unit;

and the second microprocessor unit sends the chicken state data to the front-end processor through the wireless transmission unit.

Preferably, the detection system further comprises a remote detection system connected with the data server and used for remote detection, and the remote detection system comprises a WEB monitoring client and a remote detection APP.

The invention also provides a detection method of the dead chicken disease detection system based on artificial intelligence, which comprises the following steps:

s1, the detection terminal carries out displacement calculation on acceleration data collected in the foot ring to obtain maximum displacement and three-dimensional total variance;

s2, training the maximum displacement and the three-dimensional total variance obtained in the step S1 by using a machine learning classification algorithm to obtain an optimal training model;

and S3, recognizing the activity behavior of the chicken by the data collected in the foot ring through a training model.

Preferably, the step of calculating the maximum displacement in step S1 includes:

when the displacement is calculated by integration of the acceleration signal, the obtained acceleration data is a set of discrete values, and when n is larger than 1, the velocity and displacement formula of n points is calculated by the acceleration in the discrete domain, as shown in formulas (1) and (2),

wherein, in the formulas (1) and (2), a [ n ] is the acceleration reading value at the nth moment, v [ n ] and s [ n ] are the speed and displacement values at the corresponding moments, the acquisition interval is set as Deltat, k is a point, the unidirectional movement displacement of the chicken at the n moment can be obtained according to the formulas (1) and (2), the formula (3) is calculated,

equation (3) represents the unidirectional movement displacement at time n.

Preferably, the step of calculating the three-dimensional total variance in step S1 includes:

s11, calculating the displacement from the time period t1 to t2 by sampling the acceleration value, wherein the total sampling times from t1 to t2 are n = (t2-t 1)/[ delta ] t, the real-time motion displacement is S [ n ], and the calculation formula is shown in formula (4)

S12, calculating that there are M time periods (T2-T1) in the T time, see formula (5)

M=T/(t2-t1) (5)

S13, calculating the average displacement in T time, see formula (6)

Where i represents a time period, for a total of M time periods,

s14, calculating the displacement variance of the time period T, wherein the variance describes the deviation degree and fluctuation, see formula (7)

Representing the displacement variance of the time period M.

S15, calculating the variance of displacement in three directions of x, y and z axes of the time period T, as shown in the formula (8-10)

Wherein, formula (8) represents the variance of displacement in the x-axis direction, formula (9) represents the variance of displacement in the y-axis direction, formula (10) represents the variance of displacement in the z-axis direction,

s16, calculating the three-dimensional displacement total variance and the three-dimensional total variance of the time period T, and obtaining the formula (11):

equation (11) represents the total variance of the three-dimensional displacement over T time.

Preferably, the training step of training the model in step S2 includes:

s21, setting acceleration sampling frequency, sampling primary acceleration, and firstly correcting the acceleration, wherein the corrected triaxial accelerations Gx, Gy and Gz are 0;

and S22, setting the displacement time T to calculate the displacement S, and calculating the minute displacement Sx, Sy and Sz respectively according to the formula (4). And calculating the total displacement S = Sx + Sy + Sz;

s23, calculating the accumulated M-minute triaxial displacement variance, calculating the average displacement in M minutes through a formula (6), and calculating the triaxial displacement variances Va, Vb and Vc through formulas (8), (9) and (10);

s24, calculating the accumulated M-minute solid displacement variance, and calculating the solid total variance Vtotal through a formula (11);

s25, calculating the maximum displacement Smax of the accumulated M minutes; and inputting the stereo total variance Vtotal and the maximum displacement Smax data into a machine learning classification algorithm for identifying abnormal chicken behaviors.

Preferably, the training step of training the model in step S2 includes:

first, a sample set definition is performed: sample data: x = (X1, X2), X1 is the maximum displacement Smax, X2 is the total stereo variance Vtotal, the chicken state is classified y, and training is performed by collecting a set amount of data.

Preferably, the machine learning classification algorithm includes: one of SVM, KNN, DT, NB or BP.

The invention has the technical effects and advantages that: according to the dead chicken detection system and the detection method based on artificial intelligence, the maximum displacement and the three-dimensional total variance of a chicken are measured by buckling a foot ring on each chicken, the state of the chicken is identified through a machine learning method, the dead chicken is distinguished, the foot ring of the chicken adopts a 6-axis MPU6050 sensor, the displacement in the three-dimensional direction of the chicken is calculated with high precision, the precision can reach several centimeters, a detection terminal displays the activity state of each chicken through wireless and foot ring ad hoc network, and the detection terminal is connected to the system through the network, so that the automatic detection of the dead chicken is realized, the defects in the prior art are overcome, the system is reasonable, the intelligent degree is high, and the detection method is worthy of popularization and use.

Drawings

FIG. 1 is a schematic structural view of a coop according to the present invention;

FIG. 2 is a schematic structural view of a single coop detection unit of the present invention;

FIG. 3 is a diagram of the detection system architecture of the present invention;

FIG. 4 is a three-axis acceleration data plot of the present invention;

FIG. 5 is a graph of data curves after adjustment of three-axis acceleration in accordance with the present invention;

FIG. 6 is a three-axis acceleration X-axis fluctuation data plot of the present invention;

FIG. 7 is an interface diagram of real-time data according to the present invention;

FIG. 8 is a block diagram of a foot ring assembly according to the present invention;

FIG. 9 is a block diagram of a probing terminal according to the present invention;

FIG. 10 is a flow chart of the machine learning training of the present invention;

FIG. 11 is a comparative histogram of the operating costs of the present invention;

FIG. 12 is a graph comparing the trend of the running cost of the present invention; .

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 provides an artificial intelligence based dead chicken disease detection system as shown in figures 1-12, which comprises:

the coop detection unit is used for acquiring the motion track data of running chickens and analyzing and identifying the state data of the chickens through the motion track data; the front-end processor is used for collecting data of the coop detection unit; the data server is connected with the front-end processor and processes the coop detection unit data collected by the front-end processor; wherein, chicken coop detecting element includes: the foot ring is worn on the chicken body and used for acquiring the motion track data of the running of the chicken;

the detection terminal, constitute wireless network and the activity orbit data that the foot ring conveyed with the foot ring, through the state data of activity orbit data analysis discernment chicken, the foot ring includes: the sensor unit is used for acquiring chicken movement acceleration data;

the battery unit is used for supplying power to the foot ring;

the first microprocessor unit is used for carrying out displacement calculation on acceleration data acquired by the sensor unit, obtaining the maximum displacement and the three-dimensional total variance and forming the moving track data of the running chicken, and the detection terminal comprises: the storage unit is used for storing the acquired acceleration data; the wireless transmission unit establishes a transmission channel with the foot ring and the front-end processor; the second microprocessor unit analyzes the state data of the chicken by the acquired foot ring data through an artificial intelligence learning method; the indicating lamp unit displays the state of the chicken according to the state data of the second microprocessor unit; the chicken coop detection unit consists of 8 foot rings and 1 detection terminal, the 8 foot rings and the 1 detection terminal automatically form a zigbee wireless network, the detection terminal collects sensor data and displays the activity state of the chicken through an indicator light, and the foot rings are designed mainly by a sensor unit, a battery unit and a first microprocessor unit and are as shown in figure 8;

in the embodiment, the sensor unit adopts a sensor MPU-6050 integrating 6-axis motion processing components, compared with the traditional multi-component scheme, the MPU-6050 avoids the problem of difference between a gyroscope and an accelerometer, reduces a large amount of packaging space, becomes small and high in efficiency, when a chicken ring moves, the MPU-6050 generates three-axis acceleration and calculates through a first microprocessor to obtain the maximum displacement and three-dimensional variance of a chicken, and a detection terminal receives the data and indicates the activity intensity of the chicken;

the battery unit is mainly used for supplying power to the whole chicken-ring equipment and ensuring that the equipment can run, and in terms of selecting batteries, the battery volume adopted by the invention is relatively small, so that the whole volume of the foot ring is ensured to be in a small volume amount, the chicken is comfortable to wear, and the phenomenon that the chicken intentionally breaks the chicken ring is avoided; meanwhile, the maximum electric quantity under the relative volume is ensured, and the electric quantity maximization is ensured under the limited basis of the volume, so that the chicken ring can be ensured to run for a long time without frequently replacing equipment, and the economic cost and the time cost are saved;

the first microprocessor unit, which uses the ZigBee chip CC2630 of the Texas Instruments, offers advantages over CC2530, CC2630, which is also a ZigBee bit processor: the CC2630 comprises a 32-bit ARM Cotex-M3 core, compared with an 8051 core of the CC2530, the CC2630 has stronger performance and low cost, meanwhile, the CC2630 is very unique, an ultra-low power consumption sensor controller based on the Cotex-M0, the CC2630 is provided with rich peripheral function sets, and in addition, the function sets are connected with the unique ultra-low power consumption sensor controller of the CC2630, so that the unique connection mode can ensure that the analog and digital data can be autonomously collected under the condition that a main control unit of the singlechip is not awakened, and therefore, the data collection is realized under a ZibgBee network through battery power supply;

the detection terminal mainly comprises a storage unit, a battery unit, an indicator light unit, a wireless transmission unit and a second microprocessor unit, and the schematic block diagram of the detection terminal is shown in fig. 9;

the storage unit mainly comprises a Flash chip, and the storage unit mainly has the following functions: the collected data cannot be uploaded to a front-end processor in a real-time data mode, needs to be stored and is transmitted to a main station at one time at a set time;

a battery unit for supplying power;

the wireless transmission unit mainly comprises ZigBee modules of a Wifi module, the Wifi module is mainly used for uploading acquired sensing data to a front-end processor to facilitate analysis of a master station, the Wifi module adopts ESP32 of Shanghainexin, the ESP32 is set as a terminal after being electrified and is automatically connected with a wireless router, and therefore data transmission is achieved, and the ZigBee wireless ad hoc network is a function of the second microprocessor CC 2530;

the second microprocessor unit is characterized in that CC2530 and CC2630 of the TI are also ZigBee network microprocessors, but the CC2630 is defined as a ZigBee network terminal under ultra-low power consumption by the TI, the CC2630 has a function of adding in the ZigBee network but does not have a function of establishing the ZigBee network, so that a CC2530 chip is adopted by the second microprocessor unit, the CC2530 establishes the ZigBee network through PanId and Channel, and a foot ring is added in the corresponding ZigBee network according to the two parameters to realize interaction of ZigBee messages; after the acquired foot ring data are subjected to artificial intelligence learning, the state of the chicken is analyzed, and the corresponding LED lamp is lightened to visually display the state of the chicken; meanwhile, the state information of the chicken is sent to a front-end processor master station through a Wifi module and a Wifi network, and alarm information is sent out to inform an administrator to conduct troubleshooting;

in this embodiment, 8 LEDs of the detection terminal represent the activity state of the chicken;

the state of the LED lamp represents the moving state of the chicken as shown in the table 1, and managers can visually see the state of the chicken through the LED lamp in front of the coop on site to judge the chicken died of illness and can remotely find the chicken died of illness through system software or APP;

TABLE 1 LED Lamp status indicator

A detection method of a dead chicken disease detection system based on artificial intelligence comprises the following steps:

in the expressions (1) and (2), a [ n ] is the acceleration reading at the nth time, and v [ n ] and s [ n ] are the velocity and displacement values at the corresponding times. The unidirectional movement displacement of the target object at the moment n can be obtained according to the formula (1) and the formula (2), the formula (3) is calculated,

calculating the three-dimensional total variance

Calculating the variance of unidirectional displacement of the time period T

By sampling acceleration values, calculating the displacement from the time period t1 to t2, if the acquisition interval is set as delta t =100ms, and the acceleration sampling time is 100ms and is sampled once, then the total sampling times in the time from t1 to t2 is n = (t2-t 1)/(delta t), the real-time motion displacement is s [ n ], and the calculation formula is formula (4)

The calculation has M time periods (T2-T1) in the T time, see formula (5)

M=T/(t2-t1) (5)

The average displacement in T time is calculated, see formula (6)

Calculating the displacement variance of the time period T

Since the variance is the square of the data and is too large different from the detection value, people are difficult to measure intuitively, the common variance is converted into the root number, namely the standard deviation which we say, and the variance describes the deviation degree and fluctuation, see formula (7)

Calculating the variance of displacement in three directions of x, y and z axes of the time period T

The variance of displacement in three directions, see formula (8-10):

calculating the three-dimensional displacement total variance of the time period T

Three-dimensional total variance, see formula (11):

in this embodiment, a SensorManager type is used to test the triaxial acceleration, and a triaxial acceleration data curve can be tested in real time in a static state, and due to noise interference, the curve may have some bubble fluctuation, and the data curve is shown in fig. 4;

the three-axis acceleration data curve of x, y and z is adjusted as the z axis is vertical, the acceleration of the z axis is about 9.8m/s2 and the acceleration of the x axis and the y axis is about 0.0m/s2 under the static state, and is shown in FIG. 5;

in the horizontal direction, when the mobile phone moves slightly, even a few centimeters, the acceleration of the x axis or the acceleration of the y axis fluctuates, and the fluctuation curve of the acceleration of the x axis in the three-axis acceleration is shown in fig. 6;

the method for calculating the maximum displacement and the three-dimensional total variance mainly comprises the following steps:

1) setting acceleration sampling frequency, setting delta t =100ms, sampling acceleration once, and correcting the acceleration due to different sensitivities, wherein the corrected triaxial accelerations Gx, Gy and Gz are almost 0;

2) the displacement S was calculated by setting the displacement time T, setting T =1 minute, calculating the displacement once in 60 seconds, n =60/0.1=600 times, and calculating the minute displacement by equation (4), which is Sx, Sy, Sz, respectively. And calculating the total displacement S = Sx + Sy + Sz;

3) and calculating the variance of the accumulated triaxial displacement in M minutes. Calculating the average displacement in M minutes by using a formula (6), and calculating the three-axis displacement variances Va, Vb and Vc by using formulas (8), (9) and (10);

4) the cumulative M minute solid displacement variance was calculated. Calculating a stereo total variance Vtotal through a formula (11);

5) calculating the maximum displacement Smax of accumulated M minutes;

the three-dimensional total variance Vtotal and the maximum displacement Smax are used as the input of a machine learning classification algorithm; as shown in fig. 7, the process data included on the interface is: triaxial real-time acceleration, triaxial real-time displacement, cumulative minute, triaxial minute displacement, final calculation data is: maximum minute displacement, total variance of three-dimensional total displacement;

the artificial intelligence learning is carried out by a supervised learning classification method, and the training and learning steps are shown in fig. 10. Firstly, acquiring data of a foot ring through an acceleration sensor; then, displacement calculation is carried out to obtain the maximum displacement and the three-dimensional total variance as (x1, x 2); training data through a machine learning classification algorithm to obtain an optimal training model, and identifying the activity behavior of the chicken through the training model;

in the embodiment, the data of the gateway is acquired by the foot ring sensor to judge the activity state of the chicken through machine learning, and as the activity time of the chicken is approximately 5:00 in the morning to 8:00 in the evening, the data is acquired from 5 points in the morning for 15 hours, and finally, the maximum displacement Smax and the total three-dimensional variance Vtotal are acquired;

sample set definition:

sample data: x = (X1, X2), X1 is the maximum displacement Smax, X2 is the stereo total variance Vtotal;

classification j: 0-dead chicken, 1-sick chicken, 2-weak moving chicken and 3-strong moving chicken;

example of a typical sample set:

a typical sample set of four classes is shown in table 2.

Typical sample set of Tab2

Training data: in this embodiment, 2000 sample data are collected for training test;

a machine learning algorithm is used to learn the algorithm,

the recognition effect of the SVM, KNN, DT, NB and BP for abnormal behavior recognition is realized through a 5-classification algorithm,

the SVM is a new learning method provided by cortex and Vapnik according to a statistical learning theory, and has the biggest characteristic that an optimal classification hyperplane is constructed at a maximized classification interval according to a structural risk minimization criterion to improve the generalization capability of learning, so that the problems of nonlinearity, high dimension, local minimum point and the like are well solved;

KNN is an example-based classification method, i training samples closest to an unknown sample x are found out, and x is classified into a class to which most of the i samples belong;

DT is an example-based inductive learning algorithm that focuses on inferring classification rules, denoted by DT, from a set of unordered, irregular examples. The purpose of constructing the DT is to find out the relation between the characteristic value and the category, and the DT is used for predicting the category of the unknown sample;

NB is a simple probability classification based on Bayes' theorem of independent hypothesis, and the principle is that the prior probability of a certain object is calculated by using a Bayes formula, namely the probability that the object belongs to a certain class, and the class with the maximum posterior probability is selected as the class to which the object belongs;

the BP neural network is a multilayer feedforward network trained according to error back propagation, the algorithm is called as BP algorithm, the basic idea is a gradient descent method, a gradient search technology is utilized, so that the mean square error of the actual output value and the expected output value of the network is expected to be minimum, and the BP neural network has the outstanding advantages of strong nonlinear mapping capability and flexible network structure;

the invention identifies the accuracy rate experiment result and compares and analyzes:

in this embodiment, sample data of 4 ten thousand chickens are used for testing, and 5 machine learning methods are used for comparison, and as shown in table 3, the identification results of the comparison are analyzed as follows:

1) because dead chickens are inactive, the maximum displacement Smax and the three-dimensional total variance Vtotal are almost 0, the recognition accuracy of 5 kinds of chickens is 100%, the recognition accuracy of the chickens with strong activity almost reaches 100%, the important point is that sick chickens and the chickens with weak activity are difficult to distinguish, the sick chickens and the chickens with weak activity are difficult to distinguish because the activity of some chickens is weak and the behavior of the sick chickens is similar to that of the sick chickens, but the sick chickens and the chickens with weak activity eat the feed, the maximum displacement is large, the sick chickens and the weak activity can be distinguished, and only the recognition accuracy is low;

2) by operating the 5 classification algorithms, the BP algorithm has the highest recognition rate, so that BP is selected as the final classification algorithm;

3) if sick chickens and weak moving chickens are classified into one class, the recognition rate is higher, only once more detection is carried out in actual application, but the detection of the sick chickens cannot be omitted;

and comparing identification accuracy results of Tab.35 algorithms:

the invention uses the comparative analysis of the operating cost of the farm:

assuming a chicken-raising greenhouse with 4 ten thousand chickens, 4 workers are needed for artificial breeding, 2 workers are needed for system breeding, the salary cost of one worker per year is 10 ten thousand yuan, and the material cost of 1 foot ring is 10 yuan, then the total cost of manpower and the total cost of the system are compared and analyzed, and the comparison result is shown in table 4, so that the cost is equivalent to that of the manpower when the system is operated for 2 years, and the cost can be saved after 2 years;

tab.4 Total cost of operation comparison

From the results in table 3, it can be seen that the total cost of operating the system decreases with time, the histogram comparison of the cost of operating for 4 years is shown in fig. 11, the total cost trend of operating for 8 years can also be analyzed and calculated according to the results in table 4, and compared with the manual operation cost, the cost trend graph is shown in fig. 12, and it can be seen from the cost trend comparison graph in fig. 12 that: the cost can be recovered after the system is operated for 2 years, the cost can be reduced by 25% after the system is operated for 4 years, the economic benefit can be generated by 100 ten thousand after the system is operated for 8 years, and the greater economic benefit can be generated if the system is operated in a larger scale chicken farm, so that the operation cost of the system is compared with that of manpower, the identification accuracy is improved, and the higher economic value is created;

the invention measures the three-dimensional displacement of the chicken by buckling a foot ring on each chicken, designs and calculates the three-dimensional total variance to express the activity intensity of the chicken, finally identifies the state of the chicken by a machine learning classification method to realize the correct judgment of dead chicken and sick chicken, the foot ring and the detection terminal are powered by a zigbee ad hoc network and a low-power battery, the activity state of the chicken is expressed by two characteristics of maximum displacement and the three-dimensional total variance, and experiments prove that the identification accuracy reaches 95.6 percent, thereby achieving the practical effect, reducing the labor and improving the working efficiency.

Finally, it should be noted that: although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that modifications may be made to the embodiments or portions thereof without departing from the spirit and scope of the invention.

Claims

1. The utility model provides a dead chicken disease chicken detecting system based on artificial intelligence which characterized in that: the method comprises the following steps:

the front-end processor is used for collecting data of the coop detection unit;

wherein, chicken coop detecting element includes:

2. The system of claim 1, wherein the system comprises: the foot ring includes:

the sensor unit is used for acquiring chicken movement acceleration data;

the battery unit is used for supplying power to the foot ring;

3. The system of claim 2, wherein the system comprises: the probe terminal includes:

the storage unit is used for storing the acquired acceleration data;

4. The system of claim 1, wherein the system comprises: the detection system further comprises a remote detection system connected with the data server and used for remote detection, and the remote detection system comprises a WEB monitoring client and a remote detection APP.

5. The detection method of the dead chicken detection system based on artificial intelligence as claimed in any one of claims 1-4, wherein: the method comprises the following steps:

6. The detection method of the dead chicken detection system based on artificial intelligence, which is characterized by comprising the following steps of: the step of calculating the maximum displacement in step S1 includes:

(1)

(2)

in the formulas (1) and (2), a [ n ] is an acceleration reading value at the nth moment, v [ n ] and s [ n ] are speed and displacement values at corresponding moments, the acquisition interval is set to be delta t, the unidirectional movement displacement of the chicken at the n moment can be obtained according to the formulas (1) and (2), and the calculation formula is shown as (3):

(3)。

7. the detection method of the dead chicken detection system based on artificial intelligence, which is characterized by comprising the following steps of: the step of calculating the three-dimensional total variance in step S1 includes:

(4)

M=T/(t2-t1) (5)

S13, calculating the average displacement in T time, see formula (6)

(6)

S14, calculating the variance of the displacement in T time, wherein the variance describes the deviation degree and fluctuation, see formula (7)

(7)

S15, calculating the variance of displacement in three directions of x, y and z axes in T time, see formula (8-10)

(8)

(9)

(10)

S16, calculating the three-dimensional displacement total variance in the T time, wherein the three-dimensional displacement total variance is shown in formula (11):

(11)。

8. the detection method of the dead chicken detection system based on artificial intelligence, which is characterized by comprising the following steps of: the training step of training the model in step S2 includes:

s22, setting displacement time T to calculate displacement S, calculating minute displacements Sx, Sy and Sz respectively according to a formula (4), and calculating total displacement S = Sx + Sy + Sz;

9. The detection method of the dead chicken detection system based on artificial intelligence, according to claim 8, characterized in that: the training step of training the model in step S2 includes:

10. The detection method of the dead chicken detection system based on artificial intelligence, according to claim 8, characterized in that: the machine learning classification algorithm comprises: one of SVM, KNN, DT, NB or BP.