Preserving the Tetragonula biroi

An advanced IoT & Machine Learning monitoring system designed to protect native stingless bee colonies through non-invasive acoustic and visual analysis.

Audio-Video Analysis for Colony Activity and Intrusion Alert for Tetragonula biroi Utilizing Machine-Based Learning

Abstract:

This study developed an audio-video monitoring system utilizing machine-based learning to analyze colony activity and detect intrusions in Tetragonula biroi (stingless bee) hives. The system integrates sensors and embedded devices, including an ESP32-S3 microcontroller, Raspberry Pi 4 Model B, temperature and humidity sensor, microphone, load cell, and Raspberry Pi Camera Module 3, to collect environmental, acoustic, and visual data from the hive. The collected data are processed to monitor hive conditions and identify abnormal activities or potential predator intrusions in real time.Testing was conducted through laboratory evaluation and field deployment using an active T. biroi hive.

Results showed reliable environmental monitoring, stable sensor performance, and consistent data transmission. The machine learning-based intrusion detection system achieved high detection reliability, with accuracy reaching up to 85% and real-time alert notifications delivered within approximately 3–4 seconds. A web-based dashboard was also implemented to allow users to monitor hive conditions and receive alerts across multiple devices.

The system demonstrated fast response times, reliable alert delivery, and stable operation during field testing. Overall, the proposed system provides a non-invasive and automated monitoring solution that can help beekeepers detect hive disturbances early and improve colony management.

Keywords: Stingless bee (Tetragonula biroi), machine learning, audio-video analysis, hive monitoring, intrusion detection, precision apiculture

System Capabilities

BEEROI integrates an ESP32-S3 and Raspberry Pi 4B to process multi-modal data from SHT30-B environmental sensors, HX711 weight telemetry, and INMP441 microphones. It utilizes YOLOv8 and CNN architectures for real-time predator detection and acoustic classification, delivering automated alerts to a web-based dashboard via a hybrid MQTT and Socket.IO framework.

Download Full Paper (PDF) Read Full Paper Online

In special collaboration with Hardin sa Parang

Quezon City University • BS Electronics Engineering

Academic Year 2025 - 2026

Proponents:
Aguinido, M.P. • Barela, D.J.L. • Espiritu, G.T.C. • Gironella, J.C.C. • Obumani, M.D.B. • Rabonza, S.D.P.

Hello! 👋

Authorized Personnel

Brood Temp

Target: 30°C

Humidity

Range: 60-70%

Hive Weight

Growth: +2%

Acoustics

Normal

Listening

Entrance

Active

● Live

Colony State normal_activity

Normal Activity Waiting for analysis...

Audio AI No Alert

normal • 0% Always-on acoustic monitoring active

Video AI Watching

unknown • 0% Entrance object detection standby

Microclimate Trends

Real-time

System Activity Log

Timestamp	Event Description	T/H/W Sensors	ML Conf.	Status

Preserving the Tetragonula biroi

The Native Stingless Bee

Forest Pollinators

Medicinal Pot Honey

Propolis Defense

Audio-Video Analysis for Colony Activity and Intrusion Alert for Tetragonula biroi Utilizing Machine-Based Learning

System Capabilities

Hello! 👋

Brood Temp

Humidity

Hive Weight

Acoustics

Entrance

Microclimate Trends

System Activity Log

Keeper Profile

Dashboard Configuration

Data Refresh Rate

Alert Sensitivity

Audio Notifications

The Native Stingless Bee

Forest Pollinators

Medicinal Pot Honey

Propolis Defense

Audio-Video Analysis for Colony Activity and Intrusion Alert for Tetragonula biroi Utilizing Machine-Based Learning

System Capabilities

Resources

BEEROI Login

Brood Temp

Humidity

Hive Weight

Acoustics

Entrance

Microclimate Trends

System Activity Log

Keeper Profile

Dashboard Configuration

Data Refresh Rate

Alert Sensitivity

Audio Notifications

Detected Events

Audio Visualizer

Audio Class Scores

Events

🚨 INTRUSION DETECTED!