Real-Time Object Detection using Convolution Neural Network on Raspberry Pi Embedded Systems

Object detection through colour filtering of pixel hue and saturation values is effective for objects lacking consistent distinguishable features. Traditional object recognition frameworks, however, are often passive and neglect the relationship between detection configuration and recognition performance, as well as the importance of feedback for image quality improvement. This limitation, combined with the inherent challenges of human colour detection, particularly for those with colour vision deficiencies, can impact tasks in sorting areas. To address these challenges, this project focuses on developing a real-time object detection system using convolutional neural networks (CNNs). The system integrates a camera detector, Raspberry Pi, servo motor, 5 V DC motor, infrared obstacle avoidance sensor and Liquid-Crystal Display (LCD) display. The dataset comprises 660 augmented chili images in red, green and yellow, used for both training and testing. During the final training epoch, the system achieved 100 % accuracy for both the training and validation datasets, demonstrating the efficacy of CNNs in real-time object detection and classification. To enhance the dataset, four types of image augmentations were applied: rotation along the centre, brightness increments and decrements and image sharpening. These augmentations increased the dataset size and improved the robustness of the neural network. The validation accuracy was only 3 % lower than the training accuracy, indicating minimal overfitting. This project highlights the advantages of using CNNs for precise colour-based object identification, addressing the inefficiencies of traditional methods and the limitations of human visual perception. By integrating advanced image processing techniques and machine learning algorithms, the system provides a robust and reliable solution for real-time object detection, applicable in various industrial sectors. This approach not only improves the accuracy and efficiency of object detection but also offers a practical solution for individuals with colour vision deficiencies, making it a valuable tool for diverse applications.