Behavior monitoring and correcting system for pets

Abstract

The invention relates to the technical field of pet monitoring, in particular to a pet-oriented behavior monitoring and correcting system which comprises a camera module used for capturing behaviors of a pet at home, an AI algorithm is built in the camera module, various behaviors of the pet such as splitting and disordered pulling can be recognized, and when the camera module shows that the pet has bad behaviors, the camera module is used for monitoring the pet. A notification can be immediately sent to remote control; the vibration reminding module is used for correcting bad behaviors of the pet in time; the positioning module is used for monitoring the position of the pet in real time and ensuring the safety of the pet at home; the wireless communication module is used for communicating with the camera and the remote control module; and the remote control module receives the notification sent by the camera, and performs analysis and processing according to the behavior data of the pet.

Description

Technical Field

[0001] The present invention relates to the technical field of pet monitoring, and in particular to a pet-oriented behavior monitoring and correction system. Background Art

[0002] With the development of social economy and the improvement of people's living standards, the status of pets in the family is increasing, and more and more people regard pets as important family members. However, due to busy work or long time away from home, many pet owners are unable to accompany their pets all day, which may cause pets to have behavioral problems such as destroying furniture, defecating anywhere, and excessive barking due to loneliness, anxiety or bad habits. These behaviors not only affect the hygiene of the home environment, but may also have a negative impact on the mental health of pets. Therefore, how to effectively monitor and manage the behavior of pets has become an important topic in the current field of pet care.

[0003] At present, there are some pet monitoring devices on the market, such as smart cameras, pet GPS locators, etc., but most of these devices only have basic video monitoring or location tracking functions and cannot actively identify and intervene in the bad behavior of pets. Traditional correction methods mainly rely on manual intervention, such as using spray bottles, sound control or training collars, but these methods often lack real-time and intelligence, and are difficult to work when the owner is not at home. In addition, some training devices use electric shocks or strong vibrations, which may cause unnecessary fright or harm to pets, which is not conducive to the long-term behavior correction of pets.

[0004] In recent years, with the development of artificial intelligence (AI) and Internet of Things (IoT) technologies, smart pet monitoring systems have gradually become a research hotspot. AI behavior recognition algorithms based on computer vision can analyze pets' movement patterns and determine whether they have abnormal behaviors; wearable devices (such as smart collars) can provide immediate intervention through vibration, sound, or gentle stimulation to help pets establish correct behavioral habits. However, the existing system still needs to be improved in terms of real-time performance, accuracy, and human-computer interaction experience. In particular, how to combine AI analysis with wearable devices to achieve accurate and gentle behavior correction is still a direction that needs to be optimized.

[0005] In view of the above situation, in order to overcome the above technical problems, the present invention designs a behavior monitoring and correction system for pets to solve the above technical problems. Summary of the invention

[0006] The technical purpose to be achieved by the present invention is: the present invention proposes a behavior monitoring and correction system for pets, which combines wearable smart devices to achieve real-time monitoring and intelligent intervention of pet behavior. The system can automatically identify bad behaviors of pets, and make timely and gentle corrections through a vibration reminder module, while sending real-time alarms to the owner to facilitate remote management and adjustment of training strategies. This solution not only improves the level of intelligence in pet behavior management, but also reduces negative stimulation to pets, providing modern pet owners with a more scientific and efficient pet monitoring solution.

[0007] In order to achieve the above technical objectives, the present invention provides the following technical solutions: A behavior monitoring and correction system for pets, comprising: The camera module is used to capture the pet's behavior at home. The camera module has a built-in AI algorithm that can identify various behaviors of pets, such as destroying the house and pooping. When the camera module detects bad behavior of the pet, it will immediately send a notification to the remote control; Vibration reminder module, used to correct bad behavior of pets in time; The positioning module is used to monitor the location of the pet in real time to ensure the safety of the pet at home; A wireless communication module, used for communicating with the camera and the remote control module; The remote control module receives notifications from the camera and analyzes and processes the pet's behavior data.

[0008] Preferably, the AI ​​algorithm built into the camera module adopts multimodal data fusion technology, including visual dynamic capture, infrared thermal imaging analysis and sound spectrum detection. Visual dynamic capture tracks the pet's body movements in real time and identifies destructive behaviors such as house demolition and furniture climbing; infrared thermal imaging is used to monitor changes in the pet's body temperature at night or in low light environments, and judge whether it is in an anxious state based on a preset threshold; sound spectrum detection collects the frequency of pet calls through a microphone array and matches it with abnormal sounds in the database (such as biting and impact sounds). The algorithm also includes an incremental learning mechanism, and users can upload new behavior samples (such as pet breed-specific actions) through the remote control module to dynamically update the recognition model. When bad behavior is detected, the system automatically generates an encrypted log of timestamps, behavior types, and environmental parameters, and uploads it to the cloud through the wireless communication module.

[0009] Preferably, the vibration reminder module uses a piezoelectric ceramic vibrator with adjustable intensity, and its driving signal is generated by a PWM waveform controller, which responds in three levels according to the severity of the behavior: the first level is low-frequency micro-vibration (5 - 10 Hz) for mild warnings (such as approaching a dangerous area); the second level is medium-frequency pulsed vibration (15 - 20 Hz) for persistent bad behaviors (such as scratching the sofa); the third level is high-frequency intermittent vibration (25 - 30 Hz) for urgently stopping aggressive behaviors. The vibration duration is dynamically adjusted by a fuzzy logic algorithm. If the pet does not stop the action within 3 seconds, an upgraded response is triggered. The module is built-in with an inertial sensor that real-time feeds back the pet's body state changes to the remote control module. If it detects that the pet is unbalanced due to vibration, it immediately stops the output and sends a safety alarm.

[0010] Preferably, the positioning module uses UWB (Ultra-Wideband) and IMU (Inertial Measurement Unit) integrated positioning technology with a positioning accuracy of ±2 cm. UWB anchors are deployed at the indoor corners, and the relative position of the pet-worn tag is calculated through the TOF algorithm; the IMU is used to compensate for dead reckoning when the UWB signal is blocked. The system presets virtual geographical fences. When the pet enters dangerous areas such as the kitchen and balcony, it triggers the camera module to focus and start the vibration reminder. The positioning data is superimposed and displayed on the remote control terminal with the home three-dimensional map, supporting users to manually define restricted areas or set time period limits (such as prohibiting entry into the bedroom from 22:00 to 6:00). The positioning module is built-in with an RFID reader that can identify the ID of the smart collar worn by the pet to achieve multi-pet differentiated positioning.

[0011] Preferably, the remote control module integrates a decision-making engine based on reinforcement learning. Its state space includes pet breed, age, historical behavior records, and environmental variables (such as temperature, noise level), and the action space is correction strategies such as vibration intensity, voice warning, or releasing snacks. If the same behavior repeats more than 3 times within 7 days, the correction intensity is automatically increased. The control terminal provides a visual dashboard that shows a behavior heat map, correction success rate, and pet stress index (analyzed through galvanic skin activity data). Users can manually override the system decision or set scenario modes such as "away from home / sleep" to limit the activity intervention intensity during specific periods.

[0012] Preferably, the system supports Mesh networking of multiple camera modules and vibration reminder modules. The main node camera module uses edge computing to allocate tasks. For example, when camera A identifies the pet's running trajectory, it notifies camera B to pre-align with the corridor area. The synchronization error between devices < 5 ms and is calibrated through the NTPv4 protocol. The distributed database uses CRDT (Conflict-Free Replicated Data Type) to store behavior events to ensure the eventual consistency of local data during network disconnection. Users can batch upgrade device firmware or adjust the networking topology through the remote control module.

[0013] Preferably, the system is extended with a configuration of a biosensor collar that collects the pet's heart rate (PPG technology), skin conductance (EDA), and electromyogram signals (sEMG), with a sampling rate ≥ 100 Hz. The emotion recognition model uses an LSTM network to align the physiological signals with the timing of the behavior video and outputs emotion state scores such as anxiety and excitement. When a continuous high-pressure state is detected (heart rate > 180 bpm and EDA fluctuation > 5 μS), the system automatically reduces the vibration intensity and triggers soothing measures (such as playing white noise at a specific frequency). The collar is equipped with a micro air pump that can release pheromones (such as dog appeasing pheromone) for chemical intervention, and the release amount is closed-loop regulated by a PID controller according to the emotion score.

[0014] Preferably, all behavior events and correction records are stored on a lightweight blockchain for verification by the home gateway, user's mobile phone, and cloud node. Each transaction includes the behavior video, device digital signature, and time certificate, forming an immutable electronic evidence chain. The user can authorize veterinarians or dog trainers to access data for a specific period to analyze the causes of abnormal behavior. The blockchain smart contract supports an automatic claim settlement logic. If the system fails to respond to a dangerous behavior within a specified time (such as 10 seconds), a pre-deposited deposit will be paid to the pet insurance account.

[0015] Preferably, the system adopts a dynamic load balancing strategy: simple behavior recognition (such as standing still, walking) is locally processed by the NPU of the camera module; for complex scenarios (such as multi-pet interaction), 50% of the computational workload is offloaded to the edge server through model segmentation technology. Only key event videos are stored in the cloud, and the rest of the data is retained for 72 hours after being compressed by TSMC. The power management unit (PMU) of each module supports solar charging and wireless power supply. In standby mode, the camera module switches to event-driven sampling, and the power consumption is reduced to 0.1 W. The energy consumption data is visualized in real-time, providing suggestions for the "green mode" (such as adjusting the monitoring area to reduce the number of activated devices).

[0016] The beneficial effects of the present invention are as follows: (1) Through the deep integration of intelligent wearable devices and AI technology, this system provides a complete set of intelligent behavior management solutions for modern pet owners. The system uses an intelligent collar with a lightweight design, integrating high-precision motion sensors, environmental monitoring modules, and intelligent feedback devices, enabling all-weather monitoring and precise intervention of pet behavior. In terms of real-time monitoring, the system can accurately identify more than 30 common bad behaviors such as housebreaking, chewing, and abnormal barking through a nine-axis inertial measurement unit and an AI behavior recognition algorithm, with an identification accuracy rate of over 95%. Notably, the system uses multi-modal perception technology to distinguish between playful chewing and destructive behavior by analyzing parameters such as movement trajectory, force, and duration, avoiding misjudgment.

[0017] (2) This system is especially suitable for pet-owning families in modern cities. It can not only effectively solve the behavior problems of pets when they are alone, but also help the owners better understand the needs of their pets through scientific data analysis, and establish a more harmonious human-pet relationship. As the usage time increases, the AI algorithm of the system will be continuously optimized to provide a completely personalized behavior management plan for each pet. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0019] Now the above and other aspects of the present invention will be described by way of example only with reference to the drawings, where: Figure 1 is a schematic diagram of the system flow of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] In order to better understand the above technical solutions, the following will combine the drawings in the specification and specific embodiments to describe the above technical solutions in detail.

[0021] As Figure 1 shown, an intelligent behavior monitoring and correction system for pets. This system is an intelligent pet guardianship solution based on Internet of Things technology, aiming to help pet owners remotely manage and correct the bad behaviors of their pets. The system consists of multiple functional modules, forming a complete monitoring-identification-correction closed-loop system.

[0022] The core of the system is the high-definition camera module. This module uses a 1080P wide-angle camera with night vision function, which can capture the activities of pets at home all day long. The camera module is built-in with an advanced AI behavior recognition algorithm, which can accurately identify various behavior patterns of pets through deep learning technology, including but not limited to bad behaviors such as destroying the home, soiling randomly, climbing dangerous areas, etc. When the system detects an abnormal behavior, it will immediately send a real-time alarm to the remote control end through the wireless communication module.

[0023] The vibration reminder module is the core of behavior correction of this system. It adopts a wearable design and interrupts the bad behaviors of pets through gentle vibration stimulation. This module has a multi-level intensity adjustment function, which can intelligently adjust the vibration intensity according to the pet's body size and the severity of the behavior, ensuring that the behavior can be effectively stopped without harming the pet.

[0024] The positioning module adopts UWB ultra-wideband technology, which can achieve indoor positioning with centimeter-level accuracy and track the location of pets in real time. Combined with the electronic fence function, when the pet approaches a dangerous area, the system will automatically issue a warning and activate the correction mechanism.

[0025] The wireless communication module supports Wi-Fi / Bluetooth dual-mode connection, ensuring stable data transmission between components of the system. At the same time, it supports cloud synchronization, allowing the owner to view the pet's status at any time through the mobile phone APP.

[0026] The remote control module provides a visual operation interface, supporting functions such as behavior data analysis, historical record query, and custom rule setting. The system will generate behavior reports to help the owner understand the pet's behavior pattern and formulate a more scientific training plan.

[0027] This system solves the problem of pet behavior management when alone through intelligent means, which not only ensures home safety but also helps to cultivate good habits for pets. It is an ideal choice for modern pet owners.

[0028] The AI algorithm built into the camera module adopts advanced multi-modal data fusion technology, and realizes the all-round intelligent analysis of pet behavior by integrating various sensing data. The technical architecture mainly includes three core detection dimensions: The visual dynamic capture system uses a real-time object detection algorithm improved based on YOLOv5 to accurately track the pet's limb movement trajectory at a frame rate of 30fps, and identifies typical destructive behaviors such as housebreaking and climbing furniture through the pose estimation algorithm. The infrared thermal imaging unit is equipped with a thermal sensor, which can monitor the surface temperature distribution of the pet in real time at night or in low-light environments. Combining the preset body temperature fluctuation threshold (±1.5°C) and regional temperature difference characteristics, it can effectively judge whether the pet is in a stressed or anxious state; The sound detection system uses an annular microphone array and a noise reduction algorithm to extract the voiceprint characteristics of the pet's barking, and intelligently matches them with abnormal sound samples (such as biting, scratching, hitting, etc.) in the behavior feature database through Mel frequency cepstral coefficient analysis.

[0029] This algorithm innovatively adopts an incremental learning mechanism under the federated learning framework. Users can upload newly collected behavior samples (such as unique behavior characteristics of specific dog breeds) through the remote control module. The system will automatically extract key features and generate a new training dataset. Without affecting the performance of the original model, the weight parameters of the recognition model can be dynamically updated weekly. All detected bad behavior events will automatically generate structured logs, including timestamps accurate to milliseconds, behavior type tags (according to the ICD-PET behavior classification standard), and 12-dimensional parameters such as environmental light / temperature and humidity, which are uploaded to the cloud database to ensure data security and privacy protection. The system also supports the visual backtracking of behavior data, and users can view the behavior analysis report at any time period through the time axis method.

[0030] The vibration reminder module adopts a high-precision piezoelectric ceramic vibrator array, and the drive system uses a digital PWM waveform controller. It implements a three-level gradient response strategy according to the severity determined by the AI behavior analysis module: The first-level response outputs low-frequency micro-vibrations of 5 - 10 Hz / 0.5 G for 0.5 seconds, which is suitable for early warning scenarios (such as a pet approaching a restricted area); the second-level response generates a medium-frequency pulse sequence of 15 - 20 Hz / 1.2 G (vibrations with 3 intervals of 200 ms), specifically for stubborn bad behaviors (such as continuous scratching of furniture); the third-level response activates a high-frequency intermittent mode of 25 - 30 Hz / 2.0 G (1 second ON / 0.5 second OFF cycle) to quickly block dangerous behaviors (such as attacking other pets). Detailed parameter logs of all vibration events are recorded, including the trigger time, response level, pet reaction data, etc., for evaluating the effect of later behavior correction.

[0031] The positioning module adopts a multi-source fusion positioning architecture of UWB (Ultra-Wideband) and IMU (Inertial Measurement Unit), and achieves centimeter-level (±2 cm) precise positioning through the Kalman filtering algorithm. The system consists of 4 UWB anchor nodes (DW1000 chipset), and uses a hybrid ranging algorithm of TDOA (Time Difference of Arrival) and TOF (Time of Flight) to achieve three-dimensional coordinate calculation within a 10 m × 10 m space. The IMU unit integrates a three-axis accelerometer, gyroscope, and magnetometer, and automatically switches to a MEMS-based dead reckoning mode when the UWB signal is blocked (such as blocked by furniture), and the positioning drift error is controlled within 0.1 m / min.

[0032] The system has a built-in intelligent geofence function, and pre-marks 7 types of dangerous areas such as kitchen stoves and balcony railings in the three-dimensional digital map of the home (accuracy 5 cm) constructed by SLAM technology. When a pet approaches within 1 m of a dangerous area, a three-level response mechanism is triggered: automatically adjust the PTZ camera pan-tilt for target tracking and shooting; activate the progressive warning of the collar vibration module (the frequency increases as the distance decreases); push graded alarms to the user terminal (APP / sms).

[0033] The remote control module constructs an intelligent decision-making engine using a deep reinforcement learning framework. Its state space consists of multi-dimensional feature vectors, including: pet attribute dimension (static parameters such as breed, age, weight, etc.); behavior history dimension (frequency, type, and time period distribution of bad behaviors in the past 30 days); environmental perception dimension (temperature and humidity sensor data, environmental noise decibel value, light intensity, etc.). The action space contains 8 configurable correction strategy combinations: vibration intensity (adjustable from 5 - 30 Hz), pre-recorded voice commands (3 soothing tones), snack reward delivery, and smart home linkage (such as automatically closing the door).

[0034] This module has an adaptive learning mechanism. When it detects that the same type of bad behavior occurs more than 3 times within 7 days, the system will automatically start behavior pattern analysis: First, cluster the behavior characteristics to identify potential triggering factors; then dynamically adjust the policy weights and gradually increase the correction intensity (the increase amplitude each time does not exceed 20%). At the same time, a "fatigue detection" mechanism is introduced. When it monitors that the pet stress index (calculated from the electrodermal response data collected by the EDA sensor built into the collar) exceeds the threshold for 2 consecutive hours, it automatically switches to the gentle intervention mode.

[0035] The system supports multi-level permission management: Ordinary users can manually override a single decision (such as canceling the current vibration instruction), and advanced users can configure complex scenario strategies (such as setting the "late night mode" to limit the maximum vibration intensity below 15 Hz). All operation records are stored and certified through blockchain technology to ensure the traceability of the training process. The module also provides multiple interfaces to support scene linkage with mainstream smart home platforms. For example, when the "leaving home mode" is triggered, it automatically activates the whole-house monitoring and adjusts the sensitivity of the correction strategy.

[0036] The system adopts a Mesh networking architecture, supporting dynamic networking of up to 16 camera modules and vibration reminder modules. The main node camera module is equipped with an edge computing platform, and realizes intelligent load balancing through a distributed task scheduling algorithm: allocates computing resources based on the target detection confidence; dynamically adjusts the working mode according to the remaining battery power of the device; adopts predictive pre-alignment technology - when camera A detects the pet running vector (speed > 2 m / s) through the optical flow method, it immediately notifies adjacent cameras B and C to pre-rotate the pan-tilt through the multicast UDP protocol, and increases the monitoring frame rate of the predicted path areas such as corridors and doorways to 60 fps. The network layer adopts TSN (Time-Sensitive Networking) technology to ensure that the time alignment error of multi-view video frames is < 5 ms by realizing time synchronization between devices.

[0037] The data persistence layer adopts a decentralized storage architecture. Each node has an embedded database built-in, and uses the OR-Set (Observed-Removed Set) model in CRDT (Conflict-Free Replicated Data Types) to store behavior event logs. This design realizes three major characteristics: 1) Network partition tolerance - each node can continue to run independently for 72 hours when the network is disconnected; 2) Eventual consistency - automatically merges conflict data after the connection is restored; 3) Version traceability - each event is attached with a logical timestamp. The security mechanism adopts two-factor authentication: encrypts the communication between devices and verifies the data synchronization.

[0038] The remote management system provides a topology visualization editor, supporting: firmware differential upgrade (only transmitting modified blocks); wireless signal strength heat map display; dynamically adjusting the networking routing strategy (such as setting low-battery nodes as leaf nodes). Users can re-plan the parent-child relationship between devices by dragging the interface, and the system automatically calculates the optimal TDMA time slot allocation scheme. The operation and maintenance module has a built-in self-diagnosis function, which can detect abnormal states such as network loops and clock drifts, and automatically generate a physical topology map through the LLDP protocol. All configuration changes generate blockchain evidence, supporting rollback to any historical version.

[0039] The emotion recognition engine adopts a neural network architecture enhanced by an attention mechanism. The input layer extracts temporal features from physiological signals and performs cross-modal alignment with visual behavior features (limb movement encodings extracted by 3D CNN). The output layer generates a six-dimensional emotion state score (anxiety, excitement, fear, relaxation, pleasure, neutral) through the softmax function, with a comprehensive confidence level > 85%. The system is specially designed with a stress response mechanism: when a high-pressure state is continuously detected (meeting the conditions of heart rate > 180 bpm and EDA fluctuation > 5 μS for more than 30 seconds), a three-level relaxation plan is immediately activated: 1) automatically reduce the intensity of the vibration module to below 5 Hz; 2) play pink noise with a bandwidth of 125 - 250 Hz through a bone conduction speaker; 3) activate the microfluidic chip built into the collar to release dog appeasing pheromone (DAP) with an accuracy of 0.1 μL / min.

[0040] The collar is equipped with a 1.5 mL replaceable pheromone capsule, with a battery life of up to 30 days, and records usage logs through NFC tags. All intervention records automatically generate a biofeedback report, including physiological parameter trend charts, emotion state migration matrices, etc., supporting real-time data viewing through a mobile phone APP. The system is specially designed with a security protection mechanism. When abnormal physiological signals are detected (such as heart rate continuously > 220 bpm), an emergency alarm is immediately sent and medical advice is recommended.

[0041] This system uses a hierarchical blockchain architecture to achieve full-link behavior data evidence storage, consisting of three levels of verification nodes: the home gateway (edge node) is responsible for real-time collection of raw data; the user mobile terminal (light node) conducts preliminary verification; the cloud server (full node) completes the final consensus. Each behavior record is packaged as a structured transaction, containing quadruple data: an 8-second behavior video clip encoded in H.265 (with ROI area annotation), the device hardware fingerprint, and an evaluation report on the execution effect of the correction strategy. These data are organized into an immutable evidence chain, using an improved PBFT consensus algorithm.

[0042] The data sharing mechanism is developed based on zero-knowledge proof technology. Users can generate access tokens for specific time periods through a mobile APP (such as authorizing veterinarians to view data in the most recent 72 hours). The tokens contain the following customizable permissions: time range limitation (accurate to the minute); behavior type filtering (only showing aggression or anxiety-related events); sensitive data desensitization (such as blurring the faces of family members in videos). The authorized party can view enhanced visualization analysis tools such as three-dimensional behavior trajectory reconstruction and physiological index change curves through a blockchain browser.

[0043] The smart contract module contains an automatic performance fulfillment subsystem, and its core logic is as follows: The pre-deposited deposit is managed by a dual-signature cold wallet; the response timeliness detector continuously monitors the full-link delay from behavior recognition to intervention execution; if the system fails to meet the SLA agreement (such as a 10-second timeout for responding to dangerous behaviors), then a self-verifying contract is triggered: First, the authenticity of the device log is verified through an Oracle node, and then an ERC-20 standard compensation transfer is automatically executed. The compensation amount is calculated stepwise according to the delay time (basic deposit × 2 ^ (delay seconds / 5)). The contract code has been formally verified (using the Solidity verification toolchain), and key function calls require secondary confirmation via the user's biometric characteristics (fingerprint / face) to ensure the security of funds. All claims settlement records are permanently stored through IPFS distributed storage to support the compliance audit requirements of insurance companies.

[0044] The system uses an intelligent hierarchical computing architecture to achieve dynamic load balancing and automatically allocates task resources through a computing complexity evaluation engine. Basic behavior recognition tasks (including 8 simple behaviors such as standing still, walking, and eating) are locally completed by a 4TOPS computing power NPU (neural network processor) built into the camera module, and a quantized model is used to achieve real-time processing with 95% accuracy (delay < 50ms).

[0045] The data storage system adopts a three-level hierarchical architecture: Key event videos (such as vandalism and health abnormalities) are permanently stored in their original format in the cloud, and behavior summaries (key frames + spatio-temporal cube features) are automatically generated; Regular monitoring videos are processed by the (spatio-temporal motion compression) algorithm, automatically cleared after 72 hours of retention. This algorithm achieves a compression ratio of 10:1 through optical flow analysis while maintaining 90% behavior recognition availability; Metadata (such as behavior type tags and location coordinates) are stored in a columnar storage format for long-term preservation, supporting multi-dimensional queries at the millisecond level.

[0046] The energy management system innovatively integrates multiple power supply solutions: Each module's PMU (Power Management Unit) supports 15W solar charging (conversion efficiency > 23%) and Qi wireless power supply (induction distance up to 3 cm); The Dynamic Voltage and Frequency Scaling (DVFS) technology adjusts the processor's operating point in real time according to the calculated load; In the intelligent standby mode, the camera module switches to an event-driven sampling mechanism, and only the PIR sensor and the low-power MCU maintain operation. The main processor is awakened within 50 ms after detecting a moving target, reducing the average power consumption to below 0.1W.

[0047] The system provides an energy dashboard function. After analyzing the energy consumption patterns of each device through a reinforcement learning algorithm, it gives optimization suggestions: Spatial dimension suggestions: Based on the heat map, identify low-activity areas and recommend reducing the activation number of monitoring devices by 10% - 30%; Temporal dimension strategy: Automatically generate a device scheduling plan according to the master's work and rest schedule (such as reducing the frame rate to 15fps at night); Green mode: When continuous cloudy days are detected, automatically switch to an ultra-low power consumption protocol to prioritize core security functions. All energy-saving measures are simulated in 3D visualization through a digital twin system, and users can intuitively preview the energy consumption comparison before and after adjustment.

[0048] For those of ordinary skill in the art, various modifications to the present disclosure will be apparent, and the general principles defined herein can be applied to other variations without departing from the scope of the present disclosure. Therefore, the present disclosure is not limited to the examples and designs described herein, but should be accorded the widest scope consistent with the principles and novel features disclosed herein. Although one or more exemplary embodiments of the present disclosure have been described with reference to the accompanying drawings, those of ordinary skill in the art will understand that various changes in form and detail may be made therein without departing from the spirit and scope of the present disclosure as defined by the appended claims.

Claims

1. A pet-oriented behavior monitoring and correction system, characterized in that: include: The camera module is used to capture the pet's behavior at home. The camera module has a built-in AI algorithm that can identify various behaviors of pets, such as destroying the house and pooping. When the camera module detects bad behavior of the pet, it will immediately send a notification to the remote control; Vibration reminder module, used to correct bad behavior of pets in time; Positioning module, used to monitor the location of pets in real time to ensure the safety of pets at home; A wireless communication module, used for communicating with the camera and the remote control module; The remote control module receives notifications from the camera and analyzes and processes the pet's behavior data.

2. A pet-oriented behavior monitoring and correction system according to claim 1, characterized in that: The camera module has a built-in AI algorithm, which adopts multimodal data fusion technology, including visual dynamic capture, infrared thermal imaging analysis and sound spectrum detection.

3. A pet-oriented behavior monitoring and correction system according to claim 1, characterized in that: The vibration reminder module uses a piezoelectric ceramic vibrator with adjustable intensity. Its driving signal is generated by a PWM waveform controller. It has three levels of response according to the severity of the behavior: the first level is low-frequency micro-vibration (5-10Hz), which is used for mild warnings (such as approaching a dangerous area); the second level is medium-frequency pulse vibration (15-20Hz), which is aimed at persistent bad behavior (such as scratching the sofa); the third level is high-frequency intermittent vibration (25-30Hz), which is used to urgently stop aggressive behavior.

4. A pet-oriented behavior monitoring and correction system according to claim 1, characterized in that: The positioning module calculates the relative position of the tag worn by the pet through the TOF algorithm; when the pet enters dangerous areas such as the kitchen and balcony, the camera module is triggered to focus and shoot and start the vibration reminder.

5. A pet-oriented behavior monitoring and correction system according to claim 1, characterized in that: The remote control module integrates a decision engine based on reinforcement learning, whose state space includes pet breed, age, historical behavior records and environmental variables (such as temperature and noise level), and the action space is correction strategies such as vibration intensity, voice warning or releasing snacks.

6. A pet-oriented behavior monitoring and correction system according to claim 1, characterized in that: The system supports Mesh networking of multiple camera modules and vibration reminder modules. The main node camera module uses edge computing to assign tasks. Users can batch upgrade device firmware or adjust the network topology through the remote control module.

7. A pet-oriented behavior monitoring and correction system according to claim 1, characterized in that: It also includes a biosensor collar that collects the pet's heart rate (PPG technology), skin conductance (EDA) and electromyographic signals (sEMG). The collar has a built-in micro air pump that can release pheromones for chemical intervention. The amount released is regulated in a closed loop by a PID controller based on the emotion score.

8. The pet-oriented behavior monitoring and correction system according to claim 1, characterized in that: All behavioral events and correction records are stored through lightweight blockchain and verified by the home gateway, user mobile phone and cloud node. Each transaction includes a video of the behavior, a device digital signature, and a time certificate, forming an unalterable electronic chain of evidence.

9. The pet-oriented behavior monitoring and correction system according to claim 1, characterized in that: The system adopts a dynamic load balancing strategy: simple behavior recognition (such as stillness and walking) is processed locally by the NPU of the camera module; Complex scenarios (such as multi-pet interactions) offload 50% of computing workload to edge servers through model segmentation technology.

Images