Multi-structure attitude angle acquisition terminal suitable for chicken

Abstract

The invention provides a multi-structure attitude angle acquisition terminal suitable for chickens, and aims to solve the problems of single wearing position, poor data stability and insufficient attitude correction capability of existing poultry wearing equipment. The terminal comprises a three-axis attitude angle acquisition module, a mounting position identification module, an attitude angle stability processing module, an overturning identification and attitude correction module, a data processing and edge identification module and a data caching and communication module. The terminal supports various structural forms such as a foot ring type, a leg hanging type and a back clamping type, and adapts to chicken individuals in different body types and feeding environments; through initial posture learning and real-time dynamic correction, the wearing position is automatically recognized, corresponding posture angle deviation parameters are loaded, and the consistency of posture data and the spatial motion direction of the chicken is guaranteed; the attitude angle data quality is improved through algorithms such as sliding filtering, mutation detection and stability scoring; meanwhile, the method has overturning self-inspection and correction capabilities, and ensures data validity and behavior recognition reliability. The terminal system is not integrated with an accelerometer, only takes the attitude angle acquired by the three-axis gyroscope as a unique input signal, has the technical characteristics of clear signal input boundary, flexible structure deployment and closed identification path, and is suitable for front-end data acquisition requirements of chicken behavior identification and death state monitoring in a breeding scene.

Description

Technical Field

[0001] This invention relates to the field of poultry intelligent monitoring technology, specifically to a multi-structure posture angle acquisition terminal suitable for chickens, which falls under the cross-application technology category of livestock and poultry behavior recognition, intelligent wearable devices, animal health monitoring and edge perception systems. Background Technology

[0002] With the development of intelligent livestock and poultry farming technology, behavior monitoring methods based on wearable devices are gradually being applied to the daily health management and abnormal condition identification of poultry such as chickens. Among them, three-axis attitude angles (including pitch, yaw, and roll) are important signals reflecting changes in the spatial posture of the chicken's head and body, and have been widely used in intelligent algorithms such as behavior recognition, mortality warning, and stress detection.

[0003] Existing research has largely focused on larger animals such as cattle and pigs, where wearable devices typically employ ear tags, collars, or backpack structures, offering ample installation space and stability. However, in the context of chickens, due to their smaller size, greater mobility, and more diverse behaviors, the end-wearing structure faces more practical challenges. Chickens have small legs, and their leg rings are prone to slipping or rotating, which can lead to distortion of posture angle data. Back-mounted designs are limited by the chicken's center of gravity and feather coverage, requiring lightweight designs that avoid interference. Frequent actions such as pecking and flapping can easily cause the terminal to shift or flip, affecting recognition accuracy; Although some devices in the scenario support six-axis acquisition (including acceleration and angular velocity), their algorithm recognition process does not distinguish the signal source, and cannot guarantee the stability and individual reliability of the attitude angle.

[0004] In addition, existing systems often fail to distinguish the impact of different wearing structures on posture angle stability and lack a unified installation identification and automatic correction mechanism, resulting in low data accuracy and availability in actual deployments, making it difficult to support refined behavior recognition tasks.

[0005] Therefore, there is an urgent need for a terminal device specifically designed for chickens, supporting multiple wearing structures, adaptive correction, and ensuring stable output of posture angles, to fill the gap in the structural adaptation and signal consistency guarantee of the acquisition layer in poultry behavior monitoring systems. Summary of the Invention

[0006] This invention relates to the field of livestock and poultry monitoring, specifically to a multi-structure posture angle acquisition terminal suitable for chickens, which aims to solve problems such as insufficient stability, large drift, and low accuracy in acquiring posture angle data of chickens under different wearing positions.

[0007] This invention combines a structural adaptation mechanism, a wearing position recognition mechanism, and a posture angle correction mechanism to ensure that when worn in different positions such as the feet, legs, or back, the three-axis posture angle data of Yaw, Pitch, and Roll with temporal continuity and spatial consistency can still be obtained.

[0008] In this manual, "three-axis attitude angles" specifically refers to the three-dimensional spatial angle parameters obtained through attitude calculation based on the angular velocity signals acquired by the gyroscope, which are as follows: Pitch: An angle of rotation around the X-axis, reflecting the vertical tilt of the head; Yaw: The angle of rotation around the Z-axis, reflecting the body's horizontal turning state; Roll angle: The angle of rotation around the Y-axis, reflecting the body's tilt and rollover state.

[0009] I. Purpose of the Invention

[0010] The purpose of this invention is to provide a multi-structure posture angle acquisition terminal suitable for chickens, aiming to solve the problems of existing poultry behavior monitoring equipment such as single wearing structure, difficulty in identifying installation position, poor stability of posture angle data, and difficulty in correcting flipping errors, and to build a set of intelligent terminal systems with strong structural adaptability, high consistency of posture recognition, and clear closed signal processing path.

[0011] Specifically, the present invention aims to achieve the following objectives: 1. Multi-structure adaptation: Supports multiple wearing modes such as leg ring type, leg hanging type, and back type, adapting to individual chickens of different breeds, sizes and breeding environments, improving the flexibility of equipment deployment; 2. Wearing position recognition and parameter calibration: Through the self-learning mechanism of initial posture angle features, the terminal installation position is automatically identified and the corresponding posture angle offset correction parameters are loaded to achieve rapid deployment and data consistency assurance; 3. Enhanced stability of attitude angle data: An edge processing module is constructed to execute algorithms such as sliding filtering, mutation identification, and stability scoring, thereby improving the continuity and anti-interference capability of attitude angle output; 4. Terminal flipping recognition and automatic correction: It has the ability to recognize physical state changes such as flipping up and down and flipping to the side, and supports an automatic correction mechanism for attitude angle spatial transformation to ensure that the data semantics are consistent with the direction of chicken behavior; 5. Closed signal acquisition path and clear technical boundaries: Only a three-axis gyroscope is used as the attitude angle input source. The recognition process does not call acceleration or angular velocity fusion calculation, ensuring the structural independence and closed nature of the recognition path and preventing the technology from being bypassed by six-axis fusion.

[0012] By achieving the above objectives, this invention can construct a poultry posture angle recognition terminal suitable for various breeding conditions without relying on external environmental sensing devices (such as cameras, accelerometers, etc.), providing highly reliable and adaptable basic data collection support for intelligent recognition systems such as chicken behavior monitoring, health assessment, and mortality early warning.

[0013] II. Technical Solution

[0014] (a) Multi-structure wearing adaptation mechanism

[0015] The posture angle acquisition terminal of this invention is designed with multiple optional wearing methods to suit the body structure characteristics of chickens, adapting to the differences in breeding scenarios, chicken breeds, and management needs. Specifically, it includes:

[0016] 1. Anklet-style wearing structure: This structure uses a ring-shaped design that fits snugly around the chicken's ankle, similar to the installation method of traditional chicken identification leg bands. It features a compact structure that does not interfere with daily behavior. The terminal itself can be designed as a closed ring or an open structure with a self-locking mechanism to achieve convenient wear and stable positioning. The ankle's posture changes have a strong behavioral response capability, making it particularly suitable for recognizing lower limb-dominant behaviors in chickens, such as lying down, standing, and circling.

[0017] 2. Leg-mounted wearing structure: This method uses straps, slots, or adhesive tape to secure the terminal to the side of the chicken leg, typically located on the upper outer side of the tibia or femur. This method is suitable for larger chicken breeds (such as broilers or medium-sized laying hens) and provides relatively stable posture angle data, especially sensitive to changes in pitch and roll angles. The structural design must consider anti-slip and anti-rotation features, such as a non-slip lining or a contour-fitting snap-fit ​​housing, to prevent data errors caused by displacement during movement.

[0018] 3. Clip-on wearing structure: The back clip structure features a lightweight design, using flexible clamps or webbing to attach the terminal above the chicken's back midline, near the shoulder area. This wearing method is more suitable for recognizing body-dominant behavioral patterns, such as wing flapping, falling, and head tilting. The terminal needs to be designed with a low profile and a shape that conforms to the feather structure to avoid interfering with the chicken's wing movements. To prevent displacement or rotation due to movement, the structure should integrate a dual-point fixing mechanism (such as front and rear double buckles) to improve installation stability.

[0019] To ensure the stability and effectiveness of data acquisition, this invention incorporates rotation protection mechanisms and slippage suppression structures in all three wearing structures described above, for example: The shell features a contoured design that matches the chicken's body shape. A lining material with a high coefficient of friction is used to fit snugly against the chicken's skin or feathers; An adjustable-length elastic fixing device allows for adaptation to different individual sizes; The fixed structure with symmetrical arrangement of two points prevents angular errors caused by unilateral swaying.

[0020] Through the above-mentioned structural optimization design, the present invention ensures that the attitude angle acquisition terminal does not undergo unexpected rotation or displacement during the chicken's activities, thereby effectively improving the stability and reliability of the three-axis attitude angle data of Yaw, Pitch, and Roll, and providing a solid data foundation for subsequent abnormal behavior identification and mortality status judgment.

[0021] (II) Installation location identification and calibration mechanism

[0022] To address the potential differences in attitude angle references that may arise from different wearing structures for chickens (such as leg bands, leg attachments, and back clips), this invention introduces an installation position recognition and automatic calibration mechanism into the attitude angle acquisition terminal to achieve adaptive adjustment and standardized output of the three-axis angles of Yaw, Pitch, and Roll.

[0023] Specifically, the terminal enters the attitude initialization learning phase upon initial startup or reset. During this phase, the terminal continuously collects a sequence of three-axis attitude angle data in a static state over a short period of time, extracts its statistical feature values ​​(such as average value, drift range, stability factor, etc.), and compares these features with the built-in installation position attitude template library.

[0024] The template library pre-stores reference ranges of attitude angles for typical installation positions (foot, leg side, back) in a static state, for example: Ankle band style: Pitch ≈ 90°, Roll ≈ 0°, Yaw variable; Leg-mounted type: Pitch ≈ 20°~40°, Roll ≈ ±30°; Clip-on style: Pitch ≈ 0°, Roll ≈ 0°, Yaw is variable.

[0025] By matching with a template, the system can automatically determine the current wearing position of the terminal and load corresponding attitude angle offset correction parameters based on the recognition result for dynamic adjustment of subsequent real-time data. Correction parameters include, but are not limited to: Offset compensation values ​​for the reference attitude angle (such as Yaw_offset, Pitch_offset); Three-axis coordinate system rotation matrix (used for coordinate space consistency transformation); Filter gain adjustment value (optimize response sensitivity based on dynamic stability differences at different locations).

[0026] This installation identification and calibration mechanism not only takes effect during initial deployment, but also supports periodic re-judgment during terminal operation. For example, it can perform a posture self-check once a day to deal with reference offset problems caused by factors such as loose terminal wearing, slippage, or passive rotation.

[0027] Furthermore, to enhance the reliability of identification, the system can cross-verify the installation location identification results with the behavior identification results. For example, if a continuous "back-mounted unique posture fluctuation pattern" is detected but the current calibration status is "ankle ring type," a secondary judgment process or an alarm will be triggered to prompt the user to reinstall the terminal.

[0028] Through the above-mentioned installation location identification and calibration mechanism, the present invention can achieve standardized output of posture angle data of chickens under different wearing methods without relying on manual intervention, providing a unified and highly reliable data input source for subsequent behavior recognition models.

[0029] (III) Attitude Angle Data Stability Enhancement Mechanism and Edge Recognition Implementation Path

[0030] In actual chicken farming, daily behaviors such as flapping wings, jumping, pecking, and running can cause drastic fluctuations, sudden disturbances, or continuous deviations in the three-axis data (Yaw, Pitch, Roll) of a wearable posture angle acquisition terminal. To ensure the accuracy of the subsequent abnormal behavior recognition model, this invention integrates a posture angle data stability enhancement processing mechanism into the terminal system, which is used to perform real-time processing and dynamic optimization of posture angle data during the acquisition phase.

[0031] This mechanism can run locally on the data collection terminal (i.e., at the edge), without relying on cloud processing. It features low latency, high responsiveness, and strong anti-interference capabilities, making it suitable for aquaculture scenarios with complex network environments or limited data upload capabilities. In an optional implementation, this mechanism can also be executed in batches on the cloud for historical data backtracking analysis and model retraining.

[0032] This processing mechanism mainly includes the following three sub-modules:

[0033] 1. Sliding weighted filtering mechanism

[0034] This invention introduces a sliding time window mechanism during attitude angle data acquisition to perform weighted smoothing on continuous data sequences, thereby suppressing data noise caused by instantaneous changes. In an optional embodiment, the attitude angle filtering smoothing value can be calculated as follows (taking the pitch angle as an example):

[0035] in: P iThis represents the original attitude angle value at the i-th time point; w i Decreasing weights based on time; n is the length of the sliding window (e.g., 10 samples within 5 seconds).

[0036] The above formula is an optional implementation of the present invention, and can also be replaced by other equivalent smoothing methods such as median filtering, exponential smoothing, and mean square response. It does not constitute the only limitation on the technical solution.

[0037] 2. Attitude angle change identification and shielding mechanism

[0038] To identify points of drastic and abnormal change in the data, this invention introduces a mutation detection logic. The joint change magnitude score of the three-axis attitude angles can be defined as follows:

[0039] If the score is S t If the angle exceeds a set abrupt change threshold (e.g., 30°~40°), the system considers that point a "high-perturbation data segment" and excludes it from the behavior recognition model calculation. If necessary, the system can repair this interval by fitting effective data before and after it to maintain data continuity.

[0040] The above mutation scoring method is only an example. In actual systems, other mutation detection algorithms (such as threshold discrimination, Z-score scoring, rate of change prediction, etc.) can be used to achieve equivalent functions.

[0041] 3. Stability scoring and data validity assessment To evaluate the usability of attitude angle sequences in real time, this invention introduces a stability scoring mechanism within a sliding window. In one exemplary implementation, the calculation method can be described as follows:

[0042] in: σ t This represents the standard deviation of the attitude angles within the current window; θ max The maximum fluctuation tolerance is preset for the system; The closer the stability score is to 1, the more stable the current data is.

[0043] When the stability score is below a threshold (e.g., 0.7), the data segment can be marked as a "high-risk sample," and the system can proactively remove it or prompt the user to check the wearing status.

[0044] 4. Edge processing mechanism and deployment instructions

[0045] The attitude angle stability processing module can be integrated into the embedded system of the terminal device and operates in edge processing mode, possessing the following characteristics: Real-time data acquisition and processing with millisecond-level response; This can reduce the frequency of uploading raw data, and only feature values ​​or recognition results are sent; Supports offline operation, suitable for scenarios with no or low network coverage; Enhance the terminal's independent discrimination capability and reduce system energy and data consumption.

[0046] In another alternative implementation, the above mechanism can also be deployed on a cloud platform for batch analysis, model training, and anomaly backtracking.

[0047] In summary, the posture angle stability enhancement mechanism of this invention effectively improves the acquisition quality and spatiotemporal consistency of chicken posture angle data through a multi-dimensional dynamic processing strategy, providing a stable and reliable input foundation for subsequent behavior recognition, mortality determination, etc., and forming an indispensable edge recognition module in the intelligent monitoring system.

[0048] In a preferred embodiment, the attitude angle stability enhancement processing mechanism of the present invention includes algorithm modules such as sliding filtering, mutation identification, and stability scoring, which can be integrated and deployed locally on the acquisition terminal to form part of the edge recognition module. This mechanism can be executed in real time during data acquisition, has low latency and independent judgment capabilities, effectively reduces the system's dependence on network transmission, and is suitable for offline operation requirements in various aquaculture scenarios.

[0049] In another alternative implementation, the above processing mechanism can also be deployed on a cloud platform to perform batch data processing, behavior model training, and recognition strategy calibration to meet the multi-objective optimization tasks under large-scale datasets.

[0050] This invention does not restrict the deployment location. Any processing path that achieves attitude angle stability enhancement, whether deployed on a terminal device, edge node, or remote server, falls within the protection scope of this invention.

[0051] (iv) Equipment posture correction and vertical flipping recognition mechanism

[0052] In actual farming scenarios, due to improper manual installation, the pulling of chickens in their daily behavior, or the loosening of the terminal structure itself, the attitude angle acquisition terminal may flip up and down, tilt to the side, or rotate unexpectedly, causing the three-axis attitude angle output direction to be inconsistent with the actual spatial coordinate system, which seriously affects the accuracy of subsequent behavior recognition.

[0053] To address the aforementioned issues, this invention introduces a device posture correction and flip recognition mechanism. Through self-sensing and dynamic correction, it ensures that the Yaw, Pitch, and Roll data output by the terminal are always consistent with the chicken's actual spatial movement direction and are not affected by changes in the terminal's physical orientation.

[0054] 1. Install initial flip recognition: After each cold start or restart, the terminal executes an initial attitude self-check algorithm to obtain the direction of the gravity vector when the device is stationary, and determines whether there is any installation rollover based on the absolute value distribution of the three-axis attitude angles. For example: If the initial value of the pitch angle is significantly less than 0° (e.g., -90°), and the roll angle is stable around ±180°, then it is determined to be "vertical installation"; If the pitch and roll angles deviate from the default reference coordinate system by more than the set threshold, then enter the flip confirmation mode.

[0055] In the confirmation mode, the system combines short-term stability judgment (such as no obvious attitude fluctuation within 5 seconds) and default assembly logic to automatically generate a set of coordinate correction matrices, converting the original three-axis attitude angles back to the standard coordinate system, ensuring the consistency of data in spatial direction.

[0056] 2. Automatic correction of rollover and drift during operation: The terminal continuously monitors its own attitude trend dynamically. Once a sudden change in the overall offset trend of the three-axis attitude angles is detected (such as the pitch ≈ 0° suddenly stabilizing at ±180°), the system triggers the flip self-check logic and determines whether attitude correction should be performed based on the following rules: Determine whether the directions of attitude angle changes are opposite before and after the flip; Compare whether the stability score of the attitude angle data has decreased significantly; Analyze whether the behavior recognition results show continuous recognition failures or illogical states (such as suddenly "walking in the opposite direction" or "continuous diving").

[0057] When the above judgment results meet the threshold conditions, the system automatically activates the coordinate mapping switching mechanism. Based on the identified rotation type (X, Y, or Z axis flip), it selects an appropriate spatial transformation matrix to correct the current output in real time, so as to restore it to the semantic consistency under the original coordinate reference system.

[0058] 3. Flip status recording and safety alert mechanism: To enhance system robustness and auditability, the terminal records the time point, angle offset, and correction matrix parameters of each flip identification and correction operation, and uploads this information to the management platform when necessary. The platform can use algorithms to flag individuals with "high incidence of installation anomalies" and prompt users to perform physical checks or reinstall the system.

[0059] In addition, if the flipping state occurs frequently or the number of correction failures exceeds the set threshold, the terminal can trigger low-level alarms, such as LED flashing, behavior recognition pause, signal isolation, etc., to prevent erroneous data from continuously affecting the recognition system.

[0060] In summary, this invention constructs a terminal posture consistency assurance system for chicken behavior recognition tasks through multiple mechanisms including initial recognition, dynamic flip self-calibration, spatial coordinate transformation, and status recording. This ensures that even under non-ideal wearing conditions, it can still stably output yaw, pitch, and roll posture angle data that match the actual movement direction of the chicken, providing reliable support for subsequent intelligent recognition algorithms.

[0061] (V) Overall Description of System Structure

[0062] In summary, this invention proposes a multi-structure attitude angle acquisition terminal suitable for chickens, the structure of which includes: Wearing an adapter module supports multiple forms such as leg band, leg attachment, and back clip, adapting to different body sizes and behavioral characteristics of chickens; The installation position identification and attitude calibration module is used to automatically identify the terminal installation position during the startup phase and load the corresponding attitude angle offset correction parameters. The attitude angle data stability processing module includes mechanisms such as sliding filtering, mutation identification, and stability scoring, which are used to perform real-time edge processing on attitude angle data locally on the terminal. The equipment flipping recognition and coordinate correction module is used to automatically identify installation abnormalities when vertical flipping or lateral rotation occurs, and convert the three-axis attitude angles back to the standard reference coordinate system through spatial mapping. The status recording and alarm output module is used to record the historical changes in the terminal's attitude status and output safety prompts when frequent anomalies occur.

[0063] The above modules can be integrated into a single physical terminal to form a unified structural design, or they can be functionally divided through software logic. They support low power consumption and high robustness operation, and are suitable for building a daily health monitoring and behavior recognition system in high-density chicken farming scenarios.

[0064] The multi-structure attitude angle acquisition terminal described in this invention is applicable to scenarios including but not limited to chicken behavior recognition, abnormal state monitoring, and mortality warning. It is particularly suitable for the applicant's patent "A method for recognizing abnormal behavior and mortality status of chickens based on attitude angles", serving as its key data acquisition and recognition signal input terminal, and providing high-quality input data to ensure the realization of core algorithms such as attitude angle discrimination, fluctuation detection, and status scoring in the recognition method.

[0065] The aforementioned system structure and methodological logic constitute a supporting application path, jointly constructing a complete technical solution for intelligent monitoring scenarios in poultry farming.

[0066] (vi) Signal input path limitation instructions

[0067] To ensure that the terminal system described in this invention has clear and controllable signal input boundaries in terms of structural design and functional implementation, the three-axis attitude angle acquisition module used in this system only integrates a three-axis gyroscope to directly acquire the pitch, yaw and roll angles of the chicken in its natural state, and serves as the sole signal input source for the behavior recognition module.

[0068] This system does not integrate an accelerometer, nor does it perform a fusion calculation process of acceleration and angular velocity. It does not construct a six-axis inertial navigation model (IMU Fusion). The system's path recognition does not rely on any calculated angles or indirect attitude estimation results. Instead, it uses the original three-axis attitude angles as the sole criterion for constructing the behavior state recognition logic.

[0069] Even if the terminal hardware has expansion capabilities (such as an embedded six-axis chip or support for a three-axis + three-axis structure), as long as the actual recognition logic does not call acceleration dimension related data (such as Ax, Ay, Az) and does not execute the six-axis fusion processing flow, it still falls within the three-axis attitude angle limited recognition path claimed by this invention and is within the protection scope of this invention.

[0070] Through the aforementioned design of the input path, this invention fundamentally differs from many existing animal behavior recognition systems based on six-axis fusion algorithms in terms of signal processing structure, demonstrating the following innovative features: Native data source and closed path: Directly acquire three-axis attitude angle signals without the need for fusion algorithm calculation process, reducing system complexity; The input signal has clear boundaries: the recognition logic is strictly limited to using only the three axes of Pitch, Yaw, and Roll, and does not call other dimensions of perception data; Suitable for low-power edge deployments: Simplifies the computing path, facilitates operation in lightweight terminals, and enables real-time identification and offline response; Enhanced patent protection capabilities: Effectively prevents others from circumventing the protection boundaries of this invention's technical solutions by introducing accelerometers or fusion models.

[0071] In summary, the limitation of the signal acquisition path in this invention not only reflects the structural independence and path closure of the recognition logic, but also provides a reliable data foundation for subsequent system performance optimization, algorithm deployment flexibility, and behavior recognition accuracy assurance.

[0072] (vii) Non-restrictive description of structural appearance

[0073] To further clarify the scope of protection of this invention: The "chicken multi-structure posture angle acquisition terminal" described in this invention emphasizes the implementation method of functional structure and recognition logic path, and does not impose any limiting requirements on the specific structural form of the terminal device's appearance, size, shell design, or wearing posture.

[0074] The terminal can be in the form of a foot ring, leg hanging, back clip, or any form that can be fixed to a part of the chicken's body. It can also be embedded in other carriers to achieve functional integration. All implementation paths are considered as equivalent structures of the present invention. Attached Figure Description

[0075] Figure 1 : A schematic diagram of the overall composition of the terminal structure of the present invention; Figure 2 : Schematic diagram of wearing structure for multiple terminals; Figure 3 Flowchart of initial attitude angle calibration and installation position identification; Figure 4 Flowchart of attitude angle stability enhancement process; Figure 5 Logic diagram for vertical flip recognition and coordinate correction; Figure 6 : Diagram illustrating the interaction between the identification data cache and the edge processing module. Detailed Implementation

[0076] To make the objectives, technical solutions, and beneficial effects of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. This invention is not limited to the specific embodiments described below; any equivalent substitutions or improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

[0077] Example 1: Composition and Workflow of the Data Acquisition Terminal

[0078] The following, with reference to a typical embodiment, further illustrates the composition and workflow of the multi-structure attitude angle acquisition terminal for chickens described in this invention. This embodiment is only used to illustrate the technical solution of this invention and does not constitute a limitation on the scope of protection.

[0079] In a real-world farming scenario, ten broiler chickens were selected as target individuals, and each chicken was fitted with a posture angle acquisition terminal as described in this invention. The terminal adopts a leg-ring type wearing structure, which is fixed to the chicken's ankle with a flexible elastic material to ensure that the device does not rotate or slip during daily activities.

[0080] The terminal integrates a low-power three-axis gyroscope chip (e.g., L3GD20H) to collect the three-axis attitude angles (Pitch, Yaw, Roll) of the chicken in its natural state in real time. After the terminal is powered on, it enters the installation position self-identification and calibration mode. Based on the attitude angle characteristics in the static state (Pitch ≈ 90°), it automatically determines that the installation is on the feet and loads the matching attitude correction parameters.

[0081] In actual operation, the terminal performs attitude angle sampling at a frequency of 1 Hz, forms a sliding data window every 5 seconds, and performs the following operations through the local embedded processor: The sliding weighted filtering algorithm is executed to smooth short-term disturbance fluctuations. Calculate the triaxial mutation score and remove high-interference segments that exceed a set threshold; The attitude angle stability score within the evaluation window is below 0.6, and the data is marked as unreliable. Perform real-time behavioral pattern recognition (such as lying prone, spinning, etc.) and record the recognition results; The system can identify the flip state and automatically trigger coordinate system remapping correction if the pitch is detected to have stably shifted to the ±180° range. Record calibration operation logs and upload them to the cloud platform periodically via the Wi-Fi module.

[0082] All recognition logic is based solely on three-axis attitude angle data and does not involve acceleration signal acquisition and fusion calculation. Test results show that during 48 hours of continuous operation, the system's attitude angle recognition error is less than 5°, the success rate of flip recognition and correction reaches 100%, and the total system power consumption is less than 20mAh / day, meeting the low-power deployment and edge recognition requirements of actual poultry farming scenarios.

[0083] This embodiment verifies the structural adaptability, deployment stability, and recognition accuracy of the present invention, fully demonstrating the adaptability and practicality of the system in chicken behavior recognition and terminal posture perception tasks.

[0084] Example 2: Adaptation verification of the leg-mounted wearing structure

[0085] In another set of experiments, five laying hens were selected for wearing tests using the leg-mounted structure of the terminal described in this invention. The terminal uses a flexible, strap-type fixation method, installed on the side of the chicken's thigh, avoiding the joints and areas with thick feathers.

[0086] After powering on, the terminal automatically identifies the current static posture features of Pitch ≈ 30° and Roll ≈ ±20°, determines it to be the leg mounting position, and loads the preset leg-side mounting parameter set, including the posture angle reference correction table and behavior model switching coefficients. In the leg-side position, the terminal can more sensitively identify lower limb-dominant behaviors such as jumping, walking, and kicking.

[0087] In the experiment, by comparing the results with those of manually recorded and labeled behavior, the accuracy of the terminal in recognizing behaviors such as walking, lying still, and agitation when worn on the leg reached 92.4%, and the average data stability score was 0.86, which verified the adaptability and data availability under the leg-side mounting structure.

[0088] Example 3: Automatic Recognition and Real-time Correction Test of Flip Posture

[0089] In this experiment, three chickens were selected and the terminals were intentionally installed in reverse (i.e., the leg bands were installed backwards, causing the terminals to flip upside down). After the device was started, it entered the initial posture self-check process. It was detected that Pitch ≈ -90° and Roll ≈ ±180°, which did not match the normal leg band posture. The system automatically judged it as "installed upside down".

[0090] After the terminal determines that the condition is met, it calls the spatial mapping module to automatically generate the attitude transformation matrix and perform three-axis attitude angle correction to ensure that the output yaw / pitch / roll data direction is restored to the standard state of the logical coordinate system.

[0091] During operation, the behavior recognition module was not affected by incorrect installation and could still recognize actions such as walking, lying prone, and turning normally. This demonstrates that the present invention has robustness in wearing direction and autonomous posture reconstruction capability, and can maintain stable operation of the recognition function under non-ideal wearing conditions.

[0092] Example 4: Edge Recognition Performance and Offline Early Warning Response Test

[0093] In this scenario, simulating a Wi-Fi outage in a farm, the terminal is set to offline edge mode, using only local data acquisition and recognition functions. The terminal samples the attitude angle once per second and performs stability scoring, behavior recognition, and roll correction processes locally.

[0094] When attitude angle instability occurs continuously (such as long-term extreme values ​​of pitch / roll with no fluctuation), the system continuously judges it as dead within 10 minutes, and the terminal triggers a local alarm mechanism (such as LED flashing and Bluetooth short message broadcasting), which can issue a death warning without relying on the network.

[0095] This test verifies that the terminal of the present invention can still perform a complete acquisition-judgment-output closed loop in a network-free environment, demonstrating the system's independent operation capability and practical value in edge computing scenarios.

[0096] Example 5: Status Tracking and Platform Notification Mechanism for Individuals with Frequent Abnormal Installations

[0097] In another test, the testers repeatedly adjusted the orientation of the terminal on two chickens to simulate frequent loosening or accidental installation. During 72 hours of continuous operation, the terminal detected more than 12 flip state transitions. The system recorded all flip state time points, attitude offsets, and correction matrix parameters completely and uploaded them to the platform simultaneously.

[0098] The platform identifies the individual as a "chicken with a high incidence of installation abnormalities" according to the set rules, and pushes a prompt in the management interface: "It is recommended to re-check the stability of the terminal or change the wearing method."

[0099] This embodiment demonstrates that the present invention can not only identify flipping, but also realize the functions of flipping record and risk individual labeling, thereby improving the management visualization capabilities of the system and the response efficiency of aquaculture personnel.

[0100] Example 6: Stability Test of Day and Night Behavior Comparison Recognition

[0101] To evaluate the recognition stability of the attitude angle acquisition terminal of the present invention under different lighting and behavioral rhythms, eight commercial laying hens were selected, equipped with the terminal device of the present invention, and data was continuously collected and analyzed for 72 hours under natural lighting control.

[0102] The system is configured to sample attitude angles at a frequency of 1 Hz, with a sampling window of 5 seconds, a sliding filter window size of 10 groups, and edge recognition processing every 5 minutes. The collected data is divided into daytime activity periods (6:00–18:00) and nighttime inactivity periods (18:00–6:00), and the distribution of behavior recognition results and changes in stability scores are statistically analyzed for each period.

[0103] The recognition results show: During the day, their behavior changes frequently, mainly consisting of standing, walking, and pecking, with a large fluctuation in the average posture angle. At night, they mainly lie prone and still, maintaining a pitch angle that is close to horizontal (80°~100°), with significantly reduced fluctuations. The system stability score averaged 0.94 at night and 0.81 during the day, with no instances of false triggering.

[0104] The results verify that the system has high adaptability to circadian rhythms, can stably identify state transitions, and the nighttime quiescent data has high-quality features, which can be used for death identification and behavioral baseline establishment.

[0105] Example 7: Antimutagenic adaptation test under feed disturbance conditions

[0106] Under simulated high-stimulation conditions of feed feeding, the robustness of the recognition system of this invention to interference from severe disturbances in chicken behavior was tested.

[0107] Six chickens were fitted with terminals and fed at fixed times each day (9:00, 13:00, and 17:00). During this time, the chickens exhibited highly dynamic behaviors such as flapping their wings, scrambling for food, and jumping, resulting in sudden and large-scale changes in their posture angles.

[0108] This system employs a built-in sliding window mutation identification mechanism, setting the joint mutation scoring threshold to 40°. During the data feeding period, a total of 65 mutation segments were identified, accounting for 7.4% of the total data. The system automatically masked these mutation segments and performed fitting interpolation repair to ensure that the behavior identification was not falsely interfered with.

[0109] Furthermore, the system did not falsely report a death state in the behavior recognition output, and all feed stimulus responses were correctly identified as highly active states. This indicates that the posture angle stability enhancement mechanism described in this invention has good anti-interference ability under severe disturbances and is suitable for practical application needs in real farms where there are short-term high-dynamic behaviors.

[0110] Example 8: Collaborative Operation of Terminal Edge Recognition and Communication Mechanism

[0111] To verify that the present invention has edge recognition and communication linkage functions in actual chicken farming scenarios, five chickens were selected and equipped with the terminal described in the present invention, and deployed in a regular open chicken house.

[0112] The terminal structure integrates a three-axis gyroscope acquisition module, a low-power MCU processor, a Bluetooth BLE communication module, and a Flash data cache module. The system adopts the following edge computing and data communication mechanism:

[0113] 1. Edge recognition mechanism: Attitude angles are sampled at a frequency of 1 Hz; A data window is generated every 5 seconds, and sliding filtering, mutation identification, and behavior discrimination are performed locally. Every 10 minutes, check for continuous anomalies (such as prolonged periods without fluctuations + stable extreme values). If the abnormal score exceeds the threshold, a low-power LED will be triggered to flash immediately on the terminal as a warning.

[0114] 2. Data communication mechanism: By default, the terminal uploads key behavioral events via BLE broadcast every 30 minutes. If the network is disconnected, the behavior recognition results and stability scores are cached in local Flash memory; After the BLE signal is restored, historical data will be automatically retransmitted and timestamps will be aligned with the cloud platform.

[0115] 3. Early warning output and platform linkage: If a death recognition event occurs, the terminal will proactively send a high-priority event packet via BLE. Upon receiving the data, the upper-level gateway immediately synchronizes it to the platform and triggers a notification on the APP. All identification logs can be used for subsequent behavior retrospective analysis and management evaluation.

[0116] Test results show that the terminal operated at the edge for 48 hours, experiencing two BLE signal interruptions. The terminal automatically completed cache recording and retransmission, achieving 100% data integrity on the platform. This demonstrates that the invention supports edge recognition, intermittent communication, and reliable data transmission, making it suitable for stable deployment in real-world aquaculture farms under fluctuating network conditions.

Claims

1. A multi-structure attitude angle acquisition terminal suitable for chickens, characterized in that, The terminal includes: The three-axis attitude angle acquisition module is used to acquire the pitch, yaw and roll angles of chickens in their natural state. The behavior recognition module is used to determine the behavior state or abnormal behavior of chickens based solely on the three-axis attitude angle signals. The three-axis attitude angle signal is the sole source of input data for the identification module.

2. The terminal according to claim 1, characterized in that, The three-axis attitude angle acquisition module is implemented by integrating a three-axis gyroscope. The gyroscope directly outputs the three-axis angle values ​​of Yaw, Pitch, and Roll without relying on an accelerometer for fusion calculation.

3. The terminal according to claim 1 or 2, characterized in that, The behavior recognition module does not include acceleration or angular velocity fusion algorithms, does not construct a six-axis inertial model, and does not use estimated attitude angle data.

4. The terminal according to claim 1, characterized in that, The behavior recognition module is deployed in the edge processing unit on the terminal and can complete behavior judgment and status output without network communication.

5. The terminal according to claim 1, characterized in that, The terminal further includes an attitude angle stability enhancement processing module, which performs sliding filtering, abrupt change detection, and data stability scoring to improve the continuity and recognition reliability of attitude angle data.

6. The terminal according to claim 1, characterized in that, The terminal includes an attitude correction module, which is used to identify whether the terminal wearing direction has been flipped, and to perform spatial coordinate transformation on the attitude angle data during the identification process to restore it to the standard reference system.

7. The terminal according to claim 1, characterized in that, When the terminal is started, it includes an installation position recognition mechanism, which determines the wearing position of the terminal based on the initial attitude angle state and automatically matches the corresponding attitude offset correction parameters.

8. The terminal according to claim 1, characterized in that, The terminal further includes a behavior recognition result caching module and a data synchronization module, which are used to record recognition process information and upload it to the platform via Bluetooth, Wi-Fi or other communication methods.

9. The terminal according to claim 1, characterized in that, During the execution of the identification logic, any path that calls three-axis attitude angle data as the basis for judging behavior status is considered to fall within the protection scope of this invention.

10. The terminal according to claim 1, characterized in that, The terminal is suitable for ankle clip, leg clip, or back clip wearing structures, and the recognition logic establishes a mapping relationship with the collected posture angle data, without depending on changes in the external structure of the terminal.

Images