A smart pet status perception method based on Internet of Things technology

Abstract

A smart pet status perception method based on Internet of Things technology includes: using multiple types of sensors to collect different data for different pets when they are in the device state, the device state, and the device state; using a linear classification function to classify the data, using binary classification for the pet's position and multi-classification for the pet's state; optimizing the linear function using a perceptron algorithm to find the optimal weight vector; using the two resulting linear classification functions to identify the relative position of the pet and the device, as well as the pet's own physical state; and after the pet leaves the device, feeding back the analysis results of the pet to the pet owner via the internet. Based on Internet of Things technology, the present invention achieves a one-to-one binding between the device and the user account, uses multiple sensors to collect pet status information, and pushes it to the corresponding pet owner, allowing pet owners to better understand their pet's living and physical conditions.

Description

Technical Field

[0001] The present invention relates to sensor technology and perceptron algorithms, and in particular to an intelligent pet perception method based on Internet of Things technology. Background Art

[0002] In recent years, an increasing number of young people have left their hometowns to pursue careers in large cities. The long-term loneliness has led them to yearn for friends, family, and a listening ear. Driven by this spiritual need, many have started keeping pets. Since pet owners often work during the day and are unable to care for their pets, they opt for foster care or smart pet products. However, existing smart pet devices often lack information access. Pet owners often lack access to information about their pets, or the information they receive is limited and untimely. This leads them to worry about their pets' health, their well-being, and whether they might wander off or get hurt while they're away. To meet this high-level demand for "spiritual care" for their pets, real-time feedback on their pets' living conditions and health is essential.

[0003] Internet of Things technology refers to the interconnection of ubiquitous terminal information sensing devices and facilities through various wireless and wired long-distance or short-distance communication networks according to agreed protocols, and then the exchange and communication of information to achieve intelligent identification, positioning, tracking, supervision and other functions. It has been widely used in recent years and has become the commanding heights of the new round of economic and technological development in the world.

[0004] The perceptron is a typical structure in artificial neural networks. It is used under the requirement that samples be linearly separable. It is a linear classification method that can directly obtain a linear discriminant function. The core mechanism is "reward and penalty": when the classification is correct, the correct weight vector is maintained; when the classification is incorrect, the weight vector is shifted in the direction of reward.

[0005] Intelligent sensing technology based on the Internet of Things, with its broad information perception, reliable information transmission and effective information processing, makes information interaction between objects and between objects and people more convenient. It not only has the ability to obtain external information through various sensors, but also can achieve the purpose of recognizing the environment and object categories and attributes through processes such as memory, learning, judgment, and reasoning, and then provide behavioral decision-making recommendations. Summary of the Invention

[0006] In order to overcome the problem of pet owners' insufficient understanding of their pets in the existing technology, the present invention proposes an intelligent pet perception method based on Internet of Things technology, which is compatible with the existing smart cat litter boxes on the market. Its purpose is to record the daily excretion behavior of pets and analyze the physical condition of pets.

[0007] In order to achieve the above object, the technical solution provided by the present invention is:

[0008] An intelligent pet sensing method based on Internet of Things technology includes the following steps:

[0009] Step 1: Use multiple types of sensors to collect different data when the pet is in the state of entering the device, staying inside the device, and leaving the device;

[0010] Step 2: Use linear classification function to classify the data, where pet location adopts binary classification and pet status adopts multi-classification. Use perceptron algorithm to optimize the linear function and find the optimal weight vector.

[0011] Step 3: Use the two linear classification functions obtained to identify the relative position of the pet and the device, as well as the physical condition of the pet itself. First, the position classification function determines whether a pet has entered the device. Then, the state classification function determines which pet has entered and analyzes the physical condition of the pet. Finally, the position classification function determines whether the pet has left the device.

[0012] Step 4: After the pet leaves the device, the analysis results of the pet are fed back to the pet owner via the Internet.

[0013] Furthermore, in step one, all sensors are deployed inside the device and include the following types: infrared tube sensor, infrared temperature sensor, weighing sensor, heart rate and breathing sensor, camera sensor, and the collected data include: whether the pet is inside the device, the pet's body temperature, the weight of the pet after entering the device, the weight of the pet after leaving the device, the pet's heart rate and breathing rate, and the pet's facial features.

[0014] Furthermore, in the step 2, a linear classification function is used to classify the data, and a binary classification is performed on whether the pet is inside the device based on the infrared tube data and the weighing data, and the classification categories are presence and absence; the pet's identity and physical condition are multi-classified based on the pet's body surface temperature data, the weight change data before and after the pet enters and leaves, the pet's heartbeat and breathing rate, and the pet's facial feature recognition data, and the classification categories are health, fever, diarrhea, constipation, and respiratory disease.

[0015] Furthermore, the optimal weight vector of the linear classification function in step 2 is obtained by the perception algorithm. Assume that the sample set is ((x1, y1); (x2, y2); ...; (x n ,y n )), in the case of binary classification, y i ∈{-1,+1}, then the correctly classified samples satisfy y i ω T x i >0, the weight vector is:

[0016]

[0017] Where γ is the correction vector, which corrects the error when the positive examples are mistakenly classified as negative examples or the negative examples are mistakenly classified as positive examples;

[0018] In the case of multi-classification, there are M discriminant functions {d i ,i=1,2,...,M}, if x k ∈ω i , then d i >d j , The weight vector under the perceptron algorithm is:

[0019]

[0020] Where C is a positive constant, the initial value of the weight vector ω i (l), i=1, 2, ...., M can be selected arbitrarily according to the situation.

[0021] Furthermore, in step three, during final deployment, to reduce energy consumption, the sensors do not continuously collect data, and the data collection occurs in a temporal order determined by the operational logic of the entire method. Since determining whether a pet has entered or exited the device is crucial to the present invention, the infrared pair sensor, among all sensors, collects data every 1 second. Furthermore, to eliminate malfunctions of the infrared pair sensor itself and the special case of a pet lingering at the device entrance, multiple infrared pair sensors are deployed at the front, middle, and rear of the device. After the front infrared pair sensor detects an object, the weighing sensor begins operating and continuously collects data. When it is determined that a pet has entered the device from the outside, the remaining sensors are turned on and data is collected in the order of camera acquisition, infrared temperature measurement, and heart rate and respiration acquisition. The camera collects multiple pet images until the image is recognized as a pet's face. The infrared temperature measurement and heart rate and respiration data are collected 10 times, and the average is calculated after error removal. When it is determined that the pet has left the device, the weighing sensor collects the final five data and calculates the average value. This value is then subtracted from the pet's weight after the last exit and recorded.

[0022] Furthermore, in step 4, the analysis results of the pet are fed back to the pet owner via the Internet. It is necessary to first bind the device to the terminal account one-to-one, and the device must be in an environment with wireless network coverage. The final feedback information includes the pet's identity information collected by the sensor, the pet's surface temperature, weight, heart rate, respiratory rate and other original data and the pet's physical condition obtained after classification by the linear classification function.

[0023] The beneficial effects of the present invention are mainly manifested in:

[0024] 1. Collect pet status information through various sensors, classify the data using linear classification functions, and analyze the pet's physical condition to help pet owners better understand their pets and satisfy their spiritual feelings of caring for their pets at a higher level.

[0025] 2. Reduce the operation of some sensors when they are not necessary, reduce power consumption, and be more environmentally friendly

[0026] 3. The operation of the present invention is strictly in accordance with the results collected by the sensor. The overall logic is self-consistent and has multiple safety measures, which greatly guarantees the safety of pets using the device.

[0027] The present invention is based on Internet of Things technology, realizes one-to-one binding between devices and user accounts, uses multiple sensors to collect pet status information, and pushes it to the corresponding pet owners, so that pet owners can better understand the living status and physical condition of their pets. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is the overall framework diagram of the present invention.

[0029] Figure 2 、 3 is a flow chart of a two-classification solution for pet location status according to an embodiment of the present invention, wherein Figure 2 It is used to identify whether a pet has entered the device. Figure 3 It is a flowchart for identifying whether a pet has left the device.

[0030] Figure 4 2 is a diagram of a multi-classification scheme for a pet's physical state according to an embodiment of the present invention. DETAILED DESCRIPTION

[0031] The invention will be described in detail below with reference to specific embodiments and accompanying drawings. The following examples will help those skilled in the art further understand the present invention, but are not intended to limit the present invention in any way. It should be noted that those skilled in the art may make various modifications and improvements without departing from the scope of the present invention. These modifications and improvements are all within the scope of protection of the present invention.

[0032] Reference Figure 1 , a smart pet state perception method based on Internet of Things technology, comprising the following steps:

[0033] Step 1: Use multiple sensors to collect data on different pets when they enter, remain inside, and exit the device. These sensors are all deployed inside the device and include the following types: infrared tube sensors, infrared temperature sensors, weighing sensors, heart rate and respiration sensors, and camera sensors. The collected data includes: pet location data, surface temperature data, weight change data before and after the pet enters and exits the device, heart rate and respiration data, and facial features.

[0034] Step 2: Use a linear classification function to classify the data. Based on the infrared tube data and weighing data, a binary classification is performed to determine whether the pet is inside the device, with the classification categories being present or absent. Based on the pet's surface temperature data, weight change before and after the pet enters and leaves, the pet's heart rate and respiratory rate, and facial recognition data, a multi-class classification is performed to determine the pet's identity and physical condition, with the classification categories being healthy, fever, diarrhea, constipation, and respiratory disease. The perceptron algorithm is used to optimize the linear function to find the optimal weight vector.

[0035] Step 3. Use the two linear classification functions finally obtained to identify the relative position status of the pet and the device, as well as the physical condition of the pet itself. First, the position classification function determines whether a pet has entered the device. Then, the state classification function determines which pet has entered and analyzes the physical condition of the pet. Finally, the position classification function determines whether the pet has left the device.

[0036] Step 4: After the pet leaves the device, the analysis results of the pet are fed back to the pet owner via the Internet. Since the device of the present invention adopts a WIFI network card solution, the device needs to be in an environment covered by a wireless network. At the same time, the device needs to be bound to the pet owner's terminal account one-to-one. The final feedback information includes the pet's identity information collected by the sensor, the pet's body surface temperature, heart rate, respiratory rate and other raw data and the pet's physical condition obtained after classification by the linear classification function.

[0037] Reference Figure 2 、 3 ,The two classification methods for pet location status are as follows:

[0038] First, it is assumed that the pet does not exist in the device initially. The infrared sensor inside the device collects data every 1 second. Figure 2, when the front infrared sensor collects data, the weighing sensor continuously collects data, records the changes in weight data, and combines it with the data from the infrared pair tube sensors to form a position judgment data set. Under normal circumstances, when an object enters the device from the outside, the data collected by the three pairs of infrared tubes should start from {1, 0, 0}, and then change to {1, 1, 0} and {0, 1, 1}. There are changes in the time sequence. Therefore, it is necessary to continuously record data at multiple moments. At the same time, in order to eliminate the impact of sensor failures, there is a tolerance for collection errors in some data. The final data set for judging the presence of a pet in the device is {1, x, 0, m, 0, x, x, m, 0, x, 1, m}, where x means 0 or 1, and m is the changing mass data. In addition, if m obviously does not match the normal weight of a cat, it is determined that the object entering is not a pet, and the pet's position status is still not present in the device. If Figure 3 Under normal circumstances, when a pet leaves the device, the data collected by the three pairs of infrared tubes should change from {1, 0, 0} to {0, 0, 0}. Excluding sensor fault interference, the judgment data set is {1, x, 0, m, x, 0, 0, m, 0, 0, 0, m}. The above two data sets are used as positive examples x for whether the pet enters the device and whether the pet leaves the device. The data sets of three consecutive time points of continuous collection of infrared tubes and weighing are used. In the case of binary classification, the weight vector is:

[0039]

[0040] Where γ is the correction vector, which corrects the error when positive examples are mistakenly classified as negative examples or negative examples are mistakenly classified as positive examples.

[0041] To determine the physical condition of a pet, it is necessary to first identify the pet's identity, that is, which pet's data is being collected. This step requires the pet owner to first upload a frontal photo of the pet on the terminal account. After the pet enters the device, the camera sensor will collect image data and match and identify which pet has entered the device based on the collected pet's facial image. Subsequent collected data will be classified under the pet's identity.

[0042] Common pet ailments include fever, diarrhea, constipation, and respiratory illnesses. Fever typically manifests as rapid breathing, elevated body temperature, a rapid heartbeat, and lethargy. Diarrhea typically manifests as elevated body temperature, significant weight loss, and frequent toileting. Constipation manifests as a loss of energy, decreased bowel movements, and frequent inability to defecate while squatting. Respiratory illnesses manifest as rapid breathing, sneezing, and a runny nose. Infrared temperature sensors utilize the blackbody radiation principle to measure surface temperature contactlessly, making them suitable for collecting pet temperatures. Heart rate and respiration sensors can detect heart and respiratory rates, while millimeter-wave radar heart and respiration sensors can collect data remotely and contactlessly. Weighing sensors can sensitively detect changes in mass, providing information on a pet's weight and bowel movements. Furthermore, image data captured by cameras can also reflect a pet's mental and physical condition to a certain extent.

[0043] Reference Figure 4 ,The multi-classification method for the pet’s physical state is as follows:

[0044] Taking cats as an example, the normal body temperature of cats ranges from 37.5℃ to 39℃, 39.2℃ to 39.5℃ is a mild fever, and over 40℃ is a high fever. The normal breathing rate in a resting state is 30-45 times / minute, the heart rate is 120-140 times / minute, and the excretion is normal once or twice a day, and occasionally once every two days is also normal. After collecting various types of data to form a data set, those with a significantly high respiratory rate but normal body temperature are classified as respiratory diseases, those with a high respiratory rate and heart rate and a high body temperature are classified as having a fever, those who frequently enter the device but do not defecate for a long time are classified as constipated, those who frequently enter the device and have defecation behavior, and at the same time have changes in body temperature or a significant decrease in weight are classified as having diarrhea, and the rest are classified as healthy. There are 5 discrimination modes in total, so the weight vector under the perceptron algorithm is:

[0045]

[0046] The contents described in the embodiments of this specification are merely an enumeration of the implementation forms of the inventive concept. The scope of protection of the present invention should not be regarded as limited to the specific forms described in the embodiments. The scope of protection of the present invention also extends to equivalent technical means that can be conceived by those skilled in the art based on the inventive concept.

Claims

1. A method for sensing the state of an intelligent pet based on Internet of Things technology, characterized in that: The following steps are involved: Step 1: Use multiple types of sensors to collect different data when the pet is in the state of entering the device, staying inside the device, and leaving the device; Step 2: Use linear classification function to classify the data, where pet location adopts binary classification and pet status adopts multi-classification. Use perceptron algorithm to optimize the linear function and find the optimal weight vector. Step 3: Use the two linear classification functions obtained to identify the relative position of the pet and the device, as well as the pet's status. First, the position classification function determines whether a pet has entered the device. Then, the status classification function determines which pet has entered and analyzes the pet's physical condition. Finally, the position classification function determines whether the pet has left the device. Step 4: After the pet leaves the device, the analysis results of the pet are fed back to the pet owner via the Internet; In step 1, all sensors are deployed inside the device and include the following types: infrared tube sensor, infrared temperature sensor, weighing sensor, heart rate and respiration sensor, and camera sensor. The collected data includes: whether the pet is inside the device, the pet's body temperature, the pet's weight after entering the device, the pet's weight after leaving the device, the pet's heart rate and respiration rate, and the pet's facial features; The data is classified using a linear classification function. Based on the infrared tube data and weighing data, a binary classification is performed on whether the pet is inside the device, with the classification categories being present or absent. The pet's identity is identified based on the pet's facial feature recognition data. The pet's physical condition is multi-classified based on the pet's surface temperature data, weight change data before and after the pet enters and leaves, the pet's heart rate and respiratory rate, and the pet's facial feature recognition data, with the classification categories being healthy, fever, diarrhea, constipation, and respiratory disease. The optimal weight vector of the linear classification function is obtained by the perception algorithm. Assume that the sample set is ((x1,y1); (x2,y2); ...; (x n ,y n )), in the case of binary classification, y i ∈{-1,+1}, then the correctly classified samples satisfy y i ω T x i >0, the weight vector is: Where γ is the correction vector, which corrects the error when the positive examples are mistakenly classified as negative examples or the negative examples are mistakenly classified as positive examples; In the case of multi-classification, there are M discriminant functions {d i ,i=1,2,...,M}, if x k ∈ω i ,but The weight vector under the perceptron algorithm is: Where C is a positive constant, the initial value of the weight vector ω i (l), i=1, 2, ...., M can be selected arbitrarily according to the situation; In step three, when the system is finally deployed, to reduce energy consumption, the sensors do not continuously collect data. The data is collected in a sequential order determined by the system's operational logic. Because determining whether a pet has entered or exited the device is crucial to the system, the infrared pair sensors collect data every 1 second. Furthermore, to eliminate malfunctions of the infrared pair sensors and the possibility of a pet lingering at the device entrance, multiple infrared pair sensors are deployed at the front, middle, and rear of the device. When the front infrared pair sensor detects an object, the weighing sensor begins continuous data collection. If a pet enters the device from outside, the remaining sensors are activated, and data is collected in the order of camera capture, infrared temperature measurement, and heart rate and respiration. The camera captures multiple pet images until the image is recognized as a pet's face. The infrared temperature and heart rate and respiration measurements collect data 10 times, and the average is calculated after eliminating errors. If the pet is determined to have left the device, the weighing sensor collects the final five data points and averages them. This average is then subtracted from the pet's weight after the pet's last exit and recorded. In step 4, the pet's analysis results are fed back to the pet owner via the Internet. It is necessary to first bind the device to the terminal account one-to-one, and the device must be in an environment with wireless network coverage. The final feedback information includes the pet's identity information collected by the sensor, the original data of the pet's surface temperature, weight, heart rate, and respiratory rate, and the pet's physical condition obtained after classification by the linear classification function.

Images