.jpg)
Welcome to
Samsil Arefin Mozumder's
Profile
Introduction
"A journey from datasheet to dataset... !"
I am an Electronics and Computer enthusiast. I am deeply dedicated to Embedded Systems and Machine Vision development, such as Embedded AI, Embedded Machine Vision, IoT, and Robotics. I love to solve real-world problems through my experiments.My Projects
Intelligent Inspection System for Anomaly Detection in Mining Rail Infrastructure (Master's Thesis)
Abstract: This thesis develops an intelligent two-part inspection system to enhance safety in underground coal mines by detecting rail infrastructure anomalies. A mobile transmitter unit, mounted on rail vehicles, uses a Raspberry Pi, dual cameras, and an IMU sensor to perform real-time monitoring. A YOLOv5n deep learning model, trained on a custom dataset of 1446 images, achieves 97% accuracy in detecting ceiling hazards like falling debris in steel net-protected shafts. For derailment detection, a separate model classifies vehicle states with 96% precision and 98% recall using 2195 annotated images, while integrated IMU data provides motion context via the Madgwick algorithm. Data is wirelessly transmitted via LoRa to a stationary receiver, which converts it to MODBUS protocol for integration into existing mine management systems. The system was experimentally validated for real-time operation, with results visualized through a user-friendly interface. Future work aims to expand anomaly detection capabilities, improve robustness, and conduct further field testing. Overall, this integrated solution represents a significant advancement in improving operational safety and efficiency in harsh mining environments.
A Gaussian Process-Enhanced Non-Linear Function and Bayesian Convolution–Bayesian Long Term Short Memory Based Ultra-Wideband Range Error Mitigation Method for Line of Sight and Non-Line of Sight Scenarios. (Mathematics MDPI)
Abstract: Relative positioning accuracy between two devices is dependent on the precise range measurements. Ultra-wideband (UWB) technology is one of the popular and widely used technologies to achieve centimeter-level accuracy in range measurement. Nevertheless, harsh indoor environments, multipath issues, reflections, and bias due to antenna delay degrade the range measurement performance in line-of-sight (LOS) and non-line-of-sight (NLOS) scenarios. This article proposes an efficient and robust method to mitigate range measurement error in LOS and NLOS conditions by combining the latest artificial intelligence technology. A GP-enhanced non-linear function is proposed to mitigate the range bias in LOS scenarios. Moreover, NLOS identification based on the sliding window and Bayesian Conv-BLSTM method is utilized to mitigate range error due to the non-line-of-sight conditions. A novel spatial–temporal attention module is proposed to improve the performance of the proposed model. The epistemic and aleatoric uncertainty estimation method is also introduced to determine the robustness of the proposed model for environment variance. Furthermore, moving average and min-max removing methods are utilized to minimize the standard deviation in the range measurements in both scenarios. Extensive experimentation with different settings and configurations has proven the effectiveness of our methodology and demonstrated the feasibility of our robust UWB range error mitigation for LOS and NLOS scenarios.
Research on Vertical Shaft Detection System Based on CMOS-Camera and Lidar. (IEEE Access)
Abstract: This research focuses on the pivotal role of coal mining shaft safety. Deformations and fractures in the shaft wall can cause significant risks to mine operations, which require regular safety inspection to prevent hazards. Existing shaft detection systems cannot provide any prior notifications before damage occurs, with complex systems and low intelligence. To solve these problems, our proposed system has three key indicators: shaft deformation, shaft perpendicularity, and horizontal vibration of the lifting container. This intelligent system reduces the likelihood of significant shaft damage by enhancing automation, intelligence, and efficiency. The research starts by defining the essential features required for the shaft detection system, like data acquisition system, data transmission, and data processing. Hardware includes LiDAR sensors, CMOS cameras, laser collimators, control board, batteries, and software implemented on ROS. This system collects data, simulates shaft conditions, and conducts horizontal vibration displacement detection experiments. Moreover, we validated shaft deformation and perpendicularity detection under static conditions. This research concludes by analyzing the accuracy of the detection system in actual working conditions, confirming its practicality and reliability. Overall, this research solves critical issues in shaft safety and introduces an intelligent, efficient, and smart solution.
Derailment detection of mining shaft’s rail vehicle using machine vision on edge device. (Conference Paper)
Abstract: The railway system is pivotal in the mineral transportation industry. Due to rail track shape deformation or imbalance, the situation worsens. Stone, coal, and dust can also obstruct the mine rail transportation system. Moreover, derailment is very likely to happen, but conventionally, it is difficult to determine whether the vehicle is derailed. Scientists have developed different derailment detection techniques to overcome these difficulties, but most are implemented using sensors only. In challenging conditions, such as inside a mine shaft, sensor systems can fail for various reasons such as high dust levels, humidity, and fallen stones. These factors can easily destroy the sensors, especially if they are placed near to the ground surface. Machine vision, on the other hand, can detect objects from a greater distance than sensors. In our proposed system, we introduced a machine vision approach based on a Raspberry Pi, which combines YOLOv5n deep learning model and OpenCV to detect different types of derailment. Our system can be used as a rail track detection system by analyzing realtime video data.
Smart IoT-Biofloc water management system using Decision regression tree (Conference Paper)
Abstract: The conventional fishing industry has several difficulties, such as water contamination, temperature instability, nutrition, area, expense, etc. In fish farming, Biofloc technology turns traditional farming into a sophisticated infrastructure that enables the utilization of leftover food by turning it into mi-crobial biomass. The objective of our study is to propose an intelligent IoT Bi-ofloc system that improves efficiency and production. This article introduced a system that gathers data from sensors, store data in the cloud, analyses it using a machine learning model such as a Decision regression tree model to predict the water condition, and provides real-time monitoring through an android app. The proposed system has achieved a satisfactory accuracy of 79% during the expe-riment.
IRHA: An Intelligent RSSI based Home automation System (Conference Paper)
Abstract: Human existence is getting more sophisticated and better in many areas due to remarkable advances in the fields of automation. Automated systems are favored over manual ones in the current environment.Home Automation is becoming more popular in this scenario, as people are drawn to the concept of a home environment that can automatically satisfy users' requirements.The key challenges in an intelligent home are intelligent decision making, location-aware service, and compatibility for all users of different ages and physical conditions. Existing solutions address just one or two of these challenges, but smart home automation that is robust, intelligent, location-aware, and predictive is needed to satisfy the user's demand. This paper presents a location-aware intelligent home automation system that uses Wi-Fi signals to detect the user's location and control the appliances automatically. The fingerprinting method is used to map the Wi-Fi signals for different rooms, and the machine learning method, such as Decision Tree, is used to classify the signals for different rooms. The machine learning models are then implemented in the ESP32 microcontroller board to classify the rooms based on the real-time Wi-Fi signal, and then the result is sent to the main control board through the ESP32 MAC communication protocol to control the appliances automatically. The proposed method has achieved 92% accuracy in classifying the users' location.
LoRaComm: Ultrasonic Sensor Network for Real-Time Monitoring
The system features a sender unit equipped with two ultrasonic sensors and a potentiometer, all connected to an STM32 microcontroller. This unit also includes a 1278 LoRa module for long-range communication. A separate STM32 microcontroller on the receiving end is connected to another LoRa module. This setup enables effective data capture and processing by the receiver. The received values are then displayed in Modscan. This is made possible through a TTL to Modbus module, ensuring seamless communication.
Campus Automation
This project, designed specifically for the COVID-19 situation, utilizes a variety of hardware components including an Arduino Pro Mini, Ultrasonic Sensor, Bluetooth Module, 4-Channel Relay Module, Servo Motors, MLX-90614 Sensor, MQ-2 Smoke Sensor, Buzzer, 20*4 I2C LCD, PIR Sensor, Pump, IR Sensor, and a Switching Circuit. The project is divided into five key sections. The first section uses the MLX-90614 sensor to detect human temperature from a distance. If the temperature is below 98.6°F, the campus gate opens; otherwise, a sound is emitted and the gate remains closed. The second section is a touchless water tap for sanitization, activated by an IR sensor and controlled by a switching circuit and pump. The third section is a touchless dustbin that uses an ultrasonic sensor to detect a human hand and a metal gear servo to open and close the cover. The fourth section is a voice-controlled classroom automation system that uses an Android app, Bluetooth, Arduino, Relay Module, and loads. The final section is a smoke detection system that uses the MQ-2 sensor to detect smoke or gas and emits a beep sound when detected.
Face Mask Detection Based Automatic Gate
The “Face Mask Detection Based Automatic Gate” is an embedded system that leverages Google’s Teachable Machine, an accessible online machine learning tool. By training the tool with various image samples, it can distinguish between two different types of images and generate a model link. The ESP32 Camera Module, which is internet-enabled via built-in WiFi, is used in this project. The model link is integrated into the ESP32 Cam, which captures images for analysis by the Teachable Machine. Upon making a decision, the system communicates with the Arduino to open the gate and displays relevant gate indicators on an LCD and LED.
Rubik's Cube Solver Robot
This robot, designed to solve a Rubik’s cube, is constructed from PVC board and wood. It’s programmed using two languages: Python and C++. The graphical user interface (GUI), the Kociemba algorithm, and the instructions that signal the Arduino to operate the servo are all written in Python. The Arduino, in turn, receives these signals from the PC and rotates the servos as instructed. The robot is equipped with three servos: push, hold, and rotate. Additionally, a vibration motor is utilized to intelligently manipulate the cube’s movements.
IoT weather Station
The system is equipped with several hardware components: an LDR, DHT11, BMP180, and ESP32. These sensors allow the system to take five different readings: light intensity, atmospheric pressure, temperature, and humidity. The data from these sensors is processed using Arduino code written in HTML. An IP interface has been implemented to display the readings, which include luminosity, altitude, atmospheric pressure, humidity, and temperature. These values can be accessed by pasting the IP address into any device connected to a common router.
Face Recognition & Temperature Detection Based Automatic Gate
In this system, an ESP32 Cam is used for face detection and recognition. Alongside this, an Arduino is employed to gather readings from the MLX90614 IR temperature sensor. This sensor measures temperature and ensures it falls within a safe limit. The system also checks if the detected face is recognized. If both conditions are met - the temperature is within the safe range and the face is familiar - then the servos are activated to open the gate.
Automatic Bottle Filling System
The system is equipped with several hardware components: an IR sensor, a gear motor, a water pump, an Arduino, a relay, an LCD, and push buttons. The push buttons are used to set the pump’s operation time, start and stop the system, and select digits. When a bottle is placed on the conveyor, the sensor detects it, causing the conveyor to stop and the pump to start filling the bottle with water. The system then pauses briefly to allow for water droplets to fall before the conveyor starts running again. It stops once more when it reaches another sensor, which signals that the bottle can be removed from the conveyor.
Voice Control Robot
The robot operates based on five commands: Go, Back, Turn Left, Turn Right, and Stop. The hardware components of this system include an Arduino Uno, a Bluetooth Module HC-05, and an L298N. These components work together to enable the robot’s movements. An Android app is used to provide voice commands to the robot. Upon receiving these commands, the robot activates the corresponding functions.
Mini CNC Plotter
This project features a structure composed of metal and acrylic. The stepper housings are repurposed from broken DVD drives. The Z-axis is represented by a modified pen that can move its nib vertically. Movement along the X and Y axes is facilitated by the use of a stepper motor. Two drivers are employed to control these movements. Writable files from MIT processing software are transmitted to the Arduino via serial communication and plotted according to the file orientation. The stepper and servo mechanisms are electrically interconnected through the Arduino, drawing a significant amount of current. The entire structure is mounted on an acrylic base, lending precision and aesthetic appeal to the design.
.jpg)
Heat Shrink Tube Cutter Machine Modification
This project is a safety enhancement for a machine that has a removable safety cover. There’s a risk that the cover could be opened while the machine is cutting tubes, potentially leading to finger injuries. The hardware components used include a 2N3904 BJT, a 555 Timer IC, a relay, and a limit switch. A logic circuit has been designed using relays that triggers a beep sound at 1Hz frequency when the cover is open. This audible alert serves as a safety feature to prevent accidents.
Bend Product Check Tester
The hardware components used in this project include a 2N3904 BJT, resistors, a diode, a fiber sensor, and a limit switch. This design was prompted by the discovery of some of our products being bent in Japan. As a result, I developed a testing device capable of detecting bends as small as 1.2mm in the product. The circuitry of this tester was designed exclusively with BJTs and incorporates both Normally Open (NO) and Normally Closed (NC) actions. In the absence of a product on the tester, a yellow lamp is illuminated. However, when a product is placed on the tester, the yellow lamp turns off. The tester then makes a decision, indicated by either a green or red lamp, along with a buzzer. This system effectively identifies any product deformities
My Small Projects
GSM Based Irrigation System
Height Measuring Instrument
Gas Leakage Detection Device
Level Sensing Device
Obstacle Avoiding Robot
Smart Basin,Dustbin,Door
IoT Home Automation
