Infrared night vision video acquisition and cloud storage method of a video pet feeder

Abstract

This application relates to the field of intelligent pet feeding technology, and provides a method for infrared night vision video acquisition and cloud storage of a video-enabled pet feeder. The method includes: acquiring infrared video footage of pet activity in real time in a dark environment using an infrared sensor built into the pet feeder; compressing the acquired infrared video footage in real time using a preset advanced video encoding standard to generate compressed video data; transmitting the compressed video data to a preset distributed container cloud platform; receiving and storing the compressed video data through the distributed container cloud platform, establishing a video storage index uniquely associated with the corresponding pet feeder; and when a video playback request is received from a user terminal, the distributed container cloud platform retrieves the corresponding compressed video data according to the video storage index, decodes it, and sends it to the user terminal for playback.

Description

Technical Field

[0001] This application relates to the field of intelligent pet feeding technology, and in particular to a method for infrared night vision video acquisition and cloud storage of a video-enabled pet feeder. Background Technology

[0002] With the increasing popularity of smart pet feeding devices, video-enabled pet feeders have become a core tool for remote pet monitoring. Currently, the video monitoring systems of existing video-enabled pet feeders suffer from multi-dimensional systemic technical deficiencies and have not yet formed a complete solution that can simultaneously meet the requirements of high-definition nighttime acquisition, low-power transmission, and high-concurrency cloud services.

[0003] First, there are inherent blind spots in nighttime video capture. Existing products generally use ordinary infrared sensors with weak visible light supplementation for night vision. The images captured in completely dark environments suffer from severe noise, blurred edges, and motion blur, making it impossible to clearly identify the subtle movements and status of pets. Nighttime monitoring can only achieve basic presence detection and cannot meet the needs of refined care.

[0004] Secondly, there is an imbalance between video transmission and storage costs. Existing encoding schemes have insufficient compression efficiency, resulting in excessively large amounts of high-definition video data at night. This not only consumes a large amount of network bandwidth and causes transmission delays, but also significantly increases cloud storage costs, forcing manufacturers to reduce video resolution or shorten storage cycles, sacrificing user experience.

[0005] Secondly, there are bottlenecks in the capacity of cloud services. Existing cloud platforms mostly adopt traditional virtual machine deployment architectures, which have slow elastic scaling response and low resource utilization. They are unable to support the concurrent access of a large number of users and the concurrent writing of massive amounts of video data. During peak user access periods, video stuttering, loading failures, and even data loss frequently occur, making it impossible to guarantee stable monitoring services around the clock.

[0006] The aforementioned technical deficiencies are mutually restrictive, making it impossible for existing video-enabled pet feeders to achieve true 24 / 7 high-definition remote monitoring, thus failing to meet users' core needs for pet safety monitoring. Summary of the Invention

[0007] This application provides an infrared night vision video acquisition and cloud storage method for video pet feeders, aiming to solve the problem that the existing video monitoring system of video pet feeders has multiple systemic technical defects and has not yet formed a complete solution that can simultaneously meet the requirements of high-definition nighttime acquisition, low-power transmission and high-concurrency cloud services.

[0008] In a first aspect, embodiments of this application provide an infrared night vision video acquisition and cloud storage method for a video-enabled pet feeder, applied to a pet feeder; including: The built-in infrared sensor in the pet feeder captures real-time infrared video footage of the pet's activities in dark environments; the captured infrared video footage is then compressed in real-time using a preset advanced video encoding standard to generate compressed video data. The compressed video data is transmitted to a pre-set distributed container cloud platform; the compressed video data is received and stored through the distributed container cloud platform, and a video storage index uniquely associated with the corresponding pet feeder is established. When a video playback request is received from a user terminal, the distributed container cloud platform retrieves the corresponding compressed video data based on the video storage index, decodes it, and sends it to the user terminal for playback. It also controls the distributed container cloud platform to call a pre-trained pet target detection model to detect the stored infrared video data frame by frame, identify pet eating behavior, and count the number of times the pet eats daily and the duration of each feeding. The statistical results are then pushed to the bound user terminal. Furthermore, the pet feeder is controlled to perform frame difference detection on the continuously collected infrared video frames. If no pixel change is detected for three consecutive frames, it is determined that there is no pet activity, and the infrared video acquisition frame rate is reduced to one frame per second. When a pixel change is detected and pet activity is determined, the original acquisition frame rate is restored.

[0009] In some embodiments, the real-time acquisition of infrared video footage of pet activities in a dark environment using the built-in infrared sensor of the pet feeder includes: detecting ambient light brightness using the built-in photosensor of the pet feeder, determining an environment where the ambient light brightness is lower than a preset brightness threshold as a dark environment, activating the built-in infrared sensor to acquire infrared video footage, and outputting the infrared video footage of pet activities at a fixed frame rate.

[0010] In some embodiments, the step of performing real-time compression processing on the acquired infrared video images using a preset advanced video coding standard to generate compressed video data includes: identifying static background areas and dynamic pet areas in the acquired infrared video images, using low bitrate encoding for the static background areas and high bitrate encoding for the dynamic pet areas, and splicing all coded blocks to generate compressed video data.

[0011] In some embodiments, transmitting the compressed video data to a preset distributed container cloud platform includes: detecting the real-time transmission bandwidth of the network currently accessed by the pet feeder, adjusting the sending frame rate of the compressed video data according to the real-time transmission bandwidth, adding verification information to the compressed video data in blocks, and then sending them sequentially to the distributed container cloud platform.

[0012] In some embodiments, receiving and storing the compressed video data through a distributed container cloud platform and establishing a video storage index uniquely associated with the corresponding pet feeder includes: after receiving the compressed video data on the distributed container cloud platform, extracting the unique device identifier of the sending pet feeder, recording the generation time and storage node address of the compressed video data, binding the generation time and storage node address with the device identifier as the core, generating a video storage index, and storing it in the index database.

[0013] In some embodiments, when a video playback request is received from a user terminal, the distributed container cloud platform retrieves the corresponding compressed video data according to the video storage index, decodes it, and sends it to the user terminal for playback. This includes: after receiving the playback request from the user terminal, the distributed container cloud platform extracts the device identifier and playback time range from the request, matches the corresponding storage node address according to the video storage index, retrieves the compressed video data for the corresponding time range, decodes it, transcodes it according to the requested bitrate, and sends it to the user terminal for playback.

[0014] In some embodiments, the method further includes: controlling the distributed container cloud platform to store compressed video data, extracting video clips containing pet activities using a moving target detection algorithm, deleting static video clips that do not contain pet activities, updating the video storage index, and saving cloud storage space.

[0015] In some embodiments, the method further includes: comparing the acquired infrared video footage with pre-stored facial features of authorized personnel, and generating an alarm message when an unauthorized person appears in the footage, and sending the alarm message along with the real-time captured footage to the user terminal.

[0016] This application achieves noise-free high-definition video capture in completely dark environments, completely eliminating blind spots in nighttime monitoring and enabling clear 24 / 7 display of pet activity. It employs efficient video encoding standards for real-time compression, significantly reducing video data transmission bandwidth and cloud storage costs while maintaining image clarity. Leveraging a high-concurrency distributed container cloud platform, it stably supports concurrent access from hundreds of millions of users and the writing and reading of massive amounts of video data, ensuring the 24 / 7 stability of the monitoring service. Furthermore, it enables centralized cloud storage and unified index management of video data, allowing users to accurately review historical videos of any time period anytime, anywhere via terminal devices, overcoming local storage capacity limitations.

[0017] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic flowchart illustrating the steps of an infrared night vision video acquisition and cloud storage method for a video-enabled pet feeder according to an embodiment of this application; Figure 2 This is a schematic diagram of the structure of a pet feeder provided in one embodiment of this application; Figure 3 This is a schematic block diagram of the structure of an infrared night vision video acquisition and cloud storage system for a video-enabled pet feeder provided in one embodiment of this application; Figure 4 This is a schematic block diagram of the structure of a pet feeder provided in one embodiment of this application.

[0020] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Detailed Implementation

[0021] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0023] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and the terms "first" and "second" are not necessarily different.

[0024] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0025] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0026] With the increasing popularity of smart pet feeding devices, video-enabled pet feeders have become a core tool for remote pet monitoring. Currently, the video monitoring systems of existing video-enabled pet feeders suffer from multi-dimensional systemic technical deficiencies and have not yet formed a complete solution that can simultaneously meet the requirements of high-definition nighttime acquisition, low-power transmission, and high-concurrency cloud services.

[0027] First, there are inherent blind spots in nighttime video capture. Existing products generally use ordinary infrared sensors with weak visible light supplementation for night vision. The images captured in completely dark environments suffer from severe noise, blurred edges, and motion blur, making it impossible to clearly identify the subtle movements and status of pets. Nighttime monitoring can only achieve basic presence detection and cannot meet the needs of refined care.

[0028] Secondly, there is an imbalance between video transmission and storage costs. Existing encoding schemes have insufficient compression efficiency, resulting in excessively large amounts of high-definition video data at night. This not only consumes a large amount of network bandwidth and causes transmission delays, but also significantly increases cloud storage costs, forcing manufacturers to reduce video resolution or shorten storage cycles, sacrificing user experience.

[0029] Secondly, there are bottlenecks in the capacity of cloud services. Existing cloud platforms mostly adopt traditional virtual machine deployment architectures, which have slow elastic scaling response and low resource utilization. They are unable to support the concurrent access of a large number of users and the concurrent writing of massive amounts of video data. During peak user access periods, video stuttering, loading failures, and even data loss frequently occur, making it impossible to guarantee stable monitoring services around the clock.

[0030] The aforementioned technical deficiencies are mutually restrictive, making it impossible for existing video-enabled pet feeders to achieve true 24 / 7 high-definition remote monitoring, thus failing to meet users' core needs for pet safety monitoring.

[0031] To solve the above problem, please refer to Figure 1 This application provides a method for infrared night vision video acquisition and cloud storage of a video-enabled pet feeder, applicable to applications such as... Figure 2The pet feeder shown. At the same time, it should be noted that each piece of information involved in the method provided in this application is extracted under the authorization of relevant users and in compliance with relevant regulations, and will not violate the privacy of users.

[0032] The infrared night vision video acquisition and cloud storage method for the provided video-type pet feeder includes steps S101 to step S103. Details are as follows: Step S101. Use the infrared sensor built in the pet feeder to collect real-time infrared video images of pet activities in a dark environment; perform real-time compression processing on the collected infrared video images using a preset advanced video coding standard to generate compressed video data.

[0033] Specifically, the method of the present invention is applied to an intelligent pet feeder with a remote video monitoring function. The intelligent pet feeder hardware is integrated with an environmental light sensor, a high-resolution near-infrared CMOS image sensor, an active array infrared fill light group, a hardware encoding chip, a WiFi6 / 5G dual-mode network communication module, and is communicatively connected to the distributed container cloud platform supporting the present invention at the back end. The user initiates a real-time viewing or historical video playback request through a user terminal (smartphone, tablet, PC, smart cat's eye, etc.) installed with a supporting pet care application program, and finally realizes all-weather high-definition remote pet care.

[0034] This step addresses the shortcomings of existing technologies in nighttime image acquisition, such as excessive noise, blurred edges, motion blur, and insufficient compression efficiency. It is divided into three practical sub-steps: infrared acquisition mode triggering, adaptive preprocessing of raw frames, and optimized advanced coding compression. First, the high-precision ambient light sensor integrated on the front of the pet feeder collects the illuminance value of the current environment in real time. The acquisition period is fixed at 100ms. When the illuminance value detected for three consecutive acquisition periods is lower than the preset dark environment threshold (the default threshold in this embodiment is 5 lux, but users can customize the threshold size in the accompanying APP according to the actual usage scenario, such as the lighting conditions of the placement location), it is determined that the current environment is completely dark or extremely low light, and the infrared night vision acquisition mode is automatically triggered: the visible light supplement light is turned off (to avoid wasting energy and disturbing the pet's rest at night), the built-in 850nm near-infrared high-resolution CMOS image sensor is activated, and the active array infrared supplement light group is activated at the same time. The output power of the supplement light group is adaptively adjusted according to the detected ambient illuminance: the lower the ambient illuminance, the higher the supplement light output power, with the maximum output power not exceeding 1W, which ensures sufficient supplement light and avoids unnecessary energy waste. In this embodiment, the infrared sensor adopts a 1920×1080 full HD resolution, with a fixed acquisition frame rate of 25fps. Compared to the 10-15fps acquisition frame rate commonly used in existing video feeders, this effectively matches dynamic scenes such as pets running and jumping quickly, eliminating the motion blur problem commonly found in existing technologies and ensuring the clarity of dynamic images. After acquisition, raw infrared video frames in YUV format are obtained. Since infrared acquisition only retains grayscale information, the basic data volume is inherently lower than that of color visible light images, reducing the basic load for subsequent compression processing.

[0035] To address the inherent problem of high background noise in infrared image acquisition in completely dark environments, this embodiment performs targeted preprocessing on the original frame before encoding: based on the inter-frame difference method, it identifies the moving target area (i.e., the pet's activity area) and the static background area (such as fixed areas like walls, ground, and feeder bases) in the image. For the static background area, a 5×5 median filter is used to remove large noise blocks, while only mild Gaussian denoising is performed on the moving target area, thus fully preserving the pet's movement details, fur texture, and other information. This solves the problem of blurred pet target edges and loss of details caused by uniform denoising across the entire image in existing technologies.

[0036] The preprocessed raw video frames are fed into the feeder's built-in dedicated hardware encoding chip (in this embodiment, an RTOS chip is used, which natively supports hardware encoding of the MP4 / H.265 advanced video encoding standard, with low encoding latency, low computing power consumption, and adaptability to the computing power limitations of embedded devices) for real-time compression. Addressing the characteristics of concentrated grayscale distribution and low texture contrast in infrared night vision images, this embodiment performs targeted adaptive optimization of the encoding parameters: for the pet movement target area identified in the preprocessing, smaller encoding units are used (minimum support of 8×8 encoding blocks), reducing the quantization step size and preserving more detail information to ensure the clarity of the target area; for static background areas, larger encoding units are used (maximum support of 64×64 encoding blocks), increasing the quantization step size and significantly improving compression efficiency. In this embodiment, the I-frame interval is set to 50 frames (corresponding to a 2-second duration), balancing the response speed of random video access with the overall compression rate. After final encoding, compressed video data conforming to international standard formats is generated, and each video segment is given a unique timestamp and a globally unique device identifier for the feeder, thus completing step S101.

[0037] Step S102. Transmit the compressed video data to a preset distributed container cloud platform; receive and store the compressed video data through the distributed container cloud platform, and establish a video storage index uniquely associated with the corresponding pet feeder.

[0038] Specifically, this step addresses the shortcomings of existing technologies, such as high transmission costs, poor cloud architecture elasticity, and insufficient concurrent capacity. It is divided into three sub-steps: stable streaming transmission, containerized layered storage, and distributed index construction. The pet feeder's network communication module supports WiFi 6 / 5G dual-mode communication and automatically selects the optimal communication link based on the signal strength and bandwidth of the currently available network. The compressed video data is divided into 128KB blocks and sent to the preset distributed container cloud platform through a low-latency streaming transmission protocol based on TCP. Forward error correction coding is added during transmission. When the network packet loss rate is less than 1%, there is no need for retransmission, further reducing transmission latency and ensuring the real-time performance of the video.

[0039] Distributed Container Cloud Platform with Containerized Tiered Storage: This embodiment's distributed container cloud platform is built on the Kubernetes container orchestration engine. All video access, storage, and management services are deployed as containerized microservices. Compared to the traditional virtual machine deployment architecture commonly used by existing vendors, resource utilization is improved by more than 40%, and elastic scaling response time is reduced from minutes to seconds. The platform is divided into two layers: an edge access layer and a core storage layer. The edge access layer is responsible for receiving video data uploaded by pet feeders. It adopts a dynamic weighted load balancing strategy, and monitors the current CPU utilization, memory utilization, and outbound bandwidth ratio of all access nodes in real time. New device connections are automatically scheduled to access nodes with an overall load below 70%, avoiding single-node overload and downtime. When the concurrent access volume of the entire platform exceeds a preset threshold (the current total cluster capacity exceeds 80%), the Kubernetes cluster automatically triggers elastic scaling, which can start new video access container nodes within 30 seconds to handle the new concurrent requests, completely solving the problems of slow elastic scaling response and insufficient resources during peak periods in traditional architectures. At the storage level, a hot and cold tiered storage strategy is adopted to reduce costs: newly uploaded compressed video data and hot data from the last 7 days are stored in high-performance cloud SSD solid-state storage disks to meet the requirements of low latency and fast access; cold data stored for more than 7 days is automatically migrated to a low-cost object storage cluster, with storage costs only 1 / 5 of SSD storage, which significantly reduces cloud storage costs and solves the problem that existing technologies are forced to reduce clarity and shorten storage cycles in order to control costs.

[0040] Device-unique video storage index construction: Each pet feeder is assigned a globally unique and unchanging device ID in the distributed container cloud platform. This embodiment uses a (device ID-date) sharding key rule to construct a distributed storage index. The index is stored in the highly available distributed key-value database Etcd. Each index entry corresponds to one day's video data for one feeder. The index entry content includes: the storage node address of all video shards for that day, the start timestamp, end timestamp, video resolution, and encoding format for each shard. For example, for a feeder with device ID PF-2026000123, the index storage key for April 22, 2026 is PF-2026000123_20260422, which can achieve second-level location of the target video position and avoid the latency caused by full-platform traversal queries. Step S102 is completed after the index construction is completed.

[0041] Step S103. When a video playback request is received from a user terminal, the distributed container cloud platform retrieves the corresponding compressed video data according to the video storage index, decodes it, and sends it to the user terminal for playback. By controlling the distributed container cloud platform to call the pre-trained pet target detection model, the platform detects the stored infrared video data frame by frame, identifies pet eating behavior, counts the number of times the pet eats per day and the duration of each feeding, and pushes the statistical results to the bound user terminal. The platform also controls the pet feeder to perform frame difference detection on the continuously collected infrared video frames. When no pixel change is detected for three consecutive frames, it is determined that there is no pet activity, and the infrared video acquisition frame rate is reduced to one frame per second. When a pixel change is detected and it is determined that there is pet activity, the original acquisition frame rate is restored.

[0042] Specifically, when a user initiates a request to view live or watch historical videos, the user opens the accompanying pet care APP on their user terminal (in this embodiment, a commonly used smartphone is used as an example), completes account authentication, selects the bound target pet feeder, slides to select the target time interval to be watched (the live viewing request corresponds to the current time interval), and clicks to initiate a playback request. The request is sent to the distributed container cloud platform via encrypted HTTPS protocol. The platform first performs two-way permission verification on the user's identity and device binding relationship. After confirming that the user has access to the device's video data, it proceeds to the next step. If the verification fails, it directly returns a permission error message, ensuring the security of the user's video data throughout the process.

[0043] Video data retrieval based on storage index: After successful verification, the platform quickly queries the corresponding index entry in the Etcd distributed index based on the device ID and target time range carried in the request, locating the storage address where the video segment of the target time range is located. If the target data is hot data stored in SSD, it is read directly from the corresponding storage node with a read latency of less than 100ms. If the target data is cold data stored in object storage, an automatic cold data preheating process is triggered, pulling the corresponding data to the edge cache node before subsequent processing, balancing storage cost and access speed. Simultaneously, the platform sets up an edge caching strategy to automatically cache frequently accessed video data from the most recent 24 hours to edge access nodes. User requests can return directly from the edge node without needing to go back to the core storage cluster, further reducing the load on the core storage cluster and significantly improving the platform's overall concurrent processing capabilities.

[0044] This embodiment employs an edge-cloud collaborative adaptive decoding strategy, automatically selecting the decoding location based on the user terminal's decoding capabilities: if the user terminal supports AV1 / H.265 hardware decoding (a function supported by most mainstream smartphones), the compressed video data is directly sent to the user terminal for local decoding and playback, maximizing the saving of cloud computing resources and reducing cloud operating costs; if the user terminal's decoding capabilities are insufficient (such as some older smart devices or low-configuration tablets), the decoding service container in the cloud container cluster performs the decoding and transcodes it into a bitstream adapted to the terminal's screen resolution before sending it to the terminal. The decoded video data is sent to the user terminal via the HLS low-latency streaming protocol, enabling smooth playback and supporting common operations such as dragging the progress bar, fast forwarding, pausing, taking screenshots, downloading, and sharing.

[0045] Infrared video captured in complete darkness achieves a resolution of up to 1080P, clearly identifying subtle movements and states of pets such as eating, drinking, scratching, and vomiting. This overcomes the limitations of existing nighttime monitoring technologies, which can only detect the presence of pets and cannot meet the needs of refined care. At the same resolution, the video bitrate is reduced by 50% compared to traditional solutions, and the free video storage period for users can be extended from the industry-standard 7 days to 15 days at the same storage cost, without sacrificing resolution to control costs. Cloud concurrency handling capacity is more than three times higher than traditional virtual machine architectures. During peak periods with millions of concurrent users, the video loading failure rate is reduced from 8% to below 0.2%, eliminating issues such as stuttering, loading failures, and even data loss. This supports stable 24 / 7 service for a large number of users, truly realizing 24 / 7 high-definition remote monitoring of video pet feeders and meeting users' core needs for pet safety monitoring.

[0046] Building upon the core data acquisition and storage process, an automatic pet eating behavior statistics function is added. In this embodiment, the distributed container cloud platform is set to automatically invoke a pre-trained lightweight YOLOv8n pet target detection model at 2:00 AM daily (a low-peak period for user access, with sufficient cluster computing power to ensure uninterrupted real-time transmission and playback services). This model performs target detection frame-by-frame on all infrared video data stored the previous day. The model is pre-trained using millions of infrared night vision images of pet feeders from different scenarios and pet breeds. It can accurately identify three core targets: the pet's overall outline, the pet's head, and the feeder bowl. The model has a small number of parameters, fast inference speed, and is suitable for batch processing on the cloud platform.

[0047] After the model detection is completed, the pet's eating behavior is determined based on preset rules: when the center coordinates of the pet's head area are detected within the food bowl detection frame for 3 or more consecutive frames, it is determined that a feeding behavior has begun; when the pet's head is away from the food bowl area for more than 5 minutes, the feeding behavior is determined to have ended. The duration of a single feeding is calculated, and the total number of feedings and the total daily feeding duration are accumulated to generate a daily pet eating statistics report. After the statistics report is generated, it is pushed to the user's terminal APP bound to the device via APNs push and the manufacturer's push channel, allowing the user to directly view the pet's eating status for the day. If the system detects that no eating behavior has been recorded for 12 consecutive hours, it will also proactively generate an abnormality reminder push, indicating that the pet may have problems such as loss of appetite or physical abnormalities, helping users to detect pet health abnormalities in a timely manner and meeting the user's needs for refined pet care.

[0048] Based on the core acquisition and storage process, a local acquisition frame rate adaptive adjustment function is added to reduce device power consumption and invalid data generation. In this embodiment, the local main controller of the pet feeder performs real-time frame difference detection on the continuously acquired infrared video frames. The specific judgment rule is as follows: the three most recently acquired infrared video frames are continuously taken, and the inter-frame difference operation is performed on each pair of adjacent frames. If no pixel change exceeding the threshold is detected in the pairwise comparison of the three frames (in this embodiment, the threshold is set to: the proportion of difference pixels in the whole frame is less than 0.05%), it is determined that there is no pet activity. In order to reduce power consumption and reduce invalid data generation, the infrared video acquisition frame rate is reduced from the original 25fps to 1 frame per second, maintaining only the lowest frequency of environmental monitoring. When a frame difference detection detects that the pixel change ratio exceeds 0.05%, it is determined that there is pet or human activity, and the original 25fps acquisition frame rate is immediately restored to ensure that no activity is missed in the recording.

[0049] This embodiment employs a three-frame consecutive judgment rule, which effectively avoids misjudgments caused by instantaneous interference such as light and shadow fluctuations and flying insects, thus improving judgment accuracy. This solution can significantly reduce the overall power consumption of pet feeders. For battery-powered portable pet feeders, it can effectively extend battery life, while reducing the generation and transmission of invalid data, further reducing network bandwidth and cloud storage usage, and improving the overall system operating efficiency.

[0050] In some embodiments, the real-time acquisition of infrared video footage of pet activities in a dark environment using the built-in infrared sensor of the pet feeder includes: detecting ambient light brightness using the built-in photosensor of the pet feeder, determining an environment where the ambient light brightness is lower than a preset brightness threshold as a dark environment, activating the built-in infrared sensor to acquire infrared video footage, and outputting the infrared video footage of pet activities at a fixed frame rate.

[0051] This embodiment provides a detailed description of the implementation of the core step S101 (real-time acquisition of infrared video footage of pet activity in a dark environment using an infrared sensor built into the pet feeder). In this embodiment, the pet feeder has an industrial-grade low-power digital photosensitive sensor built into the lower part of the front-end acquisition module. Its brightness detection accuracy can reach ±0.1 lux, which can meet the detection requirements of low-brightness environments. The photosensitive sensor continuously acquires the current ambient light brightness data at a fixed cycle of 100ms / time, and the acquisition results are stored in real time in the cache area of ​​the ARM architecture main control chip of the pet feeder. To avoid misjudging the dark environment due to instantaneous brightness fluctuations (such as a pet passing by and blocking the sensor, sudden brightness changes when switching on or off lights, or temporary obstruction by the user when picking up items), this embodiment sets an anti-shake judgment rule: only when the ambient light brightness output for three consecutive acquisition cycles is lower than a preset brightness threshold is the current environment officially judged as a dark environment, triggering the infrared night vision acquisition process.

[0052] In this embodiment, the default value of the preset brightness threshold is set to 5 lux, which corresponds to the indoor ambient brightness at night without artificial light source input. At the same time, users can customize and adjust the threshold according to the actual placement scenario through the mobile APP bound to the pet feeder: if the feeder is placed by a window where there is still a small amount of ambient light at night, the user can increase the threshold to 8 lux; if it is placed in a completely enclosed storage cabinet or bathroom without natural light, the threshold can be reduced to 3 lux, so as to adapt to the triggering needs of different family scenarios.

[0053] Once the environment is officially determined to be dark, the pet feeder's main controller automatically shuts off the visible light supplementary lighting to avoid wasting energy and disturbing the pet's rest at night. It then activates the built-in 1920×1080 resolution near-infrared CMOS image sensor and simultaneously turns on the 850nm wavelength active array infrared supplementary lighting, outputting infrared video footage of the pet's activities at a fixed frame rate of 25fps. This embodiment uses a fixed 25fps acquisition frame rate, which, compared to the 10-15fps frame rate commonly used in existing video pet feeders, is more adaptable to dynamic scenes such as pets running, turning their heads, and eating, effectively eliminating motion blur and ensuring image clarity and smoothness during subsequent analysis and review. The final output is a raw infrared video stream that meets processing requirements.

[0054] In some embodiments, the step of performing real-time compression processing on the acquired infrared video images using a preset advanced video coding standard to generate compressed video data includes: identifying static background areas and dynamic pet areas in the acquired infrared video images, using low bitrate encoding for the static background areas and high bitrate encoding for the dynamic pet areas, and splicing all coded blocks to generate compressed video data.

[0055] This embodiment provides a detailed description of the implementation of the core step S101 (compressing the acquired infrared video images in real time using a preset advanced video coding standard to generate compressed video data). The technical solution and implementation method are as follows: Before encoding and compression, this embodiment first divides each frame of input infrared video into static and dynamic regions. When the pet feeder first enters the infrared night vision mode, it continuously captures 10 frames of initial environmental images without pet activity. An initial static background model is established based on the Gaussian mixture background modeling algorithm. Subsequently, for each new frame of infrared video, the new frame is compared with the established static background model by performing inter-frame difference operation. Pixel regions with a grayscale difference greater than 12 grayscale levels are marked as candidate dynamic regions. Then, contour extraction and area filtering are performed on the candidate dynamic regions to remove noise and false positive regions caused by light and shadow fluctuations with an area of ​​less than 100 pixels. Finally, the accurate dynamic pet region is obtained, and all other unmarked areas in the image are determined to be static background regions.

[0056] After region segmentation, differentiated bitrate encoding is performed based on the H.265 / AV1 advanced video coding standard: For static background areas (mostly walls, floors, feeder mounting bodies, surrounding furniture, and other unchanging areas), a maximum size of 64×64 encoding blocks is used, with a quantization parameter QP of 38, employing low bitrate encoding to significantly improve the overall video compression rate; for dynamic pet areas, a minimum size of 8×8 encoding blocks is used, with a quantization parameter QP of 22, employing high bitrate encoding to fully preserve key information such as the pet's fur texture, movement details, and facial features. After all block encoding is completed, all encoding blocks are concatenated according to the bitstream rules of the advanced coding standard, ultimately generating compressed video data conforming to the standard format.

[0057] The differentiated encoding scheme in this embodiment can improve the overall video compression rate by 30% to 40% while ensuring the clarity of the pet target area. This effectively reduces the pressure on subsequent network transmission and cloud storage, and solves the defects of the existing technology's unified encoding of the whole screen, which either has insufficient clarity in key areas or low overall compression rate and consumes too many resources.

[0058] In some embodiments, transmitting the compressed video data to a preset distributed container cloud platform includes: detecting the real-time transmission bandwidth of the network currently accessed by the pet feeder, adjusting the sending frame rate of the compressed video data according to the real-time transmission bandwidth, adding verification information to the compressed video data in blocks, and then sending them sequentially to the distributed container cloud platform.

[0059] This embodiment provides a detailed description of the specific implementation of the core step S102 (transmitting the compressed video data to a preset distributed container cloud platform). The technical solution and implementation method are as follows: In this embodiment, the main control unit of the pet feeder performs a real-time transmission bandwidth detection of the currently connected home network every 2 seconds: it sends a 1KB standard probe packet to the dedicated speed test node of the distributed container cloud platform, and calculates the available real-time transmission bandwidth of the current network based on the round-trip time (RTT) and the actual packet loss rate of the probe packet. The calculation formula is: Available bandwidth = (probe packet size / RTT) × (1 - packet loss rate). Then, based on the calculated available bandwidth, the transmission frame rate of the compressed video data is adaptively adjusted. The specific adjustment rules are as follows: When the available bandwidth is ≥2Mbps, maintain the original full frame rate of 25fps to output complete high-definition video; When 1Mbps ≤ available bandwidth < 2Mbps, adjust the transmission frame rate to 15fps to reduce bandwidth usage while ensuring basic smoothness of the picture, and adapt to the scenario of multiple devices sharing bandwidth in the home. When 512Kbps ≤ available bandwidth < 1Mbps, adjust the transmission frame rate to 10fps; When the available bandwidth is less than 512Kbps, the transmission frame rate is adjusted to 5fps to ensure continuous and uninterrupted video transmission and avoid large-scale packet loss and transmission interruption.

[0060] After adjusting the transmission frame rate, the compressed video data is divided into multiple data segments at 128KB / s. A 16-bit CRC cyclic redundancy check is added to each data segment, and then the segments are sent sequentially to the distributed container cloud platform. Upon receiving each segment, the cloud platform first performs a check. If the check fails, a retransmission request is sent to the pet feeder; if the check passes, the segment is stored in the receive buffer. This mechanism ensures transmission accuracy and adapts to the real-world usage scenario of fluctuating shared bandwidth in homes, effectively avoiding transmission interruptions and data loss due to insufficient bandwidth.

[0061] In some embodiments, receiving and storing the compressed video data through a distributed container cloud platform and establishing a video storage index uniquely associated with the corresponding pet feeder includes: after receiving the compressed video data on the distributed container cloud platform, extracting the unique device identifier of the sending pet feeder, recording the generation time and storage node address of the compressed video data, binding the generation time and storage node address with the device identifier as the core, generating a video storage index, and storing it in the index database.

[0062] This embodiment provides a detailed definition of the specific implementation of the core step S102 (receiving and storing the compressed video data through a distributed container cloud platform, and establishing a video storage index uniquely associated with the corresponding pet feeder). The technical solution and implementation method are as follows: Each video pet feeder described in this invention is programmed with a globally unique 48-bit device ID in the one-time programmable (OTP) memory area of ​​the main control chip when it leaves the factory. This ID corresponds one-to-one with the device, and there will be no duplicates or tampering. When all compressed video data is sent, this device ID will be written in the data header as a unique identifier for the sending end.

[0063] After receiving the compressed video data, the distributed container cloud platform first parses the data header to extract the unique device identifier of the sending end. Then, it adds a generation timestamp accurate to milliseconds to each video segment. Based on the real-time load data of the distributed container cluster (obtaining node CPU utilization and remaining storage capacity through the node metrics interface of the Kubernetes cluster), it allocates the video segments to the storage node with the lowest current load and records four core pieces of information for the segment: the segment start timestamp, the segment end timestamp, the storage node's internal IP address, and the segment's starting offset address and length in the node's storage volume.

[0064] After all information is recorded, the generation time of all video segments for that device is bound to the storage node address, using the device identifier as the core. A video storage index is generated according to the key name rule of device ID_year_month_day, and the index is stored in the highly available distributed key-value index database Etcd. This index rule can achieve second-level location by device and by date. When querying a target video, it is not necessary to traverse the entire database; only the index item corresponding to the date needs to be matched to quickly obtain the storage address of all target segments, which greatly improves the efficiency of subsequent playback queries. At the same time, the index is uniquely bound to the device, which can clearly distinguish the video data of different users, avoid data confusion, and ensure clear ownership and isolation of user data.

[0065] In some embodiments, when a video playback request is received from a user terminal, the distributed container cloud platform retrieves the corresponding compressed video data according to the video storage index, decodes it, and sends it to the user terminal for playback. This includes: after receiving the playback request from the user terminal, the distributed container cloud platform extracts the device identifier and playback time range from the request, matches the corresponding storage node address according to the video storage index, retrieves the compressed video data for the corresponding time range, decodes it, transcodes it according to the requested bitrate, and sends it to the user terminal for playback.

[0066] This embodiment provides a detailed description of the implementation of the core step S103 (when a video playback request is received from a user terminal, the distributed container cloud platform retrieves the corresponding compressed video data according to the video storage index, decodes it, and sends it to the user terminal for playback). When a user initiates a video playback request via a smartphone, tablet, or other user terminal, the request message carries three core pieces of information: the user's authentication JWT token, the target pet feeder's device ID, and the user-selected playback start and end times. After receiving the request, the distributed container cloud platform first performs permission verification: verifying the legality and validity of the JWT token, then verifying whether the user account is bound to the target device ID. After confirming that the user has access to the device's video data, it proceeds to the subsequent processing flow. If the permission verification fails, it directly returns a permission error message, denying access and ensuring the security of the user's privacy data.

[0067] After successful permission verification, the index entry for the corresponding date is retrieved from the Etcd index database based on the target device ID and the date to which the playback time belongs. Then, all the fragment information in the index entry is traversed, and all target fragments that meet the condition of fragment start time < playback end time and fragment end time > playback start time are matched to obtain the storage node address of all target fragments. The compressed video data of the target fragments is retrieved from the corresponding storage nodes in turn, and the complete compressed video of the target time range is spliced ​​together.

[0068] After data retrieval is complete, the output bitrate and resolution are adaptively adjusted based on the screen resolution reported by the user terminal and the user terminal's current real-time network bandwidth. For example, if the user terminal is a 1080P resolution smartphone and the current available bandwidth is ≥2Mbps, then a 1080P / 2Mbps bitrate stream is output; if the current available bandwidth is only 1Mbps, then it is automatically transcoded to a 720P / 1Mbps bitrate stream to ensure smooth playback without stuttering. Finally, the transcoded video data is sent to the user terminal via the low-latency HLS streaming protocol. The user terminal can achieve simultaneous loading and playback, supporting common playback operations such as progress bar dragging, fast forward, and pause, completing the video playback process.

[0069] In some embodiments, the method further includes: controlling the distributed container cloud platform to store compressed video data, extracting video clips containing pet activities using a moving target detection algorithm, deleting static video clips that do not contain pet activities, updating the video storage index, and saving cloud storage space.

[0070] This embodiment adds an automatic cloud storage space optimization function to the core acquisition and storage process. In this embodiment, the distributed container cloud platform automatically performs invalid segment cleanup on the currently stored compressed video data every 2 hours. Users can also set it to automatically clean up every day at midnight in the APP. The specific cleanup process is as follows: perform inter-frame difference operation on consecutive video segments, calculate the proportion of difference pixels to the entire screen. If the proportion of difference pixels is less than 0.1%, the video segment is determined to be a static empty segment that does not contain pet activity; if the proportion of difference pixels is greater than 0.1%, the video segment is determined to contain pet activity and the segment is retained.

[0071] After all segments are evaluated, all static video segments that do not contain pet activity are deleted in batches. Then, the video storage index for the corresponding device is updated, and the records of the deleted segments are removed from the index to ensure consistency between the index and the actual stored data, preventing index errors in subsequent queries. This function is enabled by default and can be manually disabled by the user in the app. Users can disable this function at any time if they need to store complete videos for the entire day.

[0072] The storage optimization scheme in this embodiment can significantly save unnecessary cloud storage space: In actual use scenarios, pets only spend less than 2 hours a day near the feeder to move around and eat, while the remaining 22 hours are empty segments without pet activity. This scheme can save more than 90% of invalid storage space and can extend the storage period of user's valid videos by more than 10 times under the same storage cost, which not only reduces the storage cost of cloud service providers but also improves the user experience.

[0073] In some embodiments, the method further includes: comparing the acquired infrared video footage with pre-stored facial features of authorized personnel, and generating an alarm message when an unauthorized person appears in the footage, and sending the alarm message along with the real-time captured footage to the user terminal.

[0074] This embodiment adds an unauthorized intrusion detection and alarm function to the core data acquisition and storage process. In this embodiment, after the user completes the account binding of the pet feeder, he / she can upload the frontal face photos of all authorized personnel (pet owners, family members, regular visitors, etc.) in advance in the supporting APP. The system automatically extracts the 512-dimensional feature vector of each face based on the ArcFace face recognition algorithm and stores them in the feature library of the pet feeder and the feature library of the distributed container cloud platform respectively, thus completing the authorization information configuration.

[0075] The local controller of the pet feeder performs face detection on the currently acquired infrared video frame every 2 seconds. It uses the MTCNN algorithm to detect whether there is a face region in the image. When a face region is detected, the feature vector of the face is extracted and compared with the face features of authorized personnel stored locally. The cosine similarity of the features is calculated. If the highest similarity is lower than the preset threshold (the threshold is set to 80% in this embodiment), the person is determined to be an unauthorized person.

[0076] Upon detecting an unauthorized person, the pet feeder immediately captures a 1080P high-definition infrared image of the current frame, generating an intrusion alarm. This alarm is then pushed in real-time to the bound user terminal via a distributed container cloud platform. Users can directly access the live video to view the situation by clicking the push notification. Because this invention employs a high-definition infrared night vision acquisition solution, it can clearly capture facial contours and extract effective features even in completely dark nighttime environments. It enables 24 / 7 unauthorized personnel detection and alarm functions, combining pet care with home security. This aligns with the security needs of users placing the feeder at entrances, in yards, or entryways.

[0077] In some embodiments, this implementation is designed for multi-pet cohabitation family scenarios and creatively optimizes the existing feeding statistics function to solve the pain point that the original solution can only count the overall feeding data and cannot distinguish the feeding behavior of different individuals. In this embodiment, after the user completes the binding of the pet feeder device, the accompanying APP automatically guides the user to complete the configuration of multiple pet information: guiding the user to collect 3 to 5 infrared video screenshots of each pet from different angles (front, side, feeding posture). After the collection is completed, the pet feeder locally extracts a unique 512-dimensional individual feature vector for each pet based on the pre-trained improved PetReID pet individual weight recognition model. The feature vector is bound to the pet nickname input by the user and the preset feeding amount threshold, and a unique individual ID is assigned and stored in the local feature library. At the same time, it is synchronously backed up to the distributed container cloud platform.

[0078] When the distributed container cloud platform calls the target detection model to process infrared video daily, in addition to the original recognition of pet head and food bowl position, it extracts depth features for each detected pet target separately, compares the extracted features with the pre-stored individual feature library using cosine similarity, and only determines the highest matching result with similarity ≥85% as the corresponding individual, thus achieving accurate differentiation of multiple pet targets and avoiding feature confusion.

[0079] After individual pets are identified, the system combines the weight data of the food bowl collected by the 0.1g precision resistive weighing sensor built into the pet feeder to perform linked statistics: when a pet with a certain individual ID is detected with its head in the food bowl area for 3 consecutive frames, the initial food bowl weight at that time point is recorded; when the pet's head leaves the food bowl area for more than 5 minutes, the final food bowl weight at the end time point is recorded, and the specific weight of the pet's food intake for this feeding is calculated. At the same time, the system accumulates the total number of feedings, the average feeding time per feeding, and the total amount of food consumed by the pet for the day, and finally generates a daily feeding report for each individual pet and pushes it to the user's terminal.

[0080] Users can set personalized food intake thresholds for each pet. For example, an upper limit of 150g per day can be set for obese pets, and a reasonable range of 80-100g per day can be set for diabetic pets. When the system detects that the corresponding pet's daily food intake exceeds the threshold range, or that a sick pet's food intake is less than 60% of the threshold for two consecutive days, a personalized abnormality reminder will be generated immediately, clearly informing the user which pet has an abnormal eating habit. This helps users adjust the feeding amount in a timely manner and check the pet's health problems, solving the core pain point of multi-pet cohabitation families being unable to accurately grasp the eating situation of each pet.

[0081] In some embodiments, the original real-time transmission scheme is optimized for offline scenarios such as sudden network outages or power failures in the home, solving the problem of video data loss after network outages in the original scheme. In this embodiment, the local controller of the pet feeder has a built-in 32GB industrial-grade wide-temperature eMMC local storage chip, with at least 28GB of storage space reserved for offline caching. It can store up to 15 days of low frame rate infrared video data, meeting the caching needs of most short-duration network outage scenarios.

[0082] If the pet feeder sends probe packets to the distributed container cloud platform three times consecutively (with a 10-second interval between each time) without receiving a response, it is determined that the device is currently offline and automatically triggers the offline caching mode: maintaining the original infrared video acquisition and compression encoding process unchanged, storing the compressed video data in local eMMC at 15-minute blocks, adding a generation timestamp accurate to the second to each data block, and maintaining a temporary storage index in local non-volatile storage to record the storage address and time range of each data block. This index is written to storage immediately after each update to avoid index loss due to power failure.

[0083] The pet feeder maintains a network connectivity check every minute. Once the network connection is restored, it automatically triggers an adaptive breakpoint synchronization process: the synchronization priority is set to (nearest to oldest), prioritizing the synchronization of video data from the most recent 48 hours before synchronizing earlier historical data, to avoid long waiting times when users urgently need to watch recent videos; during the synchronization process, the bandwidth adaptive rule of Example 3 is used to adjust the size of the synchronization block and the transmission rate according to the currently available bandwidth, and a 16-bit CRC check information is added to each synchronization block. After the cloud side receives a block that has passed verification, it returns an acknowledgment frame. After receiving the acknowledgment frame, the device immediately deletes the temporary storage record of that block locally to release storage space.

[0084] If the network is interrupted again during the synchronization process, the current synchronization progress is automatically saved in the local temporary index. When the network is restored next time, the synchronization will continue directly from the point of interruption without having to start the transmission from the beginning. This avoids wasting bandwidth by repeatedly transmitting data and completely solves the problem of video data loss caused by sudden network outages and power failures, greatly improving the overall reliability of the solution.

[0085] In some embodiments, by extending the technical solution of the present invention to public scenarios where stray cats are fed at designated locations outdoors, targeted optimizations are made for the special needs of outdoor scenarios. First, the front-end acquisition module is adapted to the outdoor environment: the default brightness threshold for determining dark environments is adjusted to 10 lux to adapt to the infrared acquisition triggering requirements of low-brightness environments such as dawn, dusk, and cloudy days outdoors. The infrared sensor, photosensitive sensor, and main control module are all equipped with IP67-level waterproof and dustproof sealing covers to adapt to harsh environments such as outdoor rain, snow, high temperature, and low temperature. The device has a built-in 10000mAh high-capacity lithium battery paired with a 5W monocrystalline silicon solar charging panel to adapt to outdoor usage scenarios without fixed mains power access.

[0086] Secondly, local acquisition is further optimized for low power consumption: a secondary frame rate reduction rule is added based on the frame rate adaptive rule in Example 9. When no pixel change is detected for 10 consecutive frames, the acquisition frame rate is further reduced from 1 frame per second to 1 frame per 10 seconds, which greatly reduces the overall power consumption and enables normal operation for 7 consecutive days in rainy weather, extending the battery life.

[0087] Then, the distributed container cloud platform expands its functionality for outdoor stray cat feeding scenarios: the pre-trained target detection model adds three additional abnormal target categories: stray dogs, people tampering with the equipment, and large vehicles. When abnormal events such as people tampering with the equipment or large stray dogs breaking in and causing damage are detected, 1080P high-definition infrared images are immediately captured, and alarm information is generated and pushed to all authorized caring users who have bound the feeding point. The device has a built-in GPS positioning module, and the alarm information carries the device's current location coordinates, making it convenient for users to arrive at the scene in time or to retrieve the device if it is stolen.

[0088] Meanwhile, the cloud-based system adds a multi-user permission group management function, allowing multiple caring users to be bound to the same feeding point device. Administrators can set different permissions: administrators can modify configurations and remove members, while ordinary members can view videos, receive alarms, and view feeding statistics, solving the information sharing needs of multiple people feeding stray cats together. The cloud-based system automatically counts the number of visits and the cumulative amount of food consumed by different stray cats, making it convenient for caring users to keep track of the health status of stray cats, flexibly adjust the amount of food, and avoid environmental pollution and health problems caused by overfeeding, filling the technological gap in intelligent care in outdoor public feeding scenarios.

[0089] In some embodiments, this implementation provides a privacy-enhanced zero-trust cloud storage solution for users with high privacy requirements, addressing the risk of easy leakage of original videos in existing cloud storage solutions. After the user binds the pet feeder device for the first time, the system guides the user to enable the privacy-enhanced mode. Once enabled, an asymmetric encryption key pair is generated locally on the user's terminal based on the SM2 national cryptographic algorithm: the private key is stored only in the encrypted storage space on the user's terminal and is not uploaded to any cloud server or stored on the device. The public key is burned into the local one-time programmable storage area of ​​the pet feeder, completing the key configuration.

[0090] After the pet feeder locally compresses and encodes the infrared video, it employs a combination of symmetric and asymmetric encryption for edge-side encryption of the video data: First, a one-time symmetric encryption key is randomly generated and used to encrypt the entire compressed video data block. Then, the one-time symmetric key is encrypted using a pre-stored user public key. Finally, the encrypted symmetric key and the ciphertext video are concatenated into a complete encrypted data block, which is then transmitted to a distributed container cloud platform for storage. The entire encryption process is completed locally on the pet feeder. The cloud can only store the encrypted ciphertext data and cannot access any plaintext video information. Furthermore, the cloud lacks a private key to decrypt the content, achieving true zero-trust cloud storage. Even if the cloud server is illegally accessed, the user's original video data will not be leaked.

[0091] When a user initiates a video playback request, the distributed container cloud platform directly sends the encrypted video data to the user terminal. The entire decryption process is completed locally on the user terminal: the user terminal uses the locally stored private key to decrypt and obtain a one-time symmetric key, and then uses the symmetric key to decrypt and obtain the plaintext video data for playback. The original plaintext video will not leak out of the user terminal, fully protecting user privacy.

[0092] For AI analysis tasks that require generating structured data such as feeding statistics, this embodiment adopts an edge-side inference + only uploading structured results solution: all AI object detection and behavior recognition inference are completed on the local edge of the pet feeder, and only the structured statistical results such as the number of feedings and the amount of food consumed are uploaded to the cloud. The original video is always stored in encrypted form on the cloud, without leaking any original video privacy, which fully meets the usage requirements of users with high privacy needs.

[0093] The above description is only a preferred embodiment of the present invention and does not limit the scope of patent protection of the present invention. Any equivalent structural transformations made based on the content of the present invention specification, or direct or indirect applications in related technical fields, should be included within the scope of patent protection of the present invention.

[0094] Please see Figure 3 As shown, Figure 3 This is a schematic diagram of the infrared night vision video acquisition and cloud storage system 200 for a video-enabled pet feeder provided in this application embodiment. The infrared night vision video acquisition and cloud storage system 200 for the video-enabled pet feeder is used to execute the steps of the infrared night vision video acquisition and cloud storage method for the video-enabled pet feeder shown in the above embodiments. The infrared night vision video acquisition and cloud storage system 200 for the video-enabled pet feeder can be a single server or a server cluster, or it can be a terminal, such as a handheld terminal, a laptop computer, a wearable device, or a robot.

[0095] like Figure 3 As shown, the infrared night vision video acquisition and cloud storage system 200 for the video pet feeder includes: The image acquisition unit 201 is used to acquire infrared video images of pet activities in real time in dark environments through the infrared sensor built into the pet feeder; the acquired infrared video images are compressed in real time using a preset advanced video encoding standard to generate compressed video data. The video transmission unit 202 is used to transmit the compressed video data to a preset distributed container cloud platform; the distributed container cloud platform receives and stores the compressed video data, and establishes a video storage index that is uniquely associated with the corresponding pet feeder; The decoding and playback unit 203 is used to retrieve the corresponding compressed video data according to the video storage index when a video playback request is received from a user terminal. After decoding, the data is sent to the user terminal for playback. The distributed container cloud platform is controlled to call a pre-trained pet target detection model to detect the stored infrared video data frame by frame, identify pet eating behavior, count the number of times the pet eats per day and the duration of each feeding, and push the statistical results to the bound user terminal. The pet feeder is also controlled to perform frame difference detection on the continuously collected infrared video frames. If no pixel change is detected for three consecutive frames, it is determined that there is no pet activity, and the infrared video acquisition frame rate is reduced to one frame per second. When a pixel change is detected and it is determined that there is pet activity, the original acquisition frame rate is restored.

[0096] In some embodiments, the real-time acquisition of infrared video footage of pet activities in a dark environment using the built-in infrared sensor of the pet feeder includes: detecting ambient light brightness using the built-in photosensor of the pet feeder, determining an environment where the ambient light brightness is lower than a preset brightness threshold as a dark environment, activating the built-in infrared sensor to acquire infrared video footage, and outputting the infrared video footage of pet activities at a fixed frame rate.

[0097] In some embodiments, the step of performing real-time compression processing on the acquired infrared video images using a preset advanced video coding standard to generate compressed video data includes: identifying static background areas and dynamic pet areas in the acquired infrared video images, using low bitrate encoding for the static background areas and high bitrate encoding for the dynamic pet areas, and splicing all coded blocks to generate compressed video data.

[0098] In some embodiments, transmitting the compressed video data to a preset distributed container cloud platform includes: detecting the real-time transmission bandwidth of the network currently accessed by the pet feeder, adjusting the sending frame rate of the compressed video data according to the real-time transmission bandwidth, adding verification information to the compressed video data in blocks, and then sending them sequentially to the distributed container cloud platform.

[0099] In some embodiments, receiving and storing the compressed video data through a distributed container cloud platform and establishing a video storage index uniquely associated with the corresponding pet feeder includes: after receiving the compressed video data on the distributed container cloud platform, extracting the unique device identifier of the sending pet feeder, recording the generation time and storage node address of the compressed video data, binding the generation time and storage node address with the device identifier as the core, generating a video storage index, and storing it in the index database.

[0100] In some embodiments, when a video playback request is received from a user terminal, the distributed container cloud platform retrieves the corresponding compressed video data according to the video storage index, decodes it, and sends it to the user terminal for playback. This includes: after receiving the playback request from the user terminal, the distributed container cloud platform extracts the device identifier and playback time range from the request, matches the corresponding storage node address according to the video storage index, retrieves the compressed video data for the corresponding time range, decodes it, transcodes it according to the requested bitrate, and sends it to the user terminal for playback.

[0101] In some embodiments, the method further includes: controlling the distributed container cloud platform to store compressed video data, extracting video clips containing pet activities using a moving target detection algorithm, deleting static video clips that do not contain pet activities, updating the video storage index, and saving cloud storage space.

[0102] In some embodiments, the method further includes: comparing the acquired infrared video footage with pre-stored facial features of authorized personnel, and generating an alarm message when an unauthorized person appears in the footage, and sending the alarm message along with the real-time captured footage to the user terminal.

[0103] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the infrared night vision video acquisition and cloud storage system and each module of the video pet feeder described above can be referred to the corresponding content in the various embodiments of the infrared night vision video acquisition and cloud storage method of the video pet feeder, and will not be repeated here.

[0104] The aforementioned method for infrared night vision video acquisition and cloud storage of video-enabled pet feeders can be implemented as a computer program, which can be used in various ways, such as... Figure 3 It runs on the device shown.

[0105] Please see Figure 4 , Figure 4 This is a schematic block diagram of the structure of a pet feeder provided in an embodiment of this application. The pet feeder includes a processor, a memory, and a network interface connected via a device bus, wherein the memory may include a storage medium and internal memory.

[0106] The storage medium can store operating devices and computer programs. The computer program includes program instructions that, when executed, cause the processor to perform any method for infrared night vision video acquisition and cloud storage of video-enabled pet feeders.

[0107] The processor provides computing and control capabilities to support the operation of the entire pet feeder.

[0108] The internal memory provides an environment for the execution of computer programs in non-volatile storage media. When the computer program is executed by the processor, it enables the processor to execute any infrared night vision video acquisition and cloud storage method for video pet feeders.

[0109] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the terminal to which the present application is applied. A specific pet feeder may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0110] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0111] In one embodiment, the processor is configured to run a computer program stored in memory to perform the following steps: The built-in infrared sensor in the pet feeder captures real-time infrared video footage of the pet's activities in dark environments; the captured infrared video footage is then compressed in real-time using a preset advanced video encoding standard to generate compressed video data. The compressed video data is transmitted to a pre-set distributed container cloud platform; the compressed video data is received and stored through the distributed container cloud platform, and a video storage index uniquely associated with the corresponding pet feeder is established. When a video playback request is received from a user terminal, the distributed container cloud platform retrieves the corresponding compressed video data based on the video storage index, decodes it, and sends it to the user terminal for playback. It also controls the distributed container cloud platform to call a pre-trained pet target detection model to detect the stored infrared video data frame by frame, identify pet eating behavior, and count the number of times the pet eats daily and the duration of each feeding. The statistical results are then pushed to the bound user terminal. Furthermore, the pet feeder is controlled to perform frame difference detection on the continuously collected infrared video frames. If no pixel change is detected for three consecutive frames, it is determined that there is no pet activity, and the infrared video acquisition frame rate is reduced to one frame per second. When a pixel change is detected and pet activity is determined, the original acquisition frame rate is restored.

[0112] In some embodiments, the real-time acquisition of infrared video footage of pet activities in a dark environment using the built-in infrared sensor of the pet feeder includes: detecting ambient light brightness using the built-in photosensor of the pet feeder, determining an environment where the ambient light brightness is lower than a preset brightness threshold as a dark environment, activating the built-in infrared sensor to acquire infrared video footage, and outputting the infrared video footage of pet activities at a fixed frame rate.

[0113] In some embodiments, the step of performing real-time compression processing on the acquired infrared video images using a preset advanced video coding standard to generate compressed video data includes: identifying static background areas and dynamic pet areas in the acquired infrared video images, using low bitrate encoding for the static background areas and high bitrate encoding for the dynamic pet areas, and splicing all coded blocks to generate compressed video data.

[0114] In some embodiments, transmitting the compressed video data to a preset distributed container cloud platform includes: detecting the real-time transmission bandwidth of the network currently accessed by the pet feeder, adjusting the sending frame rate of the compressed video data according to the real-time transmission bandwidth, adding verification information to the compressed video data in blocks, and then sending them sequentially to the distributed container cloud platform.

[0115] In some embodiments, receiving and storing the compressed video data through a distributed container cloud platform and establishing a video storage index uniquely associated with the corresponding pet feeder includes: after receiving the compressed video data on the distributed container cloud platform, extracting the unique device identifier of the sending pet feeder, recording the generation time and storage node address of the compressed video data, binding the generation time and storage node address with the device identifier as the core, generating a video storage index, and storing it in the index database.

[0116] In some embodiments, when a video playback request is received from a user terminal, the distributed container cloud platform retrieves the corresponding compressed video data according to the video storage index, decodes it, and sends it to the user terminal for playback. This includes: after receiving the playback request from the user terminal, the distributed container cloud platform extracts the device identifier and playback time range from the request, matches the corresponding storage node address according to the video storage index, retrieves the compressed video data for the corresponding time range, decodes it, transcodes it according to the requested bitrate, and sends it to the user terminal for playback.

[0117] In some embodiments, the method further includes: controlling the distributed container cloud platform to store compressed video data, extracting video clips containing pet activities using a moving target detection algorithm, deleting static video clips that do not contain pet activities, updating the video storage index, and saving cloud storage space.

[0118] In some embodiments, the method further includes: comparing the acquired infrared video footage with pre-stored facial features of authorized personnel, and generating an alarm message when an unauthorized person appears in the footage, and sending the alarm message along with the real-time captured footage to the user terminal.

[0119] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to implement the steps of the infrared night vision video acquisition and cloud storage method for a video-enabled pet feeder as provided in any embodiment of this application.

[0120] The computer-readable storage medium can be the internal storage unit of the pet feeder described in the foregoing embodiments, such as the hard drive or memory of the pet feeder. Alternatively, the computer-readable storage medium can be an external storage device of the pet feeder, such as a plug-in hard drive, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the pet feeder.

[0121] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for infrared night vision video acquisition and cloud storage of a video-enabled pet feeder, characterized in that, Applications in pet feeders; including: The built-in infrared sensor in the pet feeder captures real-time infrared video footage of the pet's activities in dark environments; the captured infrared video footage is then compressed in real-time using a preset advanced video encoding standard to generate compressed video data. The compressed video data is transmitted to a pre-set distributed container cloud platform; the compressed video data is received and stored through the distributed container cloud platform, and a video storage index uniquely associated with the corresponding pet feeder is established. When a video playback request is received from a user terminal, the distributed container cloud platform retrieves the corresponding compressed video data based on the video storage index, decodes it, and sends it to the user terminal for playback. It also controls the distributed container cloud platform to call a pre-trained pet target detection model to detect the stored infrared video data frame by frame, identify pet eating behavior, and count the number of times the pet eats daily and the duration of each feeding. The statistical results are then pushed to the bound user terminal. Furthermore, the pet feeder is controlled to perform frame difference detection on the continuously collected infrared video frames. If no pixel change is detected for three consecutive frames, it is determined that there is no pet activity, and the infrared video acquisition frame rate is reduced to one frame per second. When a pixel change is detected and pet activity is determined, the original acquisition frame rate is restored.

2. The infrared night vision video acquisition and cloud storage method for the video-enabled pet feeder according to claim 1, characterized in that, The method of using the built-in infrared sensor of the pet feeder to capture real-time infrared video footage of pet activities in dark environments includes: The pet feeder uses a built-in photosensor to detect ambient light levels. If the ambient light level is below a preset threshold, it is classified as a dark environment. The built-in infrared sensor is then activated to capture infrared video footage, which is then output at a fixed frame rate.

3. The infrared night vision video acquisition and cloud storage method for the video-enabled pet feeder according to claim 1, characterized in that, The process of real-time compression of the acquired infrared video images using a preset advanced video coding standard to generate compressed video data includes: The system identifies static background areas and dynamic pet areas in the captured infrared video footage. It uses low bitrate encoding for static background areas and high bitrate encoding for dynamic pet areas, and then stitches together all the encoded blocks to generate compressed video data.

4. The infrared night vision video acquisition and cloud storage method for the video-enabled pet feeder according to claim 1, characterized in that, The step of transmitting the compressed video data to a preset distributed container cloud platform includes: The system detects the real-time transmission bandwidth of the network currently connected to the pet feeder, adjusts the frame rate of the compressed video data based on the real-time transmission bandwidth, adds verification information to the compressed video data in blocks, and then sends them sequentially to the distributed container cloud platform.

5. The infrared night vision video acquisition and cloud storage method for the video-enabled pet feeder according to claim 1, characterized in that, The step of receiving and storing the compressed video data through a distributed container cloud platform, and establishing a video storage index uniquely associated with the corresponding pet feeder, includes: After receiving the compressed video data on the distributed container cloud platform, the unique device identifier of the pet feeder at the sending end is extracted, the generation time and storage node address of the compressed video data are recorded, and the generation time and storage node address are bound together with the device identifier as the core to generate a video storage index and store it in the index database.

6. The infrared night vision video acquisition and cloud storage method for the video-enabled pet feeder according to claim 1, characterized in that, When a video playback request is received from a user terminal, the distributed container cloud platform retrieves the corresponding compressed video data according to the video storage index, decodes it, and sends it to the user terminal for playback, including: After receiving a playback request from a user terminal, the distributed container cloud platform extracts the device identifier and playback time range from the request, matches the corresponding storage node address according to the video storage index, retrieves the compressed video data for the corresponding time range, decodes it, transcodes it according to the requested bitrate, and sends it to the user terminal for playback.

7. The infrared night vision video acquisition and cloud storage method for the video-enabled pet feeder according to claim 1, characterized in that, The method further includes: Control the distributed container cloud platform to store compressed video data, extract video clips containing pet activities through moving object detection algorithms, delete static video clips that do not contain pet activities, update the video storage index, and save cloud storage space.

8. The infrared night vision video acquisition and cloud storage method for the video-enabled pet feeder according to claim 1, characterized in that, The method further includes: The system compares the captured infrared video footage with pre-stored facial features of authorized personnel. When an unauthorized person is detected in the footage, an alarm is generated and sent to the user terminal along with the real-time captured image.

Images