Cow parturient behavior identification method based on attitude angle change of gyroscope

Abstract

The invention discloses a bovine parturient behavior identification method based on gyroscope attitude angle change. According to the method, three-axis attitude angle data such as a yaw angle (Yaw), a pitch angle (Pitch) and a roll angle (Roll) are collected through a gyroscope sensor fixed to the head of a cattle and are recorded and processed according to a set time period. And in a preset time window, extracting change characteristic parameters of the attitude angle, such as a Yaw accumulated change value, Pitch high-order maintenance time, Roll fluctuation frequency and the like, comparing the change characteristic parameters with a preset parturient behavior rule, and judging whether parturient early warning is triggered or not. The method does not need image acquisition or temperature monitoring, has high robustness and low power consumption, and is suitable for a bovine parturient monitoring and intelligent breeding management system of a large-scale farm.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent management of livestock, in particular to a method for recognizing parturition behavior of cattle based on changes in attitude angle of a gyroscope, which is suitable for monitoring the behavior of cows such as dairy cows and beef cows before delivery and early warning management of parturition. BACKGROUND

[0002] In large-scale cattle farms, the parturition process of cows is often accompanied by problems such as unclear behavior characteristics, uncontrollable labor, and difficulty in monitoring at night. If the signs of delivery are not recognized in time, it may cause serious consequences such as stillbirth, maternal birth canal injury, delayed delivery, and affect the reproductive efficiency and economic benefits.

[0003] Traditional methods of identifying estrus or parturition behavior rely on manual patrol or video monitoring, which has high misjudgment rate, poor timeliness, high coverage cost, and other problems, and cannot meet the needs of modern smart farms for all-weather and high efficiency.

[0004] Currently, a small number of studies explore monitoring the delivery behavior of cows through temperature, acceleration, voiceprint recognition, etc., but there are limitations in power consumption, environmental adaptability, and complexity of the recognition model.

[0005] Therefore, the behavior recognition method based on the attitude angle changes (Yaw, Pitch, Roll) collected by the gyroscope has the advantages of low power consumption, anti-interference, and simple modeling, providing a reliable technical path for improving the intelligent recognition of cattle delivery. SUMMARY

[0006] (I) Invention purpose

[0007] The present application provides an automatic recognition method for parturition behavior based on changes in the attitude angle (Yaw, Pitch, Roll) of the head of cattle, aiming to realize continuous monitoring of key actions before delivery, behavior recognition, and intelligent early warning output, and is suitable for the all-weather, low-cost, and high-accuracy management needs of modern farms for cow delivery.

[0008] (II) Technical solution

[0009] 1. Attitude angle acquisition

[0010] A three-axis gyroscope worn on the head of a dairy cow (ear tag, ear clip, or headband) is used to collect Yaw, Pitch, Roll angle data at a period of 5-10 seconds, and generate a time series.

[0011] 2. Behavior feature extraction

[0012] Within a set statistical window (such as 30 minutes or 1 hour), the following indicators are calculated: Yaw angle cumulative change; Pitch angle high or low duration; Roll angle significant fluctuation frequency; Three-axis linkage fluctuation characteristics.

[0013] 3. Behavior rule judgment

[0014] Set multiple behavior identification rules, for example: Yaw angle 30 minutes cumulative change exceeds 1080°; Pitch angle remains above +20° for more than 10 minutes; Roll angle repeatedly switches within ±15° range more than 20 times; Three-axis simultaneous severe fluctuation without obvious interval.

[0015] The following table is a typical behavior rule:

[0016] 4. Individual baseline modeling

[0017] The present application introduces individual behavior baseline, and generates individual angle fluctuation model using non-labor history data. Real-time data exceeding the fluctuation threshold is judged as abnormal, avoiding false positives caused by using population average.

[0018] 5. Risk scoring mechanism

[0019] To realize comprehensive evaluation of multiple behavior characteristics, the following weighted scoring system is constructed: The scoring formula is as follows: Labor risk score = 0.35 × ΔYaw change ratio + 0.25 × Pitch high maintenance ratio + 0.15 × Roll frequency ratio + 0.15 × three-axis linkage factor + 0.10 × night fluctuation offset rate The judgment criteria are as follows:

[0020] 6. Early warning output.

[0021] When the scoring result reaches the threshold value, the system automatically generates early warning information, and can notify the on-duty personnel through local prompt or remote platform, realizing all-weather, unattended intelligent recognition of sow labor.

[0022] (Three) Technical effects and benefits.

[0023] Realize all-weather automatic recognition of cow labor behavior, especially suitable for unattended night scene; Adopt the combination of low-frequency sampling and attitude angle judgment, effectively reduce the device power consumption and operation and maintenance cost; Can adapt to individual differences and group baseline, improve the recognition robustness and early warning accuracy; Help to reduce the stillbirth rate and improve the reproductive efficiency, has wide popularization value.

[0024] (iv) Explanation of differences in cattle's calving behavior and the suitability of methods

[0025] Pre-partum behavior exhibits individualization and breed characteristics across different cattle breeds, particularly among common livestock breeds such as dairy cows, beef cattle, and buffalo, where significant differences are observed. To enhance the applicability and generalization ability of this invention, the system adapts to the following behavioral differences through adjustable rule thresholds and individual modeling mechanisms:

[0026] 1. Dairy cow Dairy cows are relatively docile and exhibit strong social behavior. Their pre-partum behavioral characteristics are quite obvious, making them suitable for accurate identification through changes in posture angles. Yaw angle characteristics: Frequently turning around in place to observe similar objects or outside the fence, showing a continuous high-frequency fluctuation between ±45° and ±90°; Pitch characteristics: There are nest-building attempts that alternate between looking up and looking down, and the pitch changes drastically; Roll angle characteristics: Before lying down, there are multiple struggling head tilting and rolling movements, with a significant increase in variations of ±15° or more. In summary, the pre-partum behavior of dairy cows has a clear sequence of actions, making it suitable for multi-dimensional rule fusion judgment.

[0027] 2. Beef cattle Beef cattle are large and muscular, and before calving, they frequently adjust their posture due to the increased physical strain, but limited space restricts their range of motion. Yaw angle characteristics: less large-angle head turning, but will repeatedly shift left and right within a narrow range (±30°). Pitch angle characteristics: Due to the downward shift of the center of gravity, women often maintain a downward standing posture before labor, resulting in a downward but stable pitch angle. Roll angle characteristics: Frequent lying down and getting up, slight left and right swaying, reflected as high-frequency vibration within a range of ±10°; In summary, beef cattle often exhibit repetitive postures rather than drastic changes in angle during calving, and the rules should focus on frequency and stability deviations.

[0028] 3. Water buffalo Water buffalo are docile and not very alert. Their pre-birth behavior tends to be more introverted and secretive, but their posture angles still offer some clues. Yaw angle characteristics: relatively stable, with occasional head-turning behavior, mainly ±20°; Pitch angle characteristics: Before childbirth, children usually tilt their heads back to find a quiet area, and the pitch angle shows a continuous upward trend; Roll angle characteristics: When attempting to lie down, it exhibits a slow, unilateral tilt, with infrequent changes in the Roll angle but sudden large-angle shifts; In summary, it is suitable to construct the recognition logic by combining the features of a continuously rising pitch angle and a sudden jitter in the roll angle.

[0029] (v) Mechanism for handling abnormal misjudgments.

[0030] To reduce false alarms and false negatives caused by factors such as abnormal attitude angle fluctuations, individual behavioral differences, and environmental interference, this invention further designs the following multi-level misjudgment suppression mechanism to improve the accuracy and robustness of system identification:

[0031] 1. A dual-model comparison mechanism between individual behavioral baselines and group means.

[0032] This method also establishes: Individual behavioral baseline model: Based on the historical posture angle change data of each cow during the non-partum period, dynamically model its "normal behavior fluctuation range"; Group mean model: Statistically analyze the mean changes in posture angles of individuals of the same breed, age, or within the same pen to construct a "collective behavior standard".

[0033] During the behavior recognition process, the labor behavior label is only triggered when an individual's real-time behavior deviates from both its own baseline model and the group behavior mean model, thereby reducing false alarms caused by differences in individual activity levels.

[0034] 2. Night-to-day dynamic discrimination mechanism.

[0035] Considering that cattle behavior tends to be more stable at night than during the day, the system introduces a diurnal behavior shift factor: Calculate the deviation of the nighttime attitude angle fluctuation value from the average nighttime value of the cattle over the past 7 days; If the sudden fluctuations at night are not significantly higher than the "nighttime baseline fluctuation value", no warning will be triggered.

[0036] This mechanism effectively prevents minor abnormalities such as occasional nighttime activity, turning over, or disturbance from being misinterpreted as signs of impending labor.

[0037] 3. False alarm tolerance window mechanism.

[0038] To prevent occasional occurrences at a specific point in time from directly triggering alarms, the system sets a false alarm tolerance window (e.g., an alarm is only formally triggered if the rule conditions are met for two consecutive analysis periods): Each analysis cycle can last from 10 to 30 minutes; If the abnormal behavior pattern only appears within one cycle but does not persist, it is not considered to be in labor. A formal warning signal is triggered only when abnormal behavior occurs twice or more consecutively without being interrupted by "normal behavior".

[0039] This mechanism can filter out short-term abnormal behavior or occasional attitude fluctuations caused by environmental disturbances.

[0040] 4. Mutual verification determination under multi-axis fusion strategy.

[0041] The system is configured to require that at least two of the axes' behavior rules (such as Yaw+Pitch or Pitch+Roll) be satisfied before entering the "early warning calculation channel" to enhance the system's anti-interference capabilities and recognition accuracy.

[0042] By introducing the above-mentioned misjudgment handling mechanism, the present invention can significantly reduce the following in actual deployment: False alarms due to individual differences and highly active cattle; Occasional abnormalities caused by nighttime postural disturbances or feeding interventions; The problem of excessive alarms caused by the system's high sensitivity.

[0043] At the same time, it ensures that effective early warnings can be triggered with high confidence during the high-risk stage of birthing, supporting precise intervention and intelligent response in aquaculture management. Attached Figure Description

[0044] Figure 1 System module structure diagram; Figure 2 Three-axis attitude angle interpretation diagram; Figure 3 Overall flowchart of the method; Figure 4 A structured comparison diagram of "behavioral characteristics – posture angle changes – rule determination"; Figure 5 A chart comparing "cattle breeds – angular characteristics – adaptation schemes". Detailed Implementation

[0045] 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.

[0046] Example 1: Recognition of Frequent Turning in Place

[0047] Behavioral description: Before giving birth, dairy cows often become restless and turn around in place, exhibiting frequent head turning from side to side while moving their bodies slowly but in limited space, showing obvious characteristics of turning in place.

[0048] Attitude angle characteristics: The yaw angle exhibits continuous and rapid positive and negative fluctuations, with each change ranging from ±30° to ±90°.

[0049] Decision logic: Using a 30-minute analysis window, the cumulative change in ΔYaw exceeded 1080°, which is equivalent to the cow's head turning left and right more than 3 times. Combined with the lack of significant movement in the body position, it was judged to be the early stage of labor agitation.

[0050] Example 2: Continuous Head-Up Search Behavior Recognition

[0051] Behavioral description: Some dairy cows will keep looking up at the outside of the pen or the ceiling before giving birth, accompanied by deep breathing and a decrease in their sense of security, exhibiting a "wait-and-see" state.

[0052] Attitude angle characteristics: The pitch angle should be maintained above +20° for more than 5 minutes, with the head held high. The movement rhythm should be slow but steady.

[0053] Decision logic: If the pitch angle remains high for more than a set threshold (e.g., +20° for more than 300 seconds) within a continuous time window, combined with the nighttime period or the long-term resting state of cattle, it constitutes a medium-level calving warning.

[0054] Example 3: Recognition of frequent lying down and struggling to roll over

[0055] Behavioral description: Before giving birth, dairy cows will frequently try to lie down, stand up, or roll over from side to side, showing obvious discomfort and positioning movements.

[0056] Attitude angle characteristics: The roll angle fluctuates frequently within a unit of time, with an angle change range of ±15° or more, and the number of switching times exceeds 20 times per hour.

[0057] Decision logic: If the roll angle changes by more than ±15° ≥ 20 times within 60 minutes, accompanied by slight disturbances in the yaw and pitch angles, it is considered a significant prenatal posture adjustment behavior, and real-time monitoring is recommended.

[0058] Example 4: Three-axis joint noise recognition

[0059] Behavioral description: Dairy cows enter a highly sensitive period within 6 hours before calving, and may exhibit a restless state with simultaneous fluctuations in the yaw, pitch, and roll axes, such as pacing in place, turning their heads, tilting their heads back, and swaying slightly.

[0060] Attitude angle characteristics: The Yaw, Pitch, and Roll axes all exceeded twice their respective historical average fluctuations within 5 minutes, exhibiting characteristics of synchronous and violent fluctuations.

[0061] Decision logic: When the multi-axis threshold exceeds the limit, the system marks it as a "cooperative critical behavior window" and regards it as a high-priority early warning node, which can serve as a key signal 2–4 hours before delivery.

[0062] Example 5: Identification of Active Mutations at Night

[0063] Behavioral description: When a cow that was previously quiet at night suddenly exhibits active head movements, such as tilting its head back, turning its head, or shaking its head, it may be entering the countdown to calving.

[0064] Attitude angle characteristics: The three-axis angles fluctuated significantly at night, with the average fluctuation being more than 2.5 times the historical value of the past 7 nights.

[0065] Decision logic: The three-axis attitude angle fluctuation index at night is much higher than the nighttime average baseline. At the same time, no human interference or abnormal lighting factors are detected. This can be regarded as autonomous activity in the labor stage, and the system outputs a nighttime priority warning prompt.

[0066] Example 6: Identification of Quiet Farrowing in Beef Cattle—Lying Still and Not Moving

[0067] Behavioral description: Some beef cattle do not show obvious agitation before calving, but instead lie still for a long time with their heads slightly lowered, exhibiting a "quiet waiting" behavior pattern.

[0068] Attitude angle characteristics: The pitch angle exhibits a slight downward tilt (approximately −10° to −25°) and lasts for more than 2 hours; the yaw angle fluctuates very little (within ±10°), and the roll angle is close to zero.

[0069] Decision logic: If the three-axis angles remain stable within a continuous window (e.g., 2 hours) and the fluctuation range is far below the standard deviation of the individual's daily active period, the system marks it as "quiet prepartum type" and prompts for inspection.

[0070] Applicable scenarios: Beef cattle farms and free-range areas are suitable for cattle herds that are docile and not prone to exhibiting abnormal behavior.

[0071] Example 7: Recognition of Side-lying and Rolling Behavior in Water Buffalo under High Humidity Environment

[0072] Behavioral description: When water buffalo are about to give birth in high-humidity environments such as paddy fields and muddy areas, they often exhibit side-lying and unstable lying postures, with frequent changes in their posture angles and a cushioning effect.

[0073] Attitude angle characteristics: The roll angle varies widely (above ±25°) and exhibits alternating "plateau period" and "abrupt change period" phenomena; the pitch angle undergoes abrupt positive and negative changes during the process of getting up from a side-lying position.

[0074] Decision logic: Set a side-lying judgment area (e.g., Roll ≥ ±25° and last for >180 seconds), and when the pitch angle rises sharply (>30°), it is judged as an attempt to roll over and get up, and the system generates a "muddy birth pre-birth warning".

[0075] Applicable scenarios: Water buffalo farming environments include hilly areas, paddy fields, and wetland pastures in the south.

[0076] Example 8: Recognition of calf rolling over and adjusting behavior before birth

[0077] Behavioral description: Some cows, before giving birth, will intermittently roll from side to side or push off the ground with their hooves to adjust their position, which helps the fetus descend smoothly.

[0078] Attitude angle characteristics: The roll angle changes rapidly by approximately ±20°, while the pitch and yaw angles also exhibit slight synchronous fluctuations (e.g., yaw ±15°, pitch ±10°).

[0079] Decision logic: The number of times the Roll angle was greater than ±15° was detected to be more than 10 times within 30 minutes. Combined with short-term synchronous fluctuations in Yaw / Pitch, this was identified as a characteristic behavior of pre-labor position adjustment.

[0080] Applicable scenarios: It can be used to identify calving in breeding cows and provide early warning and assistance for fetal position deviation.

[0081] Example 9: Risk Identification of "Excessive Rest" During Farrowing in Fattening Cattle

[0082] Behavioral description: Some fattening cows are quite heavy and are reluctant to move before calving. If they lie down for a long time, they are prone to problems such as suffocation and insufficient labor.

[0083] Attitude angle characteristics: The Yaw, Pitch, and Roll angles fluctuate very little over a long period of time, and are less than 20% of the individual's historical daily average.

[0084] Decision logic: If the standard deviation of the attitude angle is lower than the set threshold within a set window (e.g., 4 hours), and combined with the reproductive status information, the system will prompt "risk of excessive stillness before labor" and recommend mandatory inspection or intervention.

[0085] Applicable scenarios: High-density fattening sheds, restricted feeding areas, and facilities equipped with automatic monitoring systems.

[0086] Example 10: Recognition of Body Licking Attempts

[0087] Behavioral description: In the early stages of calving, cows may lick certain parts of their bodies (such as the abdomen or hind legs) to relieve discomfort, which is manifested by frequently turning their heads and lowering their heads with a tilting or tilting motion.

[0088] Attitude angle performance: The Yaw angle is characterized by repeated deflection to one side (±45°~±90°); Roll angle is characterized by a significant head tilt (displacement of more than ±20°); This combination of movements is repeated more than 15 times within a given time period (10 minutes).

[0089] Decision logic: If the above-mentioned angular characteristics are met and are significantly higher than the individual's daily body-licking frequency, they are considered potential pre-labor stress signals and included in the pre-labor abnormal behavior scoring model.

[0090] Example 11: Identification of Group Boundary Loitering Behavior

[0091] Behavioral description: In pens or pastures, some cows will gradually move away from the group before giving birth, appearing alone at the edge of the pen or near the fence, slowly moving or wandering around.

[0092] Attitude angle performance: The Yaw angle remains biased in one direction for a long time (e.g., +60° ±5°). The angle change is small, with a fluctuation range of <10°, but the duration exceeds 10 minutes; Their actions are unidirectional, and their behavioral patterns tend to be fixed in their observation direction.

[0093] Decision logic: Combining location data (if optional) can further confirm that it is located in a non-group center area. Yaw stable paranoia angle + behavioral space edge location can be marked as a **"group edge wandering labor signal"**, which is one of the medium-level warning behaviors.

[0094] Example 12: Recognition of looking up into the distance / high alertness behavior

[0095] Behavioral description: Before calving, some cows may become highly alert due to discomfort or instinctive vigilance, and will maintain a high level of attention to their external environment, which is manifested by continuously tilting their heads back and staring at the direction outside the pen.

[0096] Attitude angle performance: The pitch angle rose significantly, exceeding +30°, and lasted for more than 5 minutes; During this process, the fluctuations in the Yaw and Roll angles were not significant, and the posture remained stable. This is accompanied by other sedentary behaviors, such as looking around before lying down.

[0097] Decision logic: The system defines a pitch that remains at a high angle for more than a set time threshold (e.g., 5 minutes) as a candidate behavior for impending labor. This feature can be used as a static behavioral pattern to supplement dynamic recognition rules.

Claims

1. A method for recognizing the pre-partum behavior of cattle based on gyroscope attitude angle changes, characterized in that, The method includes the following steps: (1): The sensor device with integrated three-axis gyroscope is fixed to the head of the cow to collect the three-axis attitude angle data of the cow, including yaw, pitch and roll. (2): Record the sequence data of the three-axis attitude angle changing with time according to the set sampling period, which is adjustable from 5 to 10 seconds; (3): Set an early warning analysis time window, and statistically analyze the key behavioral characteristic parameters of the three-axis attitude angle within the time window, including but not limited to the cumulative change value of the yaw angle, the duration of the pitch angle, and the frequency of the roll angle switching; (4): Compare the feature parameters with the cattle's pre-partum behavior recognition rules to determine whether the animal has entered a high-sensitivity state of pre-partum behavior; (5): Output a labor warning signal based on the recognition result for terminal display or for use in the automated management system response.

2. The method according to claim 1, wherein the three-axis gyroscope sensor is fixed to a cow ear tag or other wearable device that is synchronized with the head posture.

3. The method according to claim 1, wherein the identification rule includes any one or more of the following: (1) The cumulative change of the Yaw angle within 30 minutes is greater than 1080°; (2) The pitch angle is maintained above +20° for more than 10 minutes; (3) The Roll angle deviates by more than ±15°, and the number of switching times is greater than 20 times / hour; (4) The three-axis attitude angles fluctuate violently synchronously within a unit of time, exceeding twice the historical average threshold.

4. The method according to claim 1, further comprising: A baseline model of individual behavior was constructed based on historical data of posture angle changes in cattle outside of calving season. By comparing real-time data with individual baselines, individual-level early warnings are triggered first when attitude angle changes significantly deviate from their historical fluctuation range.

5. The method according to claim 1, wherein the early warning mechanism adopts a multi-feature weighted scoring model, calculates the labor risk score based on indicators such as the change amplitude, duration, and fluctuation frequency of each axis angle, and sets thresholds to trigger high, medium, and low level early warning responses.

Images