AI Livestock Monitoring System Development

Development of an AI system for monitoring livestock production

Large livestock farms—with 500+ head of cattle or 10,000+ pigs—don't allow a veterinarian to examine every animal daily. IoT monitoring systems with AI analyze behavioral sensors, video, and feeding data to address three key issues: early disease detection, heat/calving, and feed optimization.

Sensor infrastructure

Typical sensors on an animal:

sensor_types = {
    'ear_tag_accelerometer': {
        'measures': ['activity_count', 'rumination_time', 'lying_time'],
        'vendors': ['SCR (Merck Animal Health)', 'Allflex SenseHub', 'Moocall'],
        'frequency': '2-min aggregates'
    },
    'rumen_bolus': {
        'measures': ['rumen_pH', 'rumen_temperature', 'rumen_motility'],
        'vendors': ['SmaXtec', 'Medria'],
        'frequency': '10-min readings'
    },
    'milk_meter': {
        'measures': ['yield_kg', 'conductivity', 'flow_rate', 'milk_temperature'],
        'vendors': ['DeLaval ALPRO', 'Lely T4C'],
        'frequency': 'per milking'
    },
    'weigh_scale': {
        'measures': ['body_weight_kg'],
        'frequency': 'daily at feeding station'
    }
}

Estrus Detection

Algorithm for activity and rumination:

import pandas as pd
import numpy as np
from scipy.signal import find_peaks

def detect_estrus(animal_id: str, sensor_data: pd.DataFrame,
                  baseline_days: int = 21) -> dict:
    """
    Охота у КРС: резкий рост активности + снижение руминации
    Цикл ~21 день — паттерн периодический
    """
    recent = sensor_data[sensor_data['animal_id'] == animal_id].tail(baseline_days * 24)

    # Нормировка по индивидуальному baseline
    activity_baseline = recent['activity_count'].quantile(0.5)
    rumination_baseline = recent['rumination_time'].quantile(0.5)

    current_24h = sensor_data[
        sensor_data['animal_id'] == animal_id
    ].tail(12)  # последние 2 часа (12 × 10-мин блоков)

    activity_ratio = current_24h['activity_count'].mean() / (activity_baseline + 1e-9)
    rumination_ratio = current_24h['rumination_time'].mean() / (rumination_baseline + 1e-9)

    # Охота: активность выше нормы в 2+ раза, руминация снижена
    estrus_score = activity_ratio * (2 - rumination_ratio)

    # Дополнительно: проверяем повторение 21 день назад
    prev_cycle_score = check_previous_cycle(animal_id, sensor_data, days_back=21)

    return {
        'animal_id': animal_id,
        'estrus_score': float(estrus_score),
        'estrus_detected': estrus_score > 2.5,
        'confidence': 'high' if prev_cycle_score > 2.0 else 'medium',
        'recommended_action': 'insemination_window_12-18h' if estrus_score > 2.5 else None
    }

Early detection of mastitis

Milk conductivity is the main signal:

def mastitis_risk_score(milking_data: pd.DataFrame, animal_id: str) -> float:
    """
    Мастит → воспаление → рост ионов Na+, Cl- → рост электропроводности
    Сравниваем четверти вымени между собой и с историей
    """
    today_milking = milking_data[
        (milking_data['animal_id'] == animal_id) &
        (milking_data['date'] == milking_data['date'].max())
    ]

    if today_milking.empty:
        return 0.0

    quarters = ['LF', 'RF', 'LR', 'RR']  # левая передняя, правая передняя, etc.
    conductivities = {q: today_milking[f'conductivity_{q}'].mean()
                      for q in quarters if f'conductivity_{q}' in today_milking.columns}

    if len(conductivities) < 2:
        return 0.0

    mean_cond = np.mean(list(conductivities.values()))
    max_deviation = max(abs(v - mean_cond) for v in conductivities.values())

    # Порог: >0.5 мСм/см отклонение от среднего = подозрение
    relative_deviation = max_deviation / (mean_cond + 1e-9)

    # Также: снижение удоя на четверти
    yield_deviation = check_yield_drop(animal_id, milking_data, today_milking)

    risk_score = 0.6 * (relative_deviation / 0.1) + 0.4 * yield_deviation
    return min(1.0, risk_score)

Additional signs of mastitis:

• Increase in milk temperature in the quartet (+0.5°C)
• Flocculation - change in flow rate pattern
• Decrease in overall milk yield while maintaining appetite
• Change in behavior on the milking machine (connection time)

Health monitoring by rumination

Rumen acidosis (SARA) - decreased pH:

def detect_sara_risk(rumen_bolus_data: pd.DataFrame, animal_id: str) -> dict:
    """
    SARA (Sub-Acute Ruminal Acidosis): pH < 5.8 более 3 часов в сутки
    Данные от rumen bolus SmaXtec обновляются каждые 10 минут
    """
    today_data = rumen_bolus_data[
        (rumen_bolus_data['animal_id'] == animal_id) &
        (rumen_bolus_data['date'] == pd.Timestamp.today().date())
    ]

    low_ph_minutes = len(today_data[today_data['rumen_pH'] < 5.8]) * 10

    # Снижение руминации + низкий pH = SARA
    avg_ph = today_data['rumen_pH'].mean() if not today_data.empty else 6.5
    rumination_today = today_data['rumen_motility'].mean() if not today_data.empty else 1.0

    sara_risk = 'high' if low_ph_minutes > 180 else (
        'medium' if low_ph_minutes > 60 else 'low'
    )

    return {
        'animal_id': animal_id,
        'low_ph_hours': round(low_ph_minutes / 60, 1),
        'avg_ph': round(avg_ph, 2),
        'sara_risk': sara_risk,
        'action': 'adjust_roughage_ratio' if sara_risk == 'high' else None
    }

Group analysis and feeding optimization

Automatic formation of production groups:

from sklearn.cluster import KMeans

def form_feeding_groups(herd_data: pd.DataFrame, n_groups: int = 4) -> dict:
    """
    Группировка КРС для точного кормления (TMR-рационы)
    Признаки: ДИМ (дней в молоке), удой, BCS (упитанность), вес
    """
    features = ['days_in_milk', 'milk_yield_7d_avg', 'body_condition_score', 'body_weight']
    X = herd_data[features].dropna()

    kmeans = KMeans(n_clusters=n_groups, random_state=42, n_init=10)
    herd_data.loc[X.index, 'feeding_group'] = kmeans.fit_predict(X)

    group_recommendations = {}
    for group_id in range(n_groups):
        group = herd_data[herd_data['feeding_group'] == group_id]
        group_recommendations[group_id] = {
            'n_animals': len(group),
            'avg_days_in_milk': group['days_in_milk'].mean(),
            'target_energy_NEL': calculate_nel_requirement(group),
            'crude_protein_pct': calculate_cp_requirement(group)
        }

    return group_recommendations

Dashboard and alerts: The system generates a daily list of priority animals for inspection: the top 10 by composite risk score (mastitis + disease + heat). Integration with herd management software: DairyComp 305, Herd Navigator, SAP Animal Management.

Timeframe: Estrus + mastitis detection + basic dashboard – 4-5 weeks. SARA monitoring, group feeding, predictive disease models, and DairyComp integration – 2-3 months.