AI decision-based outdoor bird feeder full-automatic feeding control method and bird feeder

Abstract

The application discloses an outdoor bird feeder full-automatic feeding control method based on AI decision and a bird feeder, relates to the technical field of data analysis, and comprises the following steps: collecting visual data of visited birds, outdoor environment data of an environment where the bird feeder is located, and current seasonal time sequence data; acquiring bird species and body size information and a visiting frequency feature, and determining environment state features and seasonal features according to the outdoor environment data and the seasonal time sequence data; inputting the bird species and body size information, the visiting frequency feature, the environment state features and the seasonal features into an AI decision model, outputting a feeding strategy from the AI decision model; and controlling a feeding mechanism of the bird feeder to execute feed feeding according to the feeding strategy. The application dynamically adjusts the frequency and single feeding amount by using an adjustment coefficient and an incremental value based on the season, the weather and the number of clusters, and cooperates with feed preference matching, automatic hibernation in bad weather and feed replenishment reminding, so that full-automatic accurate feeding is realized, and waste is reduced.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, and in particular to a fully automatic feeding control method and bird feeder for outdoor bird feeders based on AI decision-making. Background Technology

[0002] Currently, most outdoor bird feeders on the market focus on feed storage and simple mechanical dispensing in their design, with feeding methods mostly adopting fixed patterns of timed and quantitative or passive replenishment. For example, some bird feeders use simple timers to set a fixed amount of feed to be dispensed at a fixed time each day, completely disregarding the actual species and number of visiting birds and changes in external environmental conditions; others use gravity-fed designs, with feed continuously exposed at the feeding opening, allowing birds to eat at any time. However, this open-feeding method is highly susceptible to weather conditions, causing the feed to become damp and spoil, and it also makes it impossible to control the amount of feed in any way.

[0003] In practical applications, the existing outdoor bird feeders have several technical shortcomings. First, the feeding strategy is completely fixed and cannot be adaptively adjusted according to the species and size differences of the visiting birds. Birds of different sizes have drastically different feeding habits. Small songbirds are accustomed to frequent small meals, while medium and large birds tend to eat large amounts at once with longer intervals. Fixed feeding amounts and frequencies cannot simultaneously meet the needs of both types of birds, often resulting in feed waste or insufficient supply. Existing bird feeders lack the ability to sense and respond to changes in the outdoor environment and seasons. Environmental factors such as temperature, rainfall, and light directly affect the activity level and energy requirements of birds. For example, birds' energy requirements increase significantly during the brooding period and in cold seasons, while high temperatures or rainy weather can cause feed to spoil quickly. However, existing products cannot automatically adjust their feeding strategies based on these dynamic changes, resulting in low feeding accuracy. Moreover, existing bird feeders lack a real-time statistical and feedback mechanism for visit frequency and flock size. When the frequency of bird visits decreases or the number of flocks increases or decreases, the feeding amount cannot be adjusted accordingly, further exacerbating the problem of feed waste. Summary of the Invention

[0004] The main objective of this invention is to provide an AI-based decision-making fully automatic feeding control method and bird feeder for outdoor bird feeders. This invention aims to solve the technical problems of existing outdoor bird feeders having fixed feeding modes, failing to automatically adjust strategies based on visiting bird species, frequency, and seasonal environmental changes, resulting in serious feed waste, low feeding accuracy, and difficulty in dynamically meeting the habits and energy needs of different birds.

[0005] To achieve the above objectives, this invention proposes a fully automatic feeding control method for outdoor bird feeders based on AI decision-making, comprising the following steps: Visual data of visiting birds, outdoor environmental data of the environment where the bird feeder is located, and time-series data of the current season are collected. Visual data is used to identify bird species, size information, and visit frequency characteristics. Environmental status and seasonal characteristics are determined based on outdoor environmental data and seasonal time series data. Bird species and size information, visit frequency characteristics, environmental status characteristics, and seasonal characteristics are input into the AI ​​decision model, which then outputs a feeding strategy. The feeding strategy includes feeding mode, feeding frequency, and total feeding amount. The feeding mode includes a first feeding mode or a second feeding mode. The number of feedings under the first feeding mode is higher than the number of feedings under the second feeding mode. The amount of food given in a single feeding under the first feeding mode is less than the amount of food given in a single feeding under the second feeding mode. The feeding interval under the first feeding mode is shorter than the feeding interval under the second feeding mode. Based on the feeding strategy, control the feeding mechanism of the bird feeder to deliver feed.

[0006] Furthermore, visual data of visiting birds, outdoor environmental data of the bird feeder's location, and current seasonal time-series data were collected, specifically including: Visual data is collected through cameras and edge computing AI chips to identify the species, size and number of visiting birds, and to count the frequency of visits by various bird species per unit time. Outdoor environmental data is collected through temperature and humidity sensors, raindrop sensors, and light sensors. The outdoor environmental data includes ambient temperature, relative humidity, rainfall intensity, and light intensity. The bird feeder uses a built-in clock module to obtain seasonal time data to determine the current season and special periods, including at least one of the following: brooding period, molting period, and daily peak foraging times at dawn and / or dusk.

[0007] Furthermore, visual data is used to identify bird species, body size, and visit frequency characteristics. Determining environmental and seasonal characteristics based on outdoor environmental data and seasonal time-series data also includes: The collected visual data is cleaned to remove misidentified non-bird objects, abnormal visit data, and sensor malfunction data. The dominant bird species, body size classification, daily peak visit times, and average flock size within the statistical time window are extracted as bird species and body size information and visit frequency characteristics. The dominant bird species are those that are identified and recorded at the bird feeder within the statistical time window with the highest visit frequency or the most total visits. The energy requirements of birds are determined according to the current season, and the environmental conditions are divided into three levels: suitable for feeding, cautious feeding, and prohibited feeding, based on ambient temperature and rainfall intensity. The prohibited feeding level is used to trigger the device to enter a low-power sleep mode and stop feeding, while the cautious feeding level is used to trigger fine-tuning operations to reduce the feeding frequency and / or the amount of food fed at one time.

[0008] Furthermore, the AI ​​decision-making model is a weighted multi-factor fusion model, whose input factors include bird species and size factors, visit frequency factors, seasonal characteristics factors, outdoor environmental factors, and bird habit factors.

[0009] Furthermore, in the feeding strategy generation step, the AI ​​decision-making model makes a judgment based on the dominant bird species identified in the past preset time period: If the bird is identified as a small songbird and its visit frequency is higher than the visit frequency threshold, then the first feeding mode is selected. If the bird is identified as a medium to large-sized bird and its visit frequency is lower than the visit frequency threshold, then the second feeding mode is selected. If the bird is identified as a small songbird and its visit frequency is below the visit frequency threshold, or as a medium or large bird and its visit frequency is above the visit frequency threshold, it will enter hibernation mode.

[0010] Furthermore, the single feeding amount for the first feeding mode is 0.2-0.5g, and the feeding interval is 20-60 minutes; the single feeding amount for the second feeding mode is 1.5-5g, and the feeding interval is 2-6 hours.

[0011] Furthermore, the feeding strategy generation step also includes: after determining the feeding mode, multiplying the preset baseline feeding amount corresponding to the feeding mode by the adjustment coefficient corresponding to the seasonal characteristics and / or environmental state characteristics, and adding the incremental value determined by the average cluster size, thereby calculating the dynamically fine-tuned actual feeding frequency and / or single feeding amount; wherein, when the seasonal characteristics are the brooding period or the low temperature period, the adjustment coefficient is greater than 1; when the environmental state characteristics are the cautious feeding level or the seasonal characteristics are the high temperature period, the adjustment coefficient is less than 1; and, for each additional bird in the average cluster size, the single feeding amount is increased by a preset weight; the dynamically fine-tuned actual single feeding amount and feeding frequency are still limited to the feeding frequency range and single feeding amount range specified by the feeding mode.

[0012] Furthermore, the feeding execution steps include: Based on the preset feed preferences corresponding to the identified bird species, select the feed stored in layers in the bird feeder for feeding; When the environmental conditions indicate that feeding is prohibited, the control device enters a low-power sleep mode. When the remaining material in the hopper is detected to be below the threshold, a replenishment reminder is sent.

[0013] Furthermore, the method also includes iterative optimization steps: Using a 7-day iteration cycle, calculate at least one of the following based on historical feeding data: feed surplus rate, bird visitation satisfaction, and feed waste rate, to evaluate the feeding effect; The weight coefficients of each input factor in the AI ​​decision-making model are automatically adjusted based on the evaluation results of the feeding effect.

[0014] The present invention also proposes a bird feeder, comprising: A camera module is used to collect visual data of visiting birds; The environmental sensing module is used to collect outdoor environmental data of the bird feeder's location. The clock module is used to provide time-series data for the current season; The control unit is electrically connected to the camera module, the environmental sensing module and the clock module respectively. The control unit is configured to execute a fully automatic feeding control method for outdoor bird feeders based on AI decision-making. The feeding mechanism is electrically connected to the control unit and is used to perform feed feeding under the control of the control unit.

[0015] This invention integrates visual recognition, environmental sensing, and seasonal time-series data. Through a weighted multi-factor AI decision-making model, it automatically selects a high-frequency, low-volume or low-frequency, high-volume feeding mode based on the dominant bird species' size. It also dynamically fine-tunes the frequency and single feeding amount based on the season, weather, and flock size using adjustment coefficients and incremental values. Combined with feed preference matching, automatic hibernation in inclement weather, and supplementary feeding reminders, it achieves fully automatic and precise feeding, reducing waste. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the fully automatic feeding control method for outdoor bird feeders based on AI decision-making according to the present invention. Figure 2 This is a schematic diagram of the bird feeder of the present invention; The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of the present invention and are not intended to limit the present invention.

[0020] To better understand the technical solution of the present invention, a detailed description will be provided below in conjunction with the accompanying drawings and specific embodiments.

[0021] like Figure 1 As shown, Figure 1 This is a flowchart illustrating a fully automated feeding control method for outdoor bird feeders based on AI decision-making.

[0022] A fully automatic feeding control method for outdoor bird feeders based on AI decision-making, characterized by the following steps: S10 collects visual data of visiting birds, outdoor environmental data of the environment where the bird feeder is located, and current seasonal time data. S20: Identify visual data, obtain information on bird species and size, as well as visit frequency characteristics, and determine environmental status characteristics and seasonal characteristics based on outdoor environmental data and seasonal time series data; S30: Input bird species and size information, visit frequency characteristics, environmental status characteristics, and seasonal characteristics into the AI ​​decision model, and the AI ​​decision model outputs a feeding strategy. The feeding strategy includes feeding mode, feeding frequency, and total feeding amount. The feeding mode includes a first feeding mode or a second feeding mode. The number of feedings in the first feeding mode is higher than the number of feedings in the second feeding mode. The amount of food given in a single feeding in the first feeding mode is less than the amount of food given in a single feeding in the second feeding mode. The feeding time interval in the first feeding mode is less than the feeding time interval in the second feeding mode. S40 controls the feeding mechanism of the bird feeder to deliver feed according to the feeding strategy.

[0023] This invention provides a fully automated feeding control method for outdoor bird feeders based on AI decision-making. This invention integrates visual recognition, environmental sensing, and seasonal time-series data. Through a weighted multi-factor AI decision-making model, it automatically selects a high-frequency, low-volume or low-frequency, high-volume feeding mode based on the dominant bird species' size. Furthermore, it dynamically fine-tunes the frequency and single-feed amount based on season, weather, and flock size using adjustment coefficients and incremental values. Combined with feed preference matching, automatic hibernation in inclement weather, and supplementary feeding reminders, it achieves fully automated and precise feeding, significantly reducing waste. The complete technical process will be described in detail below.

[0024] S10 collects visual data of visiting birds, outdoor environmental data of the environment where the bird feeder is located, and current seasonal time data. In this embodiment, the first step of the method is multi-source data acquisition. The system acquires three types of core data in the following ways.

[0025] Specifically, in terms of visual data acquisition, visual data is collected through cameras and edge computing AI chips to identify the species, size, and number of visiting birds, and to count the frequency of visits by various bird species per unit time: the camera captures video streams at a rate of 25 frames per second, and the edge computing AI chip is equipped with a lightweight target detection model to infer each frame of the image locally, identify the bird targets in the picture, classify them by species, and estimate their size. At the same time, the chip's built-in counter and timer accumulate the number of times each identified species appears per unit time, thereby calculating the visit frequency in real time. This process of identification and statistics is completed on the device side, which greatly reduces the transmission latency and bandwidth costs of uploading video data to the cloud, while avoiding the response lag caused by network instability. This allows the bird feeder to maintain real-time and autonomous intelligent decision-making capabilities even in environments without a network.

[0026] Furthermore, in terms of environmental data acquisition, outdoor environmental data is collected through temperature and humidity sensors, raindrop sensors, and light sensors. The outdoor environmental data includes ambient temperature, relative humidity, rainfall intensity, and light intensity. The temperature and humidity sensors are digital sensors that read the ambient temperature and relative humidity once per minute. The raindrop sensors determine whether there is rainfall and its intensity by detecting the resistance changes caused by raindrops on their sensing panel. The light sensor measures the ambient illuminance to help determine the weather conditions and day-night changes. This data is summarized into an environmental state vector for subsequent decision-making. The acquisition of this data enables the bird feeder to perceive the external physical world and is a key basis for determining whether the current weather is unsuitable for feeding, such as heavy rain or high temperatures. This lays the data foundation for subsequent decisions on "suitable feeding," "cautious feeding," or "no feeding."

[0027] Furthermore, in terms of acquiring time-series data, seasonal time-series data is obtained through the clock module built into the bird feeder to determine the current season and special periods. Special periods include at least one of the following: the brooding period, the molting period, and the daily peak foraging periods at dawn and / or dusk. The clock module is a real-time clock chip with calendar functionality, capable of providing year, month, day, hour, minute, and second information. The system is pre-set with seasonal division rules, approximate calendar intervals for the brooding and molting periods of common birds in various regions, and dawn / dusk time algorithms based on sunrise and sunset times. When the clock data falls within the preset interval, the corresponding special period flag is activated. This data lays the foundation for subsequent fine-grained feeding adjustments based on the life cycle, enabling the feeding strategy to conform to the birds' internal biological rhythms. For example, during the brooding period, when parent birds need a lot of food to feed the chicks, the amount of food can be increased, and feeding can be completed in advance before the peak foraging period at dawn, thereby maximizing the satisfaction of the birds' actual needs.

[0028] S20: Identify visual data, obtain information on bird species and size, as well as visit frequency characteristics, and determine environmental status characteristics and seasonal characteristics based on outdoor environmental data and seasonal time series data; In this embodiment, the collected raw data needs to be cleaned and feature-engineered. The specific processing flow is as follows: Specifically, the collected visual data undergoes data cleaning to remove misidentified non-bird objects, abnormal visit data, and sensor malfunction data. The system maintains a confidence threshold; objects with a confidence level below the threshold in the edge computing AI chip's recognition results are considered misidentified and discarded. The system also sets reasonable upper and lower limits for visit frequency, filtering out abnormal data points outside these limits. When the sensor returns an error code or a continuously unchanged value, it is marked as faulty data and excluded from subsequent calculations. This cleaning process ensures that the data used in subsequent analysis truly reflects the actual activity patterns of birds, avoiding incorrect feeding decisions made by the AI ​​decision-making model due to accidental recognition errors or sensor noise, thus ensuring the robustness and reliability of the decision-making process.

[0029] Furthermore, the dominant bird species, body size classification, daily peak visit times, and average flock size within the statistical time window are extracted as bird species and body size information and visit frequency characteristics. The dominant bird species are those identified and recorded at the bird feeder with the highest visit frequency or the most total visits within the statistical time window. The system maintains a sliding time window data buffer in memory, storing the bird species, number, and timestamp for each visit event. At each preset calculation cycle, the system scans all records in the buffer, summarizes the visit counts by bird species, and selects the species with the most visits as the dominant bird species. The dominant bird species were identified, and their size classification was obtained by referring to a table. The average number of flocks in all visits was calculated, and the peak time period was found by statistically analyzing the frequency distribution of visits over time. This feature extraction process compresses a massive amount of raw identification records into a few feature indicators with clear ecological significance. The setting of "dominant bird species" is to ensure that the feeding strategy mainly serves the most frequently visiting bird groups, achieving optimal resource allocation and avoiding the waste caused by adjusting the entire feeding system for occasionally visiting rare bird species. Size classification provides a direct basis for subsequent selection of feeding modes.

[0030] Furthermore, the system determines the bird's energy requirement level based on the current season and categorizes environmental conditions into three levels—suitable for feeding, cautious for feeding, and prohibited for feeding—based on ambient temperature and rainfall intensity. The prohibited feeding level triggers the device to enter a low-power sleep mode and stop feeding. The cautious feeding level triggers a fine-tuning operation to reduce feeding frequency and / or the amount of food given per feeding. The system has a pre-set level determination rule table that defines the levels corresponding to different combinations of temperature and rainfall intensity. For example, when the temperature is between 15℃ and 30℃ and there is no rainfall, it is determined to be "suitable for feeding"; when the temperature is above 35℃ or the rainfall intensity is light, it is determined to be "cautious for feeding"; and when the rainfall intensity reaches a certain level... When there is moderate to heavy rain or the temperature is below -10℃ or above 40℃, feeding is prohibited. The seasonal energy demand level is determined by looking up the month and special time period flags provided by the clock module. Discretizing continuous environmental values ​​into clear decision levels simplifies the input complexity of the AI ​​model and makes the decision logic more transparent and interpretable. The "No Feeding" level is used to trigger the device to enter a low-power sleep mode and stop feeding, effectively avoiding ineffective feeding and feed mold and rot caused by severe weather. The "Cautious Feeding" level is used to trigger fine-tuning operations to reduce the feeding frequency and / or the amount of feed per feeding, minimizing waste while ensuring food supply.

[0031] S30: Input bird species and size information, visit frequency characteristics, environmental status characteristics, and seasonal characteristics into the AI ​​decision model, and the AI ​​decision model outputs a feeding strategy. The feeding strategy includes feeding mode, feeding frequency, and total feeding amount. The feeding mode includes a first feeding mode or a second feeding mode. The number of feedings in the first feeding mode is higher than the number of feedings in the second feeding mode. The amount of food given in a single feeding in the first feeding mode is less than the amount of food given in a single feeding in the second feeding mode. The feeding time interval in the first feeding mode is less than the feeding time interval in the second feeding mode. In this embodiment, the bird species and size information, visit frequency characteristics, environmental state characteristics, and seasonal characteristics obtained from the first three layers of processing are input into the AI ​​decision-making model, including: The AI ​​decision-making model is a weighted multi-factor fusion model. Its input factors include bird species and size, visit frequency, seasonal characteristics, outdoor environment, and bird behavior. The model assigns a weight coefficient to each input factor. Each factor is quantified into a standardized value between 0 and 1, multiplied by its corresponding weight, and the sum is used to obtain the decision score. The bird behavior factor is derived from a pre-set bird behavior knowledge base, which stores information such as typical feeding frequencies and food preferences for common bird species. When a dominant bird species is identified, the system automatically retrieves the corresponding behavior parameters from the knowledge base as the input value for that factor. By employing a weighted multi-factor fusion model instead of a fixed-rule engine, the system can comprehensively consider multiple factors and find a globally optimal compromise solution through weight allocation when facing complex scenarios with conflicting factors. Furthermore, the introduction of bird behavior factors ensures that the feeding strategy not only relies on real-time observation data but also incorporates expert experience and knowledge, making the decision more scientific and rational.

[0032] Furthermore, within the framework of the weighted multi-factor fusion model, strategy generation first involves determining the feeding mode. The AI ​​decision model determines the dominant bird species identified over a preset time period: if the bird is identified as a small songbird and its visit frequency is higher than the visit frequency threshold, the first feeding mode is selected; if the bird is identified as a medium to large bird and its visit frequency is lower than the visit frequency threshold, the second feeding mode is selected; if the bird is identified as a small songbird and its visit frequency is lower than the visit frequency threshold, or a medium to large bird and its visit frequency is higher than the visit frequency threshold, the bird enters a dormant mode. The first feeding mode involves a single feeding amount of 0.2-0.5g and a feeding interval of 20-60 minutes; the second feeding mode involves a single feeding amount of 1.5-5g and a feeding interval of 2-6 hours. In practice, the visit frequency threshold can be preset to, for example, 3 times per hour. The system extracts the dominant bird species and their average hourly visit frequency over the past 24 hours, and performs a lookup match in a preset 2×2 decision matrix based on the bird species size classification and whether the visit frequency crosses the threshold, outputting the corresponding mode selection instruction. This hierarchical judgment logic ensures that the basic feeding mode is precisely matched with the bird's size and actual activity level. The high-frequency, small-quantity characteristic of the first feeding mode simulates the natural feeding habit of small birds that eat small meals frequently, which can effectively avoid waste caused by large particles or large amounts of feed piling up. The low-frequency, large-quantity characteristic of the second feeding mode meets the needs of larger birds that eat more at a time. Introducing situations that do not conform to typical ecological laws into the hibernation mode provides a safety net mechanism for the system in abnormal situations.

[0033] Furthermore, after determining the feeding mode, the system enters the dynamic fine-tuning stage. After determining the feeding mode, the preset baseline feeding amount corresponding to the feeding mode is multiplied by the adjustment coefficient corresponding to the seasonal characteristics and / or environmental conditions, and then superimposed with the incremental value determined by the average cluster size. This calculates the dynamically fine-tuned actual feeding frequency and / or single feeding amount. Specifically, when the seasonal characteristics are the brooding period or low temperature period, the adjustment coefficient is greater than 1; when the environmental conditions are the cautious feeding level or the seasonal characteristics are the high temperature period, the adjustment coefficient is less than 1. Additionally, for each additional bird in the average cluster size, the single feeding amount increases by a preset weight. The actual single feeding amount and feeding frequency after dynamic fine-tuning are still limited to the feeding frequency range and single feeding amount range specified by the feeding mode. The specific calculation formula is: the actual single feeding amount equals the baseline feeding amount multiplied by the adjustment coefficient plus the cluster size multiplied by the single-bird increment. The result is compared with the minimum feeding amount of the mode by taking the larger function, and then compared with the maximum feeding amount of the mode by taking the smaller function to obtain the final constrained actual feeding amount. Taking the first feeding mode as an example, the baseline feeding amount is 0.3g, the winter adjustment coefficient is 1.3, the average cluster size is 4 birds, and the increment per bird is 0.05g. Therefore, the calculated amount is 0.3 × 1.3 + 4 × 0.05 = 0.59g. However, the maximum feeding amount per feeding in the first feeding mode is 0.5g, so the final actual feeding amount is truncated to 0.5g by boundary constraints. The feeding interval is also fine-tuned in a similar way, and is also limited to the range specified by the mode. This "qualitative mode, quantitative fine-tuning" approach ensures that the strategy has sufficient flexibility and response accuracy when dealing with seasonal changes and cluster size fluctuations. At the same time, boundary constraints ensure that the fine-tuning operation will not cause the feeding parameters to deviate from the basic characteristic range of the selected mode, thus guaranteeing the stability and security of the strategy.

[0034] S40 controls the feeding mechanism of the bird feeder to deliver feed according to the feeding strategy.

[0035] In this embodiment, the feed stored in layers within the bird feeder is selected and fed according to the preset feed preferences corresponding to the identified bird species. The bird feeder has multiple independent feed bins, each storing different types of feed. The control unit maintains a mapping table of bird species and feed preferences. Once a dominant bird species is identified, the corresponding feed type number is retrieved from the table, and a command is sent to the feeding mechanism to drive the motor to rotate to the outlet position of the corresponding feed bin, then the valve is opened for quantitative feed dispensing. This precise matching function greatly enhances the attractiveness of feeding and the utilization rate of feed, avoiding waste caused by feeding birds that do not like certain feeds, and also reducing the risk of feed spoilage and mold due to prolonged unconsumed use.

[0036] In this embodiment, when the environmental condition indicates that feeding is prohibited, the control device enters a low-power sleep mode. The "No Feeding" flag has the highest priority. When this flag is set, regardless of the feeding strategy output by the AI ​​decision model, the microcontroller will skip the feeding instruction, directly turn off the power to the feeding mechanism, shut down unnecessary peripherals, and enter a low-power sleep state until the environmental condition returns to "Suitable for Feeding" or "Cautious Feeding," at which point it will automatically wake up. This mechanism enables real-time response and automatic risk avoidance to environmental risks. It can automatically protect the equipment and feed during severe weather without human intervention, extending the equipment's lifespan and preventing health hazards caused by birds accidentally ingesting damp and spoiled feed.

[0037] In this embodiment, a refeeding reminder is sent when the remaining feed in a hopper is detected to be below a threshold. Each hopper is equipped with a weight sensor or infrared photoelectric sensor at its bottom to monitor the remaining feed level in real time. When the remaining feed in any hopper falls below a preset threshold, the control unit pushes a refeeding reminder notification to the user's mobile app via its built-in Wi-Fi or Bluetooth communication module. This function enables intelligent operation and maintenance of the bird feeder, preventing interruptions in feeding services due to feed depletion, ensuring a continuous food supply for birds, and improving the user experience.

[0038] Furthermore, the method also includes iterative optimization steps: Using a 7-day iteration cycle, calculate at least one of the following based on historical feeding data: feed surplus rate, bird visitation satisfaction, and feed waste rate, to evaluate the feeding effect; The weight coefficients of each input factor in the AI ​​decision-making model are automatically adjusted based on the evaluation results of the feeding effect.

[0039] Specifically, at the end of each 7-day cycle, data is extracted from historical feeding logs. Feed surplus rate is calculated by comparing the amount of feed given with the consumption estimated by image recognition. Bird visit satisfaction is assessed by comparing the change rate of average visit frequency between the current and previous cycles. Feed waste rate is estimated by identifying the amount of uneaten feed scattered on the ground. Based on the assessment results, the system automatically adjusts the weight coefficients of each input factor using a gradient descent method. Through repeated self-learning and optimization, the entire bird feeder system can adapt to subtle changes in different regions and seasons, becoming increasingly accurate in matching the actual needs of birds. For example, if it is found that feed waste is mainly caused by rainfall during the rainy season, the model will gradually increase the weight of outdoor environmental factors, making the system more sensitive to rainfall, thus truly becoming smarter with use.

[0040] like Figure 2 As shown, Figure 2 This is a schematic diagram of the bird feeder of the present invention.

[0041] The present invention also proposes a bird feeder, comprising: Camera module 10 is used to collect visual data of visiting birds; The environmental sensing module 20 is used to collect outdoor environmental data of the environment where the bird feeder is located; Clock module 30 is used to provide current seasonal time series data; The control unit 40 is electrically connected to the camera module 10, the environmental sensing module 20 and the clock module respectively. The control unit 40 is configured to execute a fully automatic feeding control method for outdoor bird feeders based on AI decision-making. The feeding mechanism 50 is electrically connected to the control unit 40 and is used to perform feed feeding under the control of the control unit 40.

[0042] In terms of hardware implementation, the control unit uses an embedded microcontroller as the main control chip, the camera module connects to an edge computing chip with an integrated AI accelerator via a MIPI interface, the environmental sensing module communicates with the control unit via an I2C bus, the clock module connects via an SPI interface, and the feeding mechanism uses a stepper motor to drive a screw propeller for precise quantitative feeding. All modules are encapsulated in a waterproof shell and powered by a combination of a lithium battery and a solar panel. This bird feeder constitutes a complete edge computing intelligent hardware system. All data processing and decision-making are completed locally, enabling it to independently and fully automatically run all the above methods and steps. It can complete the closed loop from perception and decision-making to execution locally without relying on an external cloud platform, making it particularly suitable for outdoor environments without stable network coverage.

[0043] The above are only some embodiments of the present invention and do not limit the patent scope of the present invention. All equivalent structural transformations made under the technical concept of the present invention using the contents of the present invention specification and drawings, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.

Claims

1. A fully automatic feeding control method for outdoor bird feeders based on AI decision-making, characterized in that, Includes the following steps: Visual data of visiting birds, outdoor environmental data of the environment where the bird feeder is located, and time-series data of the current season are collected. The visual data is identified to obtain bird species and size information as well as visit frequency characteristics, and environmental state characteristics and seasonal characteristics are determined based on the outdoor environment data and seasonal time series data. The bird species and size information, visit frequency characteristics, environmental status characteristics, and seasonal characteristics are input into the AI ​​decision model, and the AI ​​decision model outputs a feeding strategy. The feeding strategy includes a feeding mode, feeding frequency, and total feeding amount. The feeding mode includes a first feeding mode or a second feeding mode. The number of feedings in the first feeding mode is higher than the number of feedings in the second feeding mode. The amount of food given in a single feeding in the first feeding mode is less than the amount of food given in a single feeding in the second feeding mode. The feeding time interval in the first feeding mode is less than the feeding time interval in the second feeding mode. According to the feeding strategy, the feeding mechanism of the bird feeder is controlled to perform feed feeding.

2. The fully automatic feeding control method for outdoor bird feeders according to claim 1, characterized in that, The collected visual data of visiting birds, outdoor environmental data of the bird feeder's location, and current seasonal time-series data specifically include: The visual data is collected by a camera and an edge computing AI chip to identify the species, size and number of visiting birds, and to count the frequency of visits by various birds per unit time. The outdoor environmental data is collected by temperature and humidity sensors, raindrop sensors, and light sensors. The outdoor environmental data includes ambient temperature, relative humidity, rainfall intensity, and light intensity. The seasonal time data is obtained by the clock module built into the bird feeder to determine the current season and special periods, which include at least one of the following: the brooding period, the molting period, and the daily peak foraging periods at dawn and / or dusk.

3. The fully automatic feeding control method for outdoor bird feeders according to claim 1, characterized in that, The process of identifying the visual data, obtaining bird species and size information, as well as visit frequency characteristics, and determining environmental state characteristics and seasonal characteristics based on the outdoor environmental data and seasonal time series data also includes: The collected visual data is cleaned to remove misidentified non-bird objects, abnormal visit data, and sensor malfunction data. The dominant bird species, body size classification, daily peak visit times, and average flock size within the statistical time window are extracted as the bird species and body size information and visit frequency characteristics. The dominant bird species are those that are identified and recorded at the bird feeder within the statistical time window with the highest visit frequency or the most total visits. The energy requirements of birds are determined according to the current season, and the environmental conditions are divided into three levels: suitable for feeding, cautious feeding, and prohibited feeding, based on ambient temperature and rainfall intensity. The prohibited feeding level is used to trigger the device to enter a low-power sleep mode and stop feeding, while the cautious feeding level is used to trigger a fine-tuning operation to reduce the feeding frequency and / or the amount of food fed at one time.

4. The fully automatic feeding control method for outdoor bird feeders according to claim 1, characterized in that, The AI ​​decision-making model is a weighted multi-factor fusion model, whose input factors include bird species and size factors, visit frequency factors, seasonal characteristics factors, outdoor environment factors, and bird habit factors.

5. The fully automatic feeding control method for outdoor bird feeders according to claim 1 or 4, characterized in that, In the feeding strategy generation step, the AI ​​decision-making model makes a determination based on the dominant bird species identified in the past preset time period: If the bird is identified as a small songbird and its visit frequency is higher than the visit frequency threshold, then the first feeding mode is selected. If the bird is determined to be a medium to large-sized bird and its visit frequency is lower than the visit frequency threshold, then the second feeding mode is selected; If the bird is identified as a small songbird and its visit frequency is below the visit frequency threshold, or as a medium or large bird and its visit frequency is above the visit frequency threshold, it will enter hibernation mode.

6. The fully automatic feeding control method for outdoor bird feeders according to claim 5, characterized in that, The first feeding mode has a single feeding amount of 0.2-0.5g and a feeding interval of 20-60 minutes; the second feeding mode has a single feeding amount of 1.5-5g and a feeding interval of 2-6 hours.

7. The fully automatic feeding control method for outdoor bird feeders according to claim 1, characterized in that, The feeding strategy generation step further includes: after determining the feeding mode, multiplying the preset baseline feeding amount corresponding to the feeding mode by the adjustment coefficient corresponding to the seasonal characteristics and / or environmental state characteristics, and adding the incremental value determined by the average cluster size, thereby calculating the dynamically fine-tuned actual feeding frequency and / or single feeding amount; wherein, when the seasonal characteristics are the brooding period or the low temperature period, the adjustment coefficient is greater than 1; when the environmental state characteristics are the cautious feeding level or the seasonal characteristics are the high temperature period, the adjustment coefficient is less than 1; and, for each additional bird in the average cluster size, the single feeding amount is increased by a preset weight; the dynamically fine-tuned actual single feeding amount and feeding frequency are still limited to the feeding frequency range and single feeding amount range specified by the feeding mode.

8. The fully automatic feeding control method for outdoor bird feeders according to claim 1, characterized in that, The feeding execution steps include: Based on the preset feed preferences corresponding to the identified bird species, select the feed stored in layers in the bird feeder for feeding; When the environmental condition is such that feeding is prohibited, the control device enters a low-power sleep mode. When the remaining material in the hopper is detected to be below the threshold, a replenishment reminder is sent.

9. The fully automatic feeding control method for outdoor bird feeders according to claim 4, characterized in that, The method further includes an iterative optimization step: Using a 7-day iteration cycle, calculate at least one of the following based on historical feeding data: feed surplus rate, bird visitation satisfaction, and feed waste rate, to evaluate the feeding effect; The weight coefficients of each input factor in the AI ​​decision-making model are automatically adjusted based on the evaluation results of the feeding effect.

10. A bird feeder, characterized in that, include: A camera module is used to collect visual data of visiting birds; An environmental sensing module is used to collect outdoor environmental data of the environment where the bird feeder is located; The clock module is used to provide time-series data for the current season; A control unit is electrically connected to the camera module, the environmental sensing module, and the clock module, respectively, and the control unit is configured to perform the method as described in any one of claims 1 to 9; The feeding mechanism is electrically connected to the control unit and is used to perform feed feeding under the control of the control unit.

Images