Design of Blind Spot System as Collision Warning on Truck Vehicle Based on Internet of Things

In the era of globalization, technological advancements have accelerated significantly, leading to the development of various innovative technologies. The Internet of Things (IoT) enables everyday physical devices to become intelligent and interconnected through sensors and internet connectivity, allowing them to collect and process data, communicate with other devices, and be monitored and controlled remotely, while offering various applications and presenting challenges related to security, privacy, and future technological development [1]. A blind spot is an area surrounding a vehicle that cannot be directly observed by the driver. These areas can increase the risk of traffic accidents, particularly when drivers are changing lanes or maneuvering in confined spaces. Large vehicles, such as trucks, are especially susceptible to blind spot-related accidents because they generally have larger blind spot areas than smaller vehicles as shown in Figure 1. In trucks, blind spots are commonly found at the front, on both sides, and at the rear. Consequently, trucks pose a greater risk of collisions with other road users when these areas are not adequately monitored [2-4]. To address this issue, IoT-based sensor technology offers a promising solution for improving vehicle safety. Sensors can be installed at various positions on trucks, including the front, sides, and rear, to detect objects or other vehicles within blind spot areas. By providing information about nearby objects and potential collision hazards, such systems can help drivers make safer driving decisions and reduce the risk of accidents.

The hardware components used to develop the IoT-based truck blind spot detection system for collision warning include a Raspberry Pi, a Raspberry Pi Camera Module OV5647, a TF-Luna LiDAR sensor, and a buzzer. The Raspberry Pi is a compact, single-board computer developed by the Raspberry Pi Foundation to support computer science education and a wide range of computing applications. It provides processing capabilities and interfaces for integrating sensors, cameras, and other peripheral devices, making it suitable for embedded systems and IoT applications [6]. The Raspberry Pi Camera Module OV5647 uses a camera interface compatible with the Camera Serial Interface 2 (CSI-2) protocol. It supports image capture and video streaming for computer vision applications. The camera can capture images at a resolution of up to 5 megapixels and supports video recording at 1080p resolution at up to 30 frames per second (fps), depending on the camera configuration and software implementation [7]. The TF-Luna LiDAR sensor measures distances using the time-of-flight (ToF) principle, which determines distance based on the travel time of emitted light reflected from an object. The sensor provides distance measurements suitable for detecting nearby objects within its specified operating range, which is approximately 0.2 to 8 meters under appropriate conditions [8]. In this study, the sensor is used to measure the distance between the truck and objects detected within its blind spot areas. A buzzer is an electronic component that converts electrical signals into audible sound. In the proposed system, the buzzer serves as an audible warning device that alerts the driver when an object is detected in a predefined blind-spot area. This warning mechanism helps the driver become aware of potential collision hazards. This study uses the Hypertext Transfer Protocol (HTTP) as the communication protocol for the IoT-based truck blind spot detection and collision warning system. HTTP enables communication between clients and servers by facilitating the exchange of data and resources over a network. In the proposed system, HTTP supports communication between the Raspberry Pi and the monitoring website, allowing the system to transmit detection results and related information for remote monitoring. By implementing HTTP-based communication, the proposed system enables users to monitor detected objects and their corresponding distance measurements through a web interface, supporting the timely delivery of information relevant to vehicle safety. The software used in the development of the system were IDLE Python, Raspbian OS, YOLOv8, Autodesk Fusion 360, Microsoft Visio, Draw.io, and Fritzing.

The schematic diagram as shown in Figure 2 illustrates the hardware design of the IoT-based blind spot detection system for truck collision warning. The system uses TF-Luna LiDAR sensors to measure the distance to objects in the truck's blind spot areas, while Raspberry Pi cameras integrated with the YOLO object detection model capture video streams and identify objects within these areas. The Raspberry Pi processes the camera images and distance measurements to determine whether an object is within the predefined warning range. A buzzer serves as an audible warning device and is activated when the measured distance reaches the programmed warning threshold. The blind spot detection system consists of four Raspberry Pi cameras, four TF-Luna LiDAR sensors, one Arducam Multi Camera Adapter Board, one TFT LCD, and one buzzer. The camera circuit uses the Arducam Multi Camera Adapter, which enables the Raspberry Pi to connect to up to four cameras through 15-pin flexible flat cables (FFCs). The 15-pin camera interface includes ground (GND), camera power, camera clock, and camera data connections. The adapter also provides seven connection pins, including 3.3 V, GND, SDA, SCL, GPIO17, and GPIO18, which are connected to the corresponding pins on the Raspberry Pi to support camera selection and control. The TF-Luna LiDAR sensor circuit consists of four pins: VCC, GND, RX, and TX. The VCC pin is connected to the Raspberry Pi's 5 V power supply, while the GND pin is connected to the Raspberry Pi's ground. For serial communication, the sensor's RX pin is connected to the Raspberry Pi's TX pin, and the sensor's TX pin is connected to the Raspberry Pi's RX pin. These crossed connections enable bidirectional serial data transmission between the sensor and the Raspberry Pi. Finally, the LCD and buzzer circuits provide visual and audible feedback to the driver. The TFT LCD is connected to the Raspberry Pi through an HDMI cable to display the camera video stream and relevant detection information. The buzzer is connected to the appropriate GPIO and GND pins on the Raspberry Pi. When an object is detected within the predefined warning distance, the system activates the buzzer and displays the corresponding blind spot information on the LCD, helping the driver identify the location of potential collision hazards.

The Raspberry Pi Camera OV5647 was tested to evaluate whether it performed as expected in the IoT-based truck blind spot detection system for collision warning. The test results indicate that the camera operated properly and successfully displayed video streams from the truck's different blind-spot areas. The video streams displayed smoothly with satisfactory image quality, showing that the camera could capture and transmit visual information for blind spot monitoring.

The TF-Luna LiDAR sensor was tested to evaluate its performance in measuring object distances. Before testing, the distance between the object and the sensor was ensured to be 20 cm using a tape measure. During testing, as shown in Figure 3, the distance was displayed in the terminal shell. The results showed a measured distance of 20 cm, indicating that the sensor was able to detect the distance to an object at that measurement point. The displayed distance data confirmed that the sensor and its communication interface were functioning during the test. .

Real-world testing was conducted to evaluate the performance and effectiveness of the developed prototype under actual operating conditions. The TF-Luna LiDAR sensor was installed at a predetermined position on the truck to measure the distance to objects located within the driver's blind spot areas. During testing, the sensor successfully measured object distances, and the corresponding distance data were displayed on the LCD screen. As shown in Figure 4, the measured distance between the truck and the vehicle in front of it (depan) was 42 cm. This measurement provides distance information that can be used by the system to determine whether the object is within the predefined warning range. When the measured distance meets the programmed warning condition, the system activates the buzzer to alert the driver to a potential collision hazard.

Real-world testing was conducted to evaluate the performance of the Raspberry Pi Camera OV5647 when integrated into the truck blind spot detection prototype. The camera module was installed at a predetermined position facing the blind spot area to capture video streams of the surrounding environment. The YOLOv8 object detection model was integrated into the system to assist in identifying vehicles and other objects within the camera's field of view. During testing, the camera successfully captured and displayed video streams on the LCD screen as shown in Figure 5 (kanan). YOLOv8 processed the captured images and identified objects by generating bounding boxes around them. Each bounding box included a label indicating the detected object's class, such as a car or motorcycle. This visual information helps the driver identify objects within the blind spot area more easily. The test results demonstrate that the Raspberry Pi Camera OV5647 and the YOLOv8 model functioned together to provide visual monitoring and object detection. The displayed video streams, bounding boxes, and object labels support the identification of potential hazards around the truck. When integrated with the TF-Luna LiDAR sensor, the camera-based detection system can contribute to a more comprehensive blind spot monitoring and collision warning system.

The research on the design of an IoT-based blind spot detection system for truck collision warning successfully integrated hardware and software components to support system operation under both simulated and real-world conditions. The Raspberry Pi Camera, integrated with the YOLOv8 object detection model, was able to detect objects within the truck's blind spot areas, display bounding boxes, and assign labels to detected objects, such as motorcycles and cars. The TF-Luna LiDAR sensor was also able to measure object distances and provide distance data for the collision warning system. When an object entered the predefined danger zone, the system activated the buzzer to alert the driver to a potential collision hazard. The prototype demonstrated its ability to provide visual and audible warnings to support driver awareness of surrounding objects. In addition, a protective enclosure and a cooling fan were installed to protect the hardware components and help prevent the Raspberry Pi from overheating. The integration of these components enabled the system to perform its intended blind spot monitoring and collision warning functions effectively under the tested conditions.