A CNN -based Crosswalk Guidance for the Visually Impaired that Combines a Fuzzy Logic Controller and Genetic Algorithm

This study focuses on the development of a support system for visual impairments that combines computer vision techniques, Convolutional Neural Networks (CNN), Fuzzy Logic Controller (FLC), and Genetic Algorithms (GA). The proposed system combines the advantages of three intelligent logic systems: CNN, FLC, and GA. This integrated system, named the NFG (Neural Fuzzy Genetic) system, aims to detect and track objects, measure the distance between obstacles and the blind or visually impaired (BVI) person, and provide movement guidance using three ultrasonic sensors with FLC. The combination of these technologies provides a novel approach to improving the mobility and safety of individuals who are Blind or Visually Impaired (BVI). In particular, the use of object detection and tracking through CNN is implemented, with the range for obstacle detection extending up to 40 meters. The obstacle recognition system is trained on the ResNet50 model, comprising 50 million trained images and over 1000 classifications of obstacles, resulting in high accuracy in identifying obstacles. This speed in identifying, avoiding, and guiding ensures efficient navigation towards the intended destination. FLC is then utilized to provide guidance that considers uncertainty in the information used for guidance such as near, far. It assists in making appropriate decisions in the event of obstacles, navigating around them, and determining directions for movement in obstacle-free paths. The system moves at a steering angle determined by fuzzy rules. The information from CNNs and sensors with FLC is integrated to optimize guidance information using GA. This optimization process fine-tunes the rules of FLC, aiming to find the best possible solutions for effective guidance.