Intelligent pet injection system and method based on image recognition

Abstract

The invention belongs to the field of pet medical care, and provides an intelligent pet injection system and method based on image recognition, and the system comprises a net-shaped fixing device, an AI control module, an image recognition camera, a mobile injection device, a power module and a user interaction interface. By arranging the net-shaped fixing device, a pet is effectively and comfortably fixed in the injection process, and accidental injuries caused by struggling or escaping are avoided; the injection process is controlled through the AI control module, manual operation errors are reduced, and risks in the injection process are further reduced; through the combination of the AI control module, the image recognition camera and the mobile injection device, manual intervention is reduced, so that single injection operation is quicker, and meanwhile, the workload of pet medical personnel is relieved.

Description

Technical Field

[0001] The present invention relates to the field of pet medical care, and in particular to a pet intelligent injection system and method based on image recognition. Background Art

[0002] The traditional pet injection process relies on manual operation, which is not only inefficient, but also prone to inaccurate injection or pet injury due to pet struggles or human operation errors. In order to solve these problems, the pet intelligent injection system based on image recognition came into being. The rapid development of artificial intelligence technology, especially the application of deep learning algorithms, enables computers to understand and process image data more accurately. This provides strong technical support for the image recognition function of the pet injection system. With the improvement of people's living standards and the popularization of pet breeding, the demand for the pet medical industry is growing. The traditional pet injection method can no longer meet the efficiency, safety and comfort requirements of modern pet medical care.

[0003] Although the pet intelligent injection system based on image recognition has made significant progress, the existing solutions still have some technical problems: insufficient image recognition accuracy and great influence by environmental factors. Lack of real-time pet status perception and behavior analysis. Existing systems often cannot analyze dynamic factors in real time, nor can they automatically adjust injection strategies according to the pet's behavior. Existing systems are mostly stand-alone systems or local network devices, usually lacking efficient cloud platforms and communication modules, resulting in the inability to achieve real-time data transmission and processing across regions and devices. Summary of the invention

[0004] In order to overcome the deficiencies of the prior art, the object of the present invention is to provide a pet intelligent injection system and method based on image recognition to achieve a safe, accurate and low-stress vaccination or drug treatment process for pets.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A pet intelligent injection system based on image recognition, comprising: a mesh fixing device, an AI control module, an image recognition camera, a mobile injection device, a power module and a user interaction interface;

[0007] The mesh fixing device, the image recognition camera, and the mobile injection device are respectively connected to the AI ​​control module; the image recognition camera is arranged on an adjustable bracket above the inside of the mesh fixing device; the power module is respectively connected to the mesh fixing device, the AI ​​control module, the image recognition camera, the mobile injection device, and the user interaction interface; the user interaction interface is connected to the AI ​​control module;

[0008] The mesh fixing device is used to automatically adjust the size and shape according to the pet's body shape through intelligent sensing technology; the image recognition camera is used to collect images of the pet; the AI ​​control module is used to identify the target injection site of the pet based on the image data collected by the image recognition camera through the CNN image recognition algorithm, calculate the injection position based on the recognition result of the CNN image recognition algorithm through the injection control algorithm, and control the mobile injection device to perform injection according to the calculation result of the injection control algorithm.

[0009] Preferably, the mesh fixing device adopts a frame structure; the frame structure is made of a lightweight and high-strength metal material; the frame structure is covered with a mesh fabric; the mesh fabric includes: nylon material.

[0010] Preferably, the AI ​​control module includes:

[0011] A processor, used for adjusting the injection speed and injection dosage of the mobile injection device according to the injection pressure and flow rate data fed back by the preset sensor;

[0012] A memory for storing the CNN image recognition algorithm, the injection control algorithm and pet injection operation data; the pet injection operation data includes: pet body shape data, image recognition results, injection position and dosage information, pet real-time status data and feedback data of all sensors during the injection process; the feedback data includes: pressure data, flow data, temperature data, humidity data, electrical data, acceleration data and physiological data;

[0013] An algorithm unit is used to run the CNN image recognition algorithm and the injection control algorithm.

[0014] Preferably, the mobile injection device comprises: a mechanical arm, an injection drive mechanism and an injection syringe connected in sequence;

[0015] The robotic arm is used to perform multi-degree-of-freedom movement according to the target position and posture instructions, path planning and trajectory instructions, movement speed and acceleration instructions, feedback adjustment instructions and injection process control instructions issued by the AI ​​control module; the injection drive mechanism is used to control the injection dose and injection speed of the injection syringe.

[0016] Preferably, it also includes: a wireless communication module;

[0017] The wireless communication module is connected to the AI ​​control module;

[0018] The wireless communication module is used to upload feedback data during the injection process to a remote server or a mobile terminal of a pet medical staff, and send an alarm message to the remote server when a preset sensor detects an abnormal situation; the abnormal situation includes: injection failure, severe stress reaction of the pet; the severe stress reaction includes: aggressive behavior, fainting and shock.

[0019] Preferably, it also includes: an emergency processing module; the emergency processing module includes: an emergency processing program and an uninterruptible power supply;

[0020] The emergency handling program is used to detect emergencies. When the emergency is a power outage, the uninterruptible power supply is used to provide power to the power module. When the emergency is a severe allergic reaction of a pet, the mobile injection device is stopped, an emergency alarm is triggered, and an emergency handling guide is presented to the user. The severe allergic reaction includes: shortness of breath, wheezing, coughing, and mental depression.

[0021] Preferably, it also includes: a self-checking program;

[0022] The self-test program is used to check the working status of all hardware components when the AI ​​control module is started; the hardware components include the image recognition camera, the robotic arm, the sensor and the power module.

[0023] Preferably, a pet intelligent injection method based on image recognition comprises:

[0024] Adaptively fix the pet using the mesh fixing device;

[0025] Using the image recognition camera to collect the image data;

[0026] Using the AI ​​control module to identify the target injection site and calculate the injection position of the image data;

[0027] injecting the pet using the mobile injection device according to the calculation result of the injection control algorithm;

[0028] The workflow of the CNN image recognition algorithm includes:

[0029] Extracting local features of the image data using the Canny algorithm and convolution of the CNN network; the local features include ears, eyes, and legs;

[0030] Analyzing the local features using a classifier and a regressor to obtain position information and angle information of the target injection site;

[0031] Determine the behavior pattern of the pet using a behavior analysis algorithm based on the collected image and video data, sensor data, and sound data; the behavior analysis algorithm includes: any one of a recurrent neural network and a long short-term memory network;

[0032] According to the collected static images and multi-angle images, the pre-trained deep convolutional neural network is used to classify the types and postures of pets to obtain classification results;

[0033] The injection strategy is dynamically adjusted using a fully connected layer according to the position information, the angle information, the behavior pattern and the classification result.

[0034] The present invention discloses the following technical effects:

[0035] The present invention provides a pet intelligent injection system and method based on image recognition. By introducing a behavior analysis algorithm and a deep convolutional neural network, the defect that the prior art cannot automatically adjust the injection strategy according to the pet's behavior is solved, and real-time monitoring of the pet's dynamic behavior is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0037] Figure 1 A schematic diagram of the structure of a pet intelligent injection system based on image recognition provided by an embodiment of the present invention;

[0038] Figure 2 A schematic diagram of a pet intelligent injection system based on image recognition provided by an embodiment of the present invention.

[0039] Description of reference numerals:

[0040] 1-AI control module, 2-mesh fixing device, 3-image recognition camera, 4-mobile injection device, 5-power module, 6-user interaction interface. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] The purpose of the present invention is to provide a pet intelligent injection system and method based on image recognition to achieve a safe, accurate and low-stress vaccination or drug treatment process for pets.

[0043] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0044] Figure 1 A schematic diagram of the structure of a pet intelligent injection system based on image recognition provided by an embodiment of the present invention is shown in FIG. Figure 1 As shown, the present invention provides a pet intelligent injection system based on image recognition, comprising: a mesh fixing device 2, an AI control module 1, an image recognition camera 3, a mobile injection device 4, a power module 5 and a user interaction interface 6;

[0045] The mesh fixing device 2, the image recognition camera 3, and the mobile injection device 4 are respectively connected to the AI ​​control module 1; the image recognition camera 3 is arranged on an adjustable bracket above the inside of the mesh fixing device 2; the power module 5 is respectively connected to the mesh fixing device 2, the AI ​​control module 1, the image recognition camera 3, the mobile injection device 4, and the user interaction interface 6; the user interaction interface 6 is connected to the AI ​​control module 1;

[0046] The mesh fixing device 2 is used to automatically adjust the size and shape according to the pet's body shape through intelligent sensing technology; the image recognition camera 3 is used to collect images of the pet; the AI ​​control module 1 is used to identify the target injection site of the pet based on the image data collected by the image recognition camera 3 through the CNN image recognition algorithm, calculate the injection position based on the recognition result of the CNN image recognition algorithm through the injection control algorithm, and control the mobile injection device 4 to perform injection according to the calculation result of the injection control algorithm.

[0047] Preferably, the mesh fixing device 2 adopts a frame structure; the frame structure is made of a lightweight and high-strength metal material; the frame structure is covered with a mesh fabric; the mesh fabric includes: nylon material.

[0048] Specifically, the AI ​​control module 1 includes:

[0049] A processor, used for adjusting the injection speed and injection dosage of the mobile injection device according to the injection pressure and flow rate data fed back by the preset sensor;

[0050] A memory for storing CNN image recognition algorithms, injection control algorithms and pet injection operation data; the pet injection operation data includes: pet body shape data, image recognition results, injection position and dosage information, pet real-time status data and feedback data of all sensors during the injection process; the feedback data includes: pressure data, flow data, temperature data, humidity data, electrical data, acceleration data and physiological data;

[0051] Algorithm unit, used to run CNN image recognition algorithm and injection control algorithm.

[0052] Further, the mobile injection device 4 comprises: a mechanical arm, an injection drive mechanism and an injection syringe connected in sequence;

[0053] The robotic arm is used to perform multi-degree-of-freedom movement according to the target position and posture instructions, path planning and trajectory instructions, movement speed and acceleration instructions, feedback adjustment instructions and injection process control instructions issued by the AI ​​control module 1; the injection drive mechanism is used to control the injection dose and injection speed of the injection syringe.

[0054] Preferably, it also includes: a wireless communication module;

[0055] The wireless communication module is connected to the AI ​​control module 1;

[0056] The wireless communication module is used to upload feedback data during the injection process to a remote server or a mobile terminal of pet medical personnel, and to send an alarm message to the remote server when a preset sensor detects an abnormal situation; abnormal situations include: injection failure, severe stress reaction of the pet; the severe stress reaction includes: aggressive behavior, fainting and shock.

[0057] Preferably, it also includes: an emergency processing module; the emergency processing module includes: an emergency processing program and an uninterruptible power supply;

[0058] The emergency handling program is used to detect emergencies. When the emergency is a power outage, an uninterruptible power supply is used to provide power to the power module 5. When the emergency is a severe allergic reaction of the pet, the mobile injection device 4 is stopped, an emergency alarm is triggered, and an emergency handling guide is displayed on the user interaction interface 6. The severe allergic reaction includes: shortness of breath, wheezing, coughing, and mental depression.

[0059] Also included: self-test procedures;

[0060] The self-check program is used to check the working status of all hardware components when the AI ​​control module 1 is started; the hardware components include the image recognition camera 3, the robotic arm, the sensor and the power module 5.

[0061] Furthermore, a pet intelligent injection method based on image recognition comprises:

[0062] Adaptively fix the pet using the mesh fixing device 2;

[0063] Using the image recognition camera 3 to collect image data;

[0064] Using the AI ​​control module 1 to identify the target injection site and calculate the injection position of the image data;

[0065] Using the mobile injection device 4 to inject the pet according to the calculation result of the injection control algorithm;

[0066] The workflow of the CNN image recognition algorithm includes:

[0067] The local features of the image data are extracted using the Canny algorithm and the convolution of the CNN network; the local features include ears, eyes, and legs;

[0068] The local features are analyzed using classifiers and regressors to obtain the location and angle information of the target injection site;

[0069] Determine the behavior pattern of the pet using a behavior analysis algorithm based on the collected image and video data, sensor data, and sound data; the behavior analysis algorithm includes: any one of a recurrent neural network and a long short-term memory network;

[0070] According to the collected static images and multi-angle images, the pre-trained deep convolutional neural network is used to classify the types and postures of pets to obtain classification results;

[0071] The fully connected layer is used to dynamically adjust the injection strategy based on the position information, angle information, behavior pattern and classification results.

[0072] Preferably, CNN (convolutional neural network) is the core foundation of the entire image recognition and analysis. It is mainly composed of convolutional layers, pooling layers and fully connected layers. The convolutional layer is responsible for automatically extracting local features in the image, such as the pet's body contour, hair texture, ears, eyes, legs and other part features. It performs feature extraction operations on local areas by sliding the convolution kernel on the image, and different convolution kernels can learn different feature patterns. The pooling layer is used to reduce the data dimension and reduce the amount of calculation while retaining key features, such as the commonly used maximum pooling or average pooling operations. The fully connected layer is an important part of the CNN network. In CNN, the feature maps extracted by the convolutional layer and the pooling layer are flattened into a one-dimensional vector and then input into the fully connected layer. The fully connected layer integrates and transforms these local features for the final prediction, such as determining the type of pet, injection site, etc. In terms of connection, the fully connected layer is usually connected after the convolutional layer and the pooling layer of the CNN network, and is the key link for the CNN network to perform classification or regression prediction. The CNN network itself can be used as a powerful feature extractor. The features it extracts (usually the output of the fully connected layer) will be used as input to the classifier and regressor. For pet image classification tasks, such as determining whether the pet in the image is a cat or a dog, the features extracted by the CNN network will be input into a classifier (such as a Softmax classifier), and the classifier will make classification decisions based on these features. For regression tasks such as pet weight prediction or injection site location prediction, the features extracted by the CNN network will be input into a regressor (such as a linear regressor), and the regressor will make numerical predictions based on the features. CNN networks can provide basic features for behavior analysis algorithms. Behavior analysis algorithms analyze and judge the behavior of the target object (pet) based on the features extracted by CNN, combined with specific rules or models. In terms of connection, behavior analysis algorithms usually call the features extracted by CNN networks as input. For example, when analyzing the behavior of a pet, the CNN network first extracts local features of the pet image, and then the behavior analysis algorithm will use these features, combined with pre-defined behavior patterns, to determine the current behavior state of the pet, such as whether the pet is still, moving, alert, or relaxed. Deep convolutional neural network (DCNN) is a special form of CNN network, which has a deeper network structure and more convolutional layers. DCNN is essentially a CNN network, and its network structure and connection method follow the basic principles of CNN. It's just that in DCNN, the number of convolutional layers and pooling layers is greater, and the depth of the network is deeper. By increasing the depth of the network, DCNN can learn more complex and abstract features, thereby achieving better performance in many tasks. In this embodiment, for example, in the pet type and posture classification tasks, the use of pre-trained deep convolutional neural networks (such as ResNet, VGG or MobileNet) can more accurately classify pets and recognize postures.Its connection with ordinary CNN is that they both extract features based on convolutional layers, but DCNN has more advantages in the depth and complexity of feature learning. In the processing process, the features extracted by DCNN can also be passed to subsequent fully connected layers, classifiers, regressors or behavior analysis algorithms for further processing like ordinary CNN.

[0073] Specifically, the intelligent sensing technology collects the pet's body shape and image data through the intelligent sensing module and image recognition camera of the mesh fixture. The AI ​​control module processes this data and calculates the appropriate size and shape of the fixture. Finally, the mesh fixture automatically adjusts its size and shape according to the instructions of the AI ​​control module to adapt to pets of different sizes, ensuring that the pet is comfortably and effectively fixed during the injection process. The following is the collaborative work of the components:

[0074] 1) Mesh fixture: Components: The mesh fixture is a frame structure made of lightweight and high-strength metal material, and the frame is covered with a mesh fabric made of nylon. Information: The mesh fixture is connected to the AI ​​control module through an intelligent sensing module to transmit the pet's body shape data in real time. These data include information such as the pet's body shape, shape, and position. Processing: After receiving the pet's body shape data, the AI ​​control module analyzes and processes these data through intelligent sensing technology to calculate the size and shape of the fixture suitable for the current pet. Adjustment: Based on the processing results, the mesh fixture will automatically adjust its size and shape to adapt to pets of different sizes. This adjustment is achieved by controlling the expansion and contraction of the frame structure and the tightness of the mesh fabric through the intelligent sensing module.

[0075] 2) AI control module: Components: The AI ​​control module includes a processor, a memory, and an algorithm unit. Information: The AI ​​control module receives the pet's body shape data from the mesh fixture and performs a comprehensive analysis in combination with the image data collected by the image recognition camera. Processing: The AI ​​control module processes the image data through the CNN image recognition algorithm, identifies the target injection site of the pet, and calculates the injection position through the injection control algorithm. Adjustment: Based on the processing results, the AI ​​control module will control the mesh fixture to adjust its size and shape to ensure that the pet is comfortably and effectively fixed during the injection process.

[0076] 3) Image recognition camera: Components: The image recognition camera is installed on an adjustable bracket above the inside of the mesh fixture. Information: The camera captures the pet's image data in real time, including the pet's body outline, posture, injection site and other information. Processing: The camera transmits the collected image data to the AI ​​control module, which processes the image data through the CNN image recognition algorithm to identify the pet's injection site and posture. Adjustment: Based on the image recognition results, the AI ​​control module will further adjust the size and shape of the mesh fixture to ensure that the pet remains stable during the injection process.

[0077] Preferably, the injection control algorithm involves multi-degree-of-freedom motion control of the robotic arm, which is closely related to the robot motion control algorithm (such as inverse kinematics, path planning, etc.). The related technology refers to the patent in the field of robotics: US20180029234A1 (Robot motion control method). The injection control algorithm involves precise dose control and injection speed control, which is related to the automated injection system in medical equipment. Reference paper: "Automated Drug Delivery Systems: A Review" in the Journal of Medical Systems.

[0078] Optionally, the frame structure is made of a lightweight and high-strength metal material, the density of which is less than 5.0 g / cm 3 , with a tensile strength greater than 400MPa, and is composed of several high-strength aluminum alloys (such as 6061 series) and titanium alloys, which are widely used in high-precision mechanical devices due to their excellent strength and lightweight characteristics. The design of the frame follows the modular principle, and all metal rods are connected or welded by high-strength bolts to ensure the overall rigidity and stability. The outside of the frame is covered with a layer of mesh fabric (such as nylon) to further enhance the stability of the structure and provide additional support.

[0079] Specifically, the sensors used in this embodiment include: an image sensor (such as a camera), which is used to obtain pet images and support the CNN image recognition algorithm to analyze the pet's body shape, position and injection site; a pressure sensor, which is used to monitor the pressure data during the injection process to ensure the safety of the injection process; a flow sensor, which is used to measure the flow rate of drug injection to ensure that the dosage and injection speed meet the predetermined requirements; a temperature sensor, which is used to monitor the temperature of the drug and the temperature of the injection site to ensure that the environment is suitable; a position sensor, which is used to accurately locate the pet and the injection device to ensure the accuracy of the injection position; a humidity sensor, which is used to monitor the ambient humidity to avoid the impact of humidity changes on the operation of the equipment; a current / voltage sensor, which is used to monitor the system power supply to ensure the stable operation of the equipment; an accelerometer and gyroscope (IMU sensor), which is used to monitor the movement state of the injection device to ensure that the equipment is stable and without offset; a biosensor, which is used to monitor the pet's physiological data (such as heart rate, respiratory rate, etc.) to ensure the safety of the injection process.

[0080] Furthermore, the injection drive mechanism is used to accurately control the injection process of the drug, and its structure includes a drive source, a transmission system, a piston, a control module and an injection syringe: Drive source: An electric drive system (such as a stepper motor or a DC motor) or a pneumatic drive system (such as a compressed air system) is used to provide a power source to control the push of the drug during the injection process. Transmission system: The drive source converts the rotational power into a linear driving force through a screw transmission system or a gear transmission system to ensure that the piston moves accurately in the injection syringe. Piston: The piston is connected to the transmission system and is used to push the drug in the injection syringe. Control module: Equipped with position sensors, pressure sensors, etc., to monitor the injection process in real time, and the feedback system adjusts the drive source to ensure that the injection dose and speed are accurate.

[0081] Preferably, the robotic arm operates according to the following command data issued by the AI ​​control module: Target position and posture instructions: According to the pet's body shape and injection site, the AI ​​control module calculates and issues the target position and angle to guide the end effector of the robotic arm to accurately align with the injection site. Path planning and trajectory instructions: The AI ​​control module calculates the path for the robotic arm to reach the target position from the current position, ensuring that obstacles are avoided and adjusting the motion trajectory according to the surrounding environment. Movement speed and acceleration instructions: According to the injection requirements, the AI ​​control module issues speed and acceleration control instructions to ensure that the movement of the robotic arm is smooth and meets the precise requirements of the injection. Feedback adjustment instructions: According to sensor feedback, the AI ​​control module adjusts the motion trajectory in real time to correct any deviations caused by external factors. Injection process control instructions: The AI ​​control module adjusts the drug injection rate based on real-time feedback such as pressure and flow to ensure that the drug injection process is stable and accurate.

[0082] Specifically, the injection drive mechanism accurately controls the injection dose and speed of the injection syringe by collecting data from multiple sensors (such as flow sensors, pressure sensors, position sensors, etc.) in real time. Specifically: Injection dose control: The flow sensor monitors the drug flow rate, and the injection drive mechanism adjusts the injection speed to ensure that the drug is accurately injected according to the set dose. At the same time, the remaining amount of drug in the injection syringe will also be fed back to the drive mechanism to prevent insufficient or excessive injection. Injection speed control: Through the pressure sensor and flow sensor, the drive mechanism adjusts the drug flow rate in real time to ensure that the pressure during the injection process remains within a safe range and to avoid too fast or too slow injection speeds causing adverse effects on pets.

[0083] Optionally, the data during the injection process include but are not limited to the following: basic information of the pet (such as body shape, weight, health status, etc.); image recognition data (such as pet body shape, injection target position, posture, etc.); injection dosage data (such as the amount of drug injected, the amount injected); injection speed data (drug flow rate); injection pressure data (pressure monitoring data); equipment status data (equipment power, working status, etc.); position and motion data (robotic arm motion trajectory, speed, etc.); environmental data (such as temperature, humidity and other environmental parameters); real-time feedback data (pet's vital signs, health status); injection process log (detailed operation records); these data will be uploaded to a remote server or the mobile terminal of the pet medical staff through the wireless communication module to facilitate remote monitoring, data storage and subsequent medical decision-making.

[0084] Furthermore, the system monitors various data during the injection process in real time through multiple sensors, and sends an alarm message to the remote server when an abnormal situation is detected. Specific detection devices and methods include: Flow sensor: detect abnormal drug flow rate (too fast or too slow) to ensure the accuracy of the injection dose. Pressure sensor: monitor whether the pressure is too high or too low during the injection process to avoid equipment failure or harm to the pet due to abnormal pressure. Position sensor: monitor the position and movement state of the robotic arm to ensure that it accurately reaches the injection target position. Temperature sensor: detect whether the device temperature and drug temperature are within the safe range to prevent the device from overheating or drug deterioration. Biosensor: monitor the pet's vital signs (such as heart rate, body temperature, etc.) in real time to detect whether there is an allergic reaction or other abnormal physiological state. Battery power sensor: monitor the battery power of the device to ensure that the device has enough power to complete the injection task. Fault diagnosis system: detect internal faults or system abnormalities of the device to ensure the normal operation of the device.

[0085] Specifically, severe stress reactions can include the following specific reactions: Severe struggling: The pet struggles violently due to fear or pain during the injection process, which may result in injection failure or injury to the pet. For example: The pet attempts to break free from the mesh fixture, causing the injection needle to fall off or the injection site to deviate. Aggressive behavior: The pet exhibits aggressive behavior due to fear or pain, such as biting and scratching. For example: The pet attempts to bite medical personnel or scratch the injection device. Fainting or shock: The pet may faint or be in shock due to excessive tension or pain, which may be life-threatening. For example: The pet suddenly loses consciousness or shows symptoms of shock (such as cold limbs, weak pulse, etc.)

[0086] Furthermore, severe allergic reactions can include the following specific reactions: Respiratory system symptoms: including shortness of breath, wheezing, coughing, and even symptoms such as dyspnea and cyanosis. This is due to airway narrowing caused by respiratory mucosal edema, bronchospasm and other reasons, which affects gas exchange and can be life-threatening in severe cases. For example, a pet's respiratory rate suddenly increases significantly after injection, accompanied by wheezing, which may indicate a severe allergic reaction involving the respiratory system. Neurological symptoms: Some pets may experience neurological symptoms such as lethargy, drowsiness, irritability, convulsions, and coma. This may be caused by insufficient blood supply to the brain, electrolyte disorders or neurotransmitter imbalances caused by allergic reactions. For example, a pet suddenly becomes unusually quiet or extremely excited after injection, and has abnormal behaviors such as limb convulsions. The impact of severe allergic reactions on the nervous system needs to be considered.

[0087] Specifically, the Canny algorithm and the convolution of the CNN network are used to extract local features of the image data. Local features can be expressed in many ways. On the one hand, they refer to the features of specific body parts such as ears, eyes, and legs. These features can accurately reflect the morphological details of the object. On the other hand, they also include body contour information and hair texture information. The body contour information can outline the overall shape of the object, while the hair texture information can reflect the texture and growth characteristics of the object's hair. By comprehensively extracting these local features, the information contained in the image can be grasped more comprehensively and meticulously.

[0088] Optionally, when using a behavior analysis algorithm such as a recurrent neural network (RNN) or a long short-term memory network (LSTM) to determine the behavior pattern of a pet, the following types of data are usually analyzed:

[0089] Image and video data:

[0090] 1) Movement posture information: From the pet’s video or continuous image frames, the position and posture changes of each part of its body can be extracted. For example, by analyzing the extension degree of the pet’s limbs and the bending angle of the body, it can be determined whether the pet is standing, sitting, running or jumping. If the pet’s limbs are straight and the body is leaning forward, it may be running; if the limbs are curled up and the body is lying down, it may be resting.

[0091] 2) Body movement trajectory: Tracking the pet's movement trajectory over a period of time can help us understand its activity range and movement patterns. For example, whether the pet often wanders in a specific area, whether it walks in a straight line or moves in circles, etc. This helps to determine whether the pet is exploring the environment, looking for food, or playing.

[0092] 3) Expression and demeanor: A pet’s facial expression and eyes can also reflect its behavior patterns. For example, a pet with eyes wide open and ears erect may mean that it is alert, while a pet with squinting eyes and a relaxed expression may mean that it is comfortable. By analyzing the pet’s facial features in the image, these subtle changes in expression can be identified.

[0093] Sensor data: 1) Accelerometer data: The accelerometer worn on the pet can record the acceleration changes of its movement. According to the magnitude and direction of the acceleration, the intensity and type of the pet's movement can be judged. 2) Gyroscope data: The gyroscope can measure the rotation angle and angular velocity of the pet's body. Combined with the accelerometer data, the pet's movement posture and movements can be analyzed more accurately. 3) Environmental sensor data: Environmental sensor data such as temperature, humidity, and light can also assist in analyzing the pet's behavior.

[0094] Sound data: Voice characteristics: The calls of pets contain rich information, such as the frequency, intensity, and duration of the calls. Different voice characteristics may correspond to different behavioral needs or emotional states. By analyzing the sound data, these voice characteristics can be identified to infer the behavior patterns of pets. Sound environment: In addition to the calls of pets themselves, the surrounding sound environment may also affect the behavior of pets. By analyzing the changes in the sound environment, we can understand the reactions of pets to different sound stimuli and further determine their behavior patterns.

[0095] Furthermore, when using the pre-trained deep convolutional neural network to classify the types and postures of pets, the following types of data are mainly analyzed:

[0096] Image data: 1) Static images: A large number of static images of pets are common analysis data. These images need to clearly show the appearance of the pet so that the network can learn and distinguish the morphological differences between different types of pets. For example, cats and dogs have obvious differences in head shape, ear shape, body proportions, etc. By analyzing these features in static images, the network can learn how to accurately classify pet types. At the same time, static images can also show specific postures of pets, such as standing, lying, running, etc. The network can classify postures according to the relative positions and angles of various parts of the pet's body.

[0097] 2) Multi-angle images: In order to more comprehensively identify the type and posture of the pet, images of the pet taken at different angles are used. Images from different angles can provide feature information from different sides of the pet. For example, the facial features of the pet may be clearer when viewed from the front, which is helpful for determining the type; while the side view is more conducive to observing the pet's body contour and limb posture, which is very helpful for posture classification. For example, taking images of the pet from multiple angles such as the front, side, and back can allow the network to learn richer features and improve the accuracy of classification.

[0098] 3) Images in different environments: Pets may be in a variety of different environments, such as on a sofa indoors or on a lawn outdoors. Using images in different environments for analysis allows the network to learn the characteristics of pets in various backgrounds and enhance the robustness of the network. Because the environment may interfere with the appearance of the pet, by learning in multiple environments, the network can better focus on the characteristics of the pet itself and accurately classify the type and posture.

[0099] Video data: 1) Continuous action sequence: Video data contains the pet's continuous action information, which is particularly important for posture classification. The network can analyze the movement trajectory and changes of various parts of the pet's body between different frames in the video to determine whether the pet is performing dynamic postures such as walking, jumping, and playing. For example, in a video of a pet playing, the network can identify whether it is chasing a toy or playing with its companions by analyzing the movement rhythm and amplitude of the pet's limbs.

[0100] 2) Behavior cycle data: Video data can also reflect the behavior cycle of pets, that is, the various stages that pets go through to complete a complete behavior. For example, the process of a pet eating may include approaching food, sniffing, eating, and leaving. By analyzing the behavior cycle in the video, the network can have a deeper understanding of the pet's posture changes and improve the accuracy of posture classification. At the same time, different types of pets may have differences in behavior cycles and posture performance, which also helps the network to classify species.

[0101] Labeled data: 1) Category labeling: For the image or video data used for training, there needs to be clear pet category labeling. The labeling information is usually presented in the form of labels, such as "cat", "dog", "rabbit", etc. These labeled data provide the learning target for the network. The network continuously adjusts its own parameters to make the classification result of the input data as close as possible to the labeled category label, thereby achieving accurate category classification.

[0102] 2) Posture annotation: Similarly, the data also needs to include annotation information of the pet’s posture. Posture annotation can be specific descriptions, such as “standing”, “sitting”, “lying down”, “running”, etc., or more detailed posture descriptions, such as “standing with bent front legs”, “lying down sideways”, etc. Based on these posture annotations, the network learns the characteristic patterns of different postures and then performs posture classification on the newly input data.

[0103] Preferably, the deep convolutional neural network is defined as follows:

[0104] Network Depth

[0105] The notable feature of a deep convolutional neural network is that it has a deep network structure. Generally speaking, the network has many layers (including convolutional layers, pooling layers, fully connected layers, etc.). It is generally believed that convolutional neural networks with more than ten layers or even more can be called deep convolutional neural networks. For example, the classic deep convolutional neural network model AlexNet has 8 layers, VGGNet has 16 or 19 layers, and ResNet has 34, 50, 101 or even 152 layers. This deep structure enables the network to gradually extract features from low-level to high-level, from simple to complex from the original input data.

[0106] Feature learning ability

[0107] Deep convolutional neural networks have powerful feature learning capabilities and can automatically learn highly abstract feature representations from large-scale data. When dealing with pet species and posture classification tasks, shallow convolutional layers can learn some basic features, such as edges, textures, etc.; as the network depth increases, subsequent convolutional layers can combine these low-level features into higher-level and more representative features, such as pet facial features, body contours, etc., so as to more accurately classify pet species and postures.

[0108] Training methods and data requirements

[0109] Due to the large number of parameters in deep convolutional neural networks, in order to avoid overfitting and learn effective features, some special training techniques are usually required, such as stochastic gradient descent (SGD) and its variants (such as Adam, Adagrad, etc.), regularization methods (such as L1, L2 regularization, Dropout, etc.). At the same time, deep convolutional neural networks have high requirements for the scale and diversity of training data. In the pet classification task, a large amount of pet images or video data of different types, different postures, and different environments is needed to train the network to ensure that the network can learn comprehensive and accurate features.

[0110] Model architecture design

[0111] Deep convolutional neural networks are often more complex and sophisticated in model architecture design. In addition to the basic convolutional layers, pooling layers, and fully connected layers, some special structures are introduced to improve network performance. For example, the residual block is widely used in deep convolutional neural networks such as ResNet. It solves the gradient vanishing and gradient exploding problems in the deep network training process by introducing skip connections, allowing the network to be trained deeper; the Inception module is used in networks such as GoogLeNet. It increases the width of the network and the diversity of features by using convolution kernels and pooling operations of different sizes in parallel in the same layer.

[0112] Preferably, the method of dynamically adjusting the injection strategy by the fully connected layer will be described in detail in combination with the position information, angle information, behavior pattern and classification results:

[0113] 1) Data input and preprocessing:

[0114] First, the position information, angle information, behavior pattern, and classification results are passed as input to the fully connected layer. This information may exist in different data formats. For example, the position information may be a coordinate value, the angle information may be an angle degree, and the behavior pattern and classification results may be encoded labels. Before entering the fully connected layer, these data need to be preprocessed, such as normalization operations, mapping data of different ranges to the same interval (such as [0,1]) to ensure that the influence of each feature on the model is relatively balanced, so that the fully connected layer can effectively learn and process.

[0115] 2) Learning and mapping of the fully connected layer:

[0116] The fully connected layer receives the preprocessed data and contains multiple neurons, each of which is connected to all neurons in the input layer. Through a large number of training samples, the fully connected layer learns the complex mapping relationship between the input data and the injection strategy. During the training process, a suitable loss function (such as mean square error loss function, cross entropy loss function, etc.) is used to measure the difference between the predicted injection strategy and the actual optimal injection strategy, and the weights and biases between neurons in the fully connected layer are continuously adjusted through the back propagation algorithm, so that the value of the loss function gradually decreases, thereby optimizing the performance of the model.

[0117] Dynamic adjustment based on different information

[0118] Adjustments based on location information:

[0119] Proximity to the target location: If the location information shows that the pet is far away from the target injection location of the injection device, the fully connected layer may adjust the injection strategy, such as slowing down the injection speed to wait for the pet to get closer; or adjusting the movement path of the injection device to make it closer to the pet. Conversely, if the pet is very close to the target injection location, the injection speed can be appropriately increased to improve the injection efficiency.

[0120] Position stability: If the pet's position is constantly changing and the amplitude is large, the fully connected layer may decide to suspend the injection operation to avoid inaccurate injection or harm to the pet due to the pet's movement; the injection will be continued after the pet's position is relatively stable. When the pet's position is stable in the target area, the normal injection plan will be followed.

[0121] Adjustment based on angle information:

[0122] Injection angle adaptation: Based on the angle information, the fully connected layer will adjust the angle of the injection device to determine the posture and orientation of the pet's body, ensuring that the injection needle can penetrate the pet's body at the best angle. For example, if the pet is standing sideways, the injection device may need to adjust the angle accordingly to inject perpendicular to the pet's body surface, improving the accuracy and safety of the injection.

[0123] Angle deviation correction: When it is detected that the actual injection angle deviates from the preset optimal injection angle, the fully connected layer will issue instructions in a timely manner to fine-tune the angle of the injection device so that it can return to the appropriate injection angle range as soon as possible.

[0124] Adjustments based on behavior patterns:

[0125] Active state adjustment: If the behavior pattern shows that the pet is in an active state, such as running, playing, etc., the fully connected layer may reduce the injection dose or delay the injection time to avoid injection when the pet is exercising vigorously and reduce the risk of injury to the pet. When the pet enters a relatively quiet state, such as resting or eating, the normal dose will be injected.

[0126] Stress response management: If the pet exhibits stress behavior, such as tension, resistance, etc., the fully connected layer will adjust the injection strategy, such as using a gentler injection method, such as slowly pushing the drug; or giving the pet a certain amount of comfort time, and then injecting after its stress response is relieved.

[0127] Adjustments based on classification results:

[0128] Differences in pet species: Different types of pets may have different tolerance and responses to drugs. Based on the classification results, if it is a small pet, the fully connected layer may reduce the injection dose accordingly to avoid excessive drug intake and harm to the pet; for large pets, the injection dose is appropriately increased to achieve the therapeutic effect.

[0129] Health status and special needs: If the classification results show that the pet has certain special health conditions, such as a history of allergies, chronic diseases, etc., the fully connected layer will adjust the injection strategy based on this information, such as selecting a more appropriate type of drug, adjusting the injection frequency, etc., to meet the pet's personalized treatment needs.

[0130] Real-time feedback and continuous adjustment:

[0131] During the injection process, the fully connected layer will receive new position information, angle information, behavior patterns, and classification results in real time, and continuously evaluate and adjust the injection strategy. If the current injection strategy is found to be ineffective, such as bleeding at the injection site or abnormal reactions of the pet, the fully connected layer will immediately recalculate and adjust the injection strategy based on the new data to ensure the safety and effectiveness of the injection process. Through the above methods, the fully connected layer can dynamically adjust the injection strategy based on position information, angle information, behavior patterns, and classification results to adapt to the different states and needs of the pet, and improve the accuracy and safety of the injection.

[0132] Specifically, in terms of products:

[0133] Intelligent mobile injection device 4: Based on the image recognition injection system, the robot arm of the mobile injection device 4 is called to perform multi-degree-of-freedom movement according to the instruction to accurately adjust the position and posture of the injection syringe;

[0134] Mesh Fixation Device 2: A fixation device used in conjunction with the image recognition injection system for surgery to prevent pets from being startled and reacting violently, which would affect the surgical operation. It is made of soft but tough material, which can effectively restrict the pet's movement while ensuring that the pet is comfortable and harmless.

[0135] High-precision image recognition camera 3: installed on an adjustable bracket above the fixture, electrically connected to the AI ​​control unit; has auto focus and zoom functions, can work under different lighting conditions, captures the pet's body image in real time and transmits it to the AI ​​control unit;

[0136] Intelligent system and hardware integration: pet intelligent injection system based on image recognition, such as AI control unit (core control part, including processor, memory and algorithm unit), wireless communication module, light sensing module uninterruptible power supply (UPS), etc.

[0137] User interaction device: Operate and obtain information through this interface, such as querying the basic information of the pet, displaying the injection preparation status, estimated injection time and other information, so that users can understand the injection progress; send maintenance reminders to users, display detailed error information, and prompt users to take corresponding measures.

[0138] Furthermore, in terms of methods:

[0139] Intelligent device calling method: the mobile injection device 4 accurately controls the injection dosage and speed according to the preset program to complete the drug injection, and the method includes the mobile injection device 4 and the related driving mechanism;

[0140] Personalized fixing method: The mesh fixing device 2 with intelligent sensing technology automatically adjusts the size and shape according to the pet's body shape to achieve personalized fixing; the method includes the mesh fixing device 2 and the related intelligent sensing module;

[0141] Intelligent Image Recognition Camera 3 Method: Capture pet images in real time and accurately identify specific injection sites (e.g., subcutaneous, intramuscular, etc.) under various lighting conditions (with night working mode). This method includes integrating advanced image recognition algorithms and light sensing modules;

[0142] Design and implementation of AI control algorithm: The algorithm in the AI ​​control unit processes the image data, calculates the precise injection position and path, and controls the mechanical arm of the mobile injection device 4 to move to the specified position; the method includes running the CNN-based image recognition algorithm and injection control algorithm, the processor is responsible for efficiently processing a large amount of image data and algorithm calculation tasks, the memory provides storage space for the algorithm model and data, and the algorithm unit realizes the training, optimization and real-time operation of the CNN algorithm;

[0143] User interaction and personalized service method: The system is provided with a user interaction interface 6, through which the pet owner can operate and obtain information; the method includes querying the basic information of the pet, displaying the injection preparation status, the estimated injection time, viewing the system's historical injection records and data analysis results, etc., and personalized recommendations for the pet owner on diet and exercise plans that are more suitable for puppies.

[0144] refer to Figure 2 , method scheme:

[0145] Step 1: Mesh Fixture 2 Design:

[0146] Execution entity: Design an adjustable mesh fixing device that can adapt to pets of different sizes2,

[0147] Step Description: The device is made of soft but tough material, which can effectively restrict the pet's movement while ensuring the pet is comfortable and harmless. Through intelligent sensing technology, the device can automatically adjust the size and shape according to the pet's body shape to achieve personalized fixation.

[0148] Step 2: Image recognition camera 3 integration:

[0149] Execution subject: High-resolution camera

[0150] Step description: A high-resolution camera is integrated inside or near the fixture, equipped with advanced image recognition algorithms, which can capture and identify the pet's specific injection site (such as subcutaneous, intramuscular, etc.) in real time. The camera also has a night working mode to ensure accurate recognition under various lighting conditions.

[0151] Step 3: AI control algorithm design and implementation

[0152] Execution body: AI control unit

[0153] Step description: The AI ​​control unit receives the image data transmitted by the camera, the image recognition algorithm of CNN and the injection control algorithm, performs image recognition and injection position calculation, and then controls the mobile injection device 4 to complete the precise injection. At the same time, the system data can be stored and managed to realize the intelligence and automation of the system. The algorithm has been pre-trained with a large amount of pet image data and can accurately identify the type of pet, body posture and different types of injection sites. The algorithm first pre-processes the image, including image enhancement, denoising and other operations to improve the image quality for subsequent analysis. Then the pet's body features are extracted through the feature extraction network, the pet is edge detected, and the Canny algorithm is used to assist the CNN network to better detect the pet's outline and key parts. Local feature learning automatically extracts key local features (such as ears, eyes, legs, etc.) through the convolution layer of the CNN network to help the model accurately identify the pet's body posture and injection site. The classifier and regressor are then used to determine the specific location and angle information of the injection site. At the same time, the AI ​​control unit also combines the real-time dynamic information of the pet (such as whether it is quiet, whether there are abnormal movements, etc.), uses a recurrent neural network (RNN) or a long short-term memory network (LSTM) to analyze the behavior pattern of the pet, and uses a pre-trained deep convolutional neural network (such as ResNet, VGG or MobileNet) to classify the type and posture of the pet based on the collected static images and multi-angle images. The last few layers of the network are classified through the fully connected layer (FC) to determine the type of pet (such as cat, dog) and its body posture (such as sitting, lying, standing), and dynamically adjust the injection strategy. Transfer learning on existing large pet image datasets (such as ImageNet or special pet datasets) can accelerate the training process of the algorithm and improve accuracy. When the pet is more agitated, appropriately increase the fixed force or delay the injection operation to ensure the safety and accuracy of the injection process. Based on the features and posture recognition results in the image, the AI ​​control unit uses a regression network to calculate the specific injection site location (such as subcutaneous, muscle, etc.) and estimate the injection angle. Use structures such as a fully convolutional network (FCN) or U-Net to make pixel-level predictions and accurately locate the injection point. The system stores image data, injection parameters, pet status information, etc. in real time, and regularly inputs the data into self-supervised learning or meta-learning models to further enhance the self-learning ability of the AI ​​algorithm.

[0154] Step 4: Mobile Injection Device 4 Control Logic

[0155] Execution body: Call the robot arm of the mobile injection device 4

[0156] Description of steps: According to the injection position and angle information calculated by the AI ​​algorithm, the AI ​​control unit generates a control instruction and sends it to the mobile injection device 4. The robotic arm of the mobile injection device 4 performs multi-degree-of-freedom movement according to the instruction to accurately adjust the position and posture of the injection syringe. During the movement, the robotic arm is equipped with a high-precision position sensor and force sensor, which feeds back the position and force information to the AI ​​control unit in real time. The AI ​​control unit makes real-time adjustments based on this feedback information to ensure that the injection syringe can reach the designated injection site accurately. At the same time, the injection drive mechanism accurately controls the injection process of the drug according to the preset injection dose and speed parameters, and continuously monitors the injection pressure during the injection process. Once an abnormality is found (such as excessive injection resistance, which may be a needle blockage, etc.), the injection is stopped immediately and an alarm is issued.

[0157] Step 5: Data storage and analysis

[0158] Execution body: Memory of AI control unit

[0159] Step description: During the entire injection process, the system will record all relevant data, including pet body shape data, image recognition results, injection location and dosage information, pet real-time status data, and various sensor feedback data during the injection process. These data are stored in the memory of the AI ​​control unit. Periodically, the system will analyze the stored data to explore the potential patterns and associations in the data. For example, analyze the best fixation method and injection parameters for different breeds of pets during injection, or the relationship between different injection sites and drug absorption effects. Through these analysis results, the AI ​​control algorithm and various parameter settings of the system are further optimized to improve the system's adaptability and accuracy to different pets and injection scenarios.

[0160] Step 6: Remote monitoring and diagnostic functions

[0161] Execution entity: Wireless communication module

[0162] Description of steps: The system is equipped with a wireless communication module, which can transmit key data during the injection process (such as the various data mentioned above) to a remote server or a mobile terminal of a pet medical staff in real time. Pet medical staff can remotely monitor the injection process and promptly discover and deal with possible problems. At the same time, when the system detects abnormal conditions (such as injection failure, severe stress reactions in pets; the severe stress reactions include: aggressive behavior, fainting, and shock, etc.), it will automatically send an alarm message to the remote server, and the server will notify relevant personnel to handle it according to preset rules. In addition, remote experts can also remotely diagnose the system by analyzing the uploaded data, and promptly update the system's algorithms and parameters to ensure the continued stable operation of the system.

[0163] Step 7: User Interface 6 Design

[0164] Execution body: User interaction interface 6

[0165] Description of steps: The system is provided with a user interaction interface 6, through which pet owners or medical personnel can operate and obtain information. Before the injection, the basic information of the pet (such as breed, age, weight, etc.) can be input through the interface, and the system will automatically adjust the relevant parameter settings according to this information. The interface will also display information such as injection preparation and estimated injection time, so that users can understand the injection process. After the injection is completed, the interface will present an injection result report, including information such as injection site, dosage, success, etc., and can provide follow-up care suggestions. At the same time, users can view the system's historical injection records and data analysis results through the interface to track and manage the health status of the pet.

[0166] Step 8: System self-check and maintenance reminder

[0167] Execution body: Self-check program

[0168] Step description: Each time the system is started, it will automatically perform a self-check procedure to check whether the working status of each hardware component (such as camera, robotic arm, sensor, power supply, etc.) is normal, and whether the software algorithm is running correctly. During the self-check process, if any abnormality is found, detailed error information will be displayed on the user interaction interface 6, and the user will be prompted to take corresponding measures. In addition, the system will calculate the wear and tear of each component based on the usage time and operating status. When maintenance or replacement of components is required (such as cleaning the camera lens, battery power decay and need to be replaced, etc.), the user will be given a maintenance reminder on the interface in advance to ensure that the system is always in the best working condition and to ensure the safety and accuracy of pet injections.

[0169] Step 9: Emergency Response Mechanism

[0170] Execution body: Emergency procedures and uninterruptible power supply (UPS)

[0171] Step description: Although the system is designed to ensure safe and stable operation, emergencies may still occur. For example, in the event of a sudden power outage, the system is equipped with an uninterruptible power supply (UPS), which can maintain system operation for a short period of time to ensure that the injection process can be safely completed (such as completing the current injection step or safely retracting the injection syringe, etc.), while saving the system status information and uploading it to the server after power is restored. If the pet has a severe allergic reaction or other emergency health condition, the system will immediately stop the injection and trigger an emergency alarm, and display an emergency treatment guide (such as how to perform first aid, contact a nearby veterinarian, etc.) on the user interface 6 to help pet owners or medical personnel quickly take correct measures to deal with the emergency.

[0172] Furthermore, the overall structure:

[0173] Mesh fixture 2: adopts a frame structure, the frame is made of lightweight and high-strength metal material to ensure stability; the frame is covered with soft and tough mesh fabric, such as special nylon material, which is connected to the AI ​​control unit through an intelligent sensing module to achieve the function of automatically adjusting the size and shape according to the pet's body shape;

[0174] High-precision image recognition camera 3: installed on an adjustable bracket above the inside of the fixture, electrically connected to the AI ​​control unit, and used to capture the pet's body image in real time; the camera is equipped with auto focus and zoom functions to adapt to pets of different distances and sizes;

[0175] AI control unit: It is the core control part of the system, including processor, memory and algorithm unit; the processor is responsible for data processing and command issuance, the memory stores image recognition algorithm and pet injection-related data, and the algorithm unit runs image recognition and injection control algorithms, where the image recognition algorithm is based on convolutional neural network (CNN);

[0176] Mobile injection device 4: It consists of a robotic arm, an injection syringe and an injection drive mechanism; the robotic arm is connected to the AI ​​control unit, receives its control instructions to achieve multi-degree-of-freedom movement, and ensures that the injection syringe can accurately reach the designated injection position; the injection syringe uses a high-precision medical syringe, and the injection drive mechanism can accurately control the injection dose and speed;

[0177] Power module 5: Provides stable power supply for all parts of the system. It can use rechargeable lithium batteries, installed in the battery compartment at the bottom of the system, and equipped with power monitoring and charging management circuits;

[0178] User interaction interface 6: Operations are performed and information is obtained through this interface, providing an intuitive and convenient operation interface for pet owners and medical personnel.

[0179] Furthermore, the connection relationship between the overall structure:

[0180] The intelligent sensing module of the mesh fixing device 2 and the AI ​​control module 1: transmit the pet's body shape data in real time through wireless connection;

[0181] The high-precision image recognition camera 3 and the AI ​​control module 1 are connected via a video transmission line to transmit the captured image data to the AI ​​control unit in real time for processing;

[0182] Mobile injection device 4 and AI control module 1: The AI ​​control unit is connected to the mechanical arm and injection drive mechanism of the mobile injection device 4 through a control signal line to control its movement and injection operation;

[0183] Power module 5 and intelligent injection system: The power module 5 supplies power to the entire system through a power line, is connected to the power interface of each power-consuming component, and is controlled by the power management instructions of the AI ​​control unit;

[0184] User interaction interface 6 and intelligent injection system: The interface will display information such as surgical injection preparation status and estimated injection time through the system, so that users can understand the injection process; after the injection is completed, the interface will present an injection result report, including injection site, dosage, success or failure, etc., and can provide follow-up care suggestions.

[0185] Specifically, the functions of each part and the implementation of related algorithms are as follows:

[0186] Mesh Fixation Device 2: Safely and comfortably fixes pets during injections to prevent them from moving, while its smart sensing technology allows for personalized fixation, improving pets’ comfort and safety;

[0187] High-precision image recognition camera 3: acquires pet body images in real time and provides accurate image data to the AI ​​control unit for image recognition;

[0188] AI control unit: receives camera image data and sensing data from the mesh fixing device 2, runs the CNN-based image recognition algorithm and injection control algorithm, performs image recognition and injection position calculation, and then controls the mobile injection device 4 to complete precise injection. At the same time, it can store and manage system data to realize the intelligence and automation of the system;

[0189] Mobile injection device 4: Under the control of the AI ​​control unit, the drug is accurately injected into the designated part of the pet. The flexible movement of its robotic arm and the precise control of the injection drive mechanism ensure the accuracy and safety of the injection;

[0190] Power module 5: Provides stable and reliable power for the entire system to ensure normal operation of the system. The power monitoring and charging management circuits can ensure the reasonable use and timely charging of the battery;

[0191] User Interface 6: Allows pet owners and medical personnel to easily input pet information, monitor injection progress, view injection results, and obtain follow-up care recommendations.

[0192] The beneficial effects of the present invention are as follows:

[0193] (1) Significantly improve safety: Through the design of the mesh fixing device, the pet is effectively and comfortably fixed during the injection process, avoiding accidental injuries caused by struggling or escaping. The mobile injection device controlled by the AI ​​algorithm can accurately calculate and guide the injection position, reducing human operating errors and further reducing the risks during the injection process.

[0194] (2) Significantly improve efficiency: The integration of image recognition cameras and AI algorithms enables rapid and accurate identification of pet injection sites, greatly shortening the preparation time before treatment. The automated injection process reduces manual intervention, making single injection operations faster and reducing the workload of pet medical staff.

[0195] (3) Significantly enhanced comfort: The mesh fixation device uses soft but tough materials, combined with intelligent sensing technology, to achieve personalized fixation, reducing the discomfort of pets during injection. Precise injection control strategies, such as optimizing the injection path and force, further reduce the pain and stress response of pets during injection.

[0196] (4) Promote intelligent upgrades: The system has self-learning capabilities and can continuously accumulate and optimize injection data to improve the overall performance and intelligence level of the system. This intelligent upgrade not only improves the accuracy and reliability of the system.

[0197] (5) The system uses a multi-task learning method to simultaneously process multiple related tasks (such as species recognition, posture recognition, and injection site location), and improves the accuracy of each task through joint training.

[0198] (6) The system uses reinforcement learning to optimize the control strategy based on real-time feedback, achieving adaptive adjustment of the injection process and improving the accuracy and safety of the injection process.

[0199] (7) The system architecture design is highly modular and scalable, which makes it easy to deploy on different hardware platforms and can also be expanded in functionality at a later stage.

[0200] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0201] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A pet intelligent injection system based on image recognition, characterized in that: include: Mesh fixture, AI control module, image recognition camera, mobile injection device, power module and user interface; The mesh fixing device, the image recognition camera, and the mobile injection device are respectively connected to the AI ​​control module; the image recognition camera is arranged on an adjustable bracket above the inside of the mesh fixing device; the power module is respectively connected to the mesh fixing device, the AI ​​control module, the image recognition camera, the mobile injection device, and the user interaction interface; the user interaction interface is connected to the AI ​​control module; The mesh fixing device is used to automatically adjust the size and shape according to the pet's body shape through intelligent sensing technology; the image recognition camera is used to collect image data of the pet; the AI ​​control module is used to identify the pet's target injection site based on the image data through a CNN image recognition algorithm, calculate the injection position based on the recognition result of the CNN image recognition algorithm through an injection control algorithm, and control the mobile injection device to perform injection according to the calculation result of the injection control algorithm; the power module is used to provide power for the mesh fixing device, the AI ​​control module, the image recognition camera, the mobile injection device and the user interaction interface; the user interaction interface is used to collect the operating instructions of medical personnel and display the injection result report.

2. According to claim 1, a pet intelligent injection system based on image recognition is characterized in that: The mesh fixing device adopts a frame structure; the frame structure is made of a lightweight and high-strength metal material; the frame structure is covered with a mesh fabric; the mesh fabric includes: nylon material.

3. The pet intelligent injection system based on image recognition according to claim 1, characterized in that: The AI ​​control module includes: A processor, used for adjusting the injection speed and injection dosage of the mobile injection device according to the injection pressure and flow rate data fed back by the preset sensor; A memory for storing the CNN image recognition algorithm, the injection control algorithm and pet injection operation data; the pet injection operation data includes: pet body shape data, image recognition results, injection position and dosage information, pet real-time status data and feedback data of all sensors during the injection process; the feedback data includes: pressure data, flow data, temperature data, humidity data, electrical data, acceleration data and physiological data; An algorithm unit is used to run the CNN image recognition algorithm and the injection control algorithm.

4. The pet intelligent injection system based on image recognition according to claim 1, characterized in that: The mobile injection device comprises: a mechanical arm, an injection drive mechanism and an injection syringe connected in sequence; The robotic arm is used to perform multi-degree-of-freedom movement according to the target position and posture instructions, path planning and trajectory instructions, movement speed and acceleration instructions, feedback adjustment instructions and injection process control instructions issued by the AI ​​control module; the injection drive mechanism is used to control the injection dose and injection speed of the injection syringe.

5. The pet intelligent injection system based on image recognition according to claim 1, characterized in that: Also includes: Wireless communication module; The wireless communication module is connected to the AI ​​control module; The wireless communication module is used to upload the feedback data during the injection process to a remote server or a mobile terminal of a pet medical staff, and send an alarm message to the remote server when a preset sensor detects an abnormal situation; The abnormal conditions include: injection failure, severe stress reaction of the pet; the severe stress reaction includes: aggressive behavior, fainting and shock.

6. The pet intelligent injection system based on image recognition according to claim 1, characterized in that: Also includes: Emergency handling module; The emergency processing module includes: an emergency processing program and an uninterruptible power supply; The emergency handling program is used to detect emergencies. When the emergency is a power outage, the uninterruptible power supply is used to provide power to the power module. When the emergency is a severe allergic reaction of a pet, the mobile injection device is stopped, an emergency alarm is triggered, and an emergency handling guide is presented to the user. The severe allergic reaction includes: shortness of breath, wheezing, coughing, and mental depression.

7. A pet intelligent injection system based on image recognition according to any one of claims 3 and 4, characterized in that: Also includes: Self-check procedures; The self-test program is used to check the working status of all hardware components when the AI ​​control module is started; the hardware components include the image recognition camera, the robotic arm, the sensor and the power module.

8. A pet intelligent injection method based on image recognition, characterized in that: A pet intelligent injection system based on image recognition as described in any one of claims 1 to 7, the method comprising: Adaptively fix the pet using the mesh fixing device; Using the image recognition camera to collect the image data; Using the AI ​​control module to identify the target injection site and calculate the injection position of the image data; injecting the pet using the mobile injection device according to the calculation result of the injection control algorithm; The workflow of the CNN image recognition algorithm includes: Extracting local features of the image data using the Canny algorithm and convolution of the CNN network; the local features include ears, eyes, and legs; Analyzing the local features using a classifier and a regressor to obtain position information and angle information of the target injection site; Determine the behavior pattern of the pet using a behavior analysis algorithm based on the collected image and video data, sensor data, and sound data; the behavior analysis algorithm includes: any one of a recurrent neural network and a long short-term memory network; According to the collected static images and multi-angle images, the pre-trained deep convolutional neural network is used to classify the types and postures of pets to obtain classification results; The injection strategy is dynamically adjusted using a fully connected layer according to the position information, the angle information, the behavior pattern and the classification result.

Images