This study presents an advanced crosswalk guidance technology created to let visually impaired persons (VIPs) move around safely and independently when they’re outside, as the proposed system combines a CNN with GA and FLC into one system. The contributions of this research are an innovative modern approach to detecting obstacles static and moving using CNN and used Genetic Algorithms(GA) with Convolutional Neural Networks (CNN) to segment and identify the obstacles clearly, measure the distance between the person and object up to 40 m using computer vision methods, and improve the identification and tracking of moving object using Region of Interest (ROI) and Kalman Filter, and FLC defines the most effective moving path to movement guide and avoid obstacles, Also, using the effectiveness of transfer learning architectures like SqueezeNet for the detection of moving object such as cars and persons by using transfer learning by fine-tuning pre-trained CNNs architectures was highly effective in classification images. The models achieved an accuracy of training at 99.85%, validation at 100%, and testing at 98.92%. Finally, the system provides voice guidance based on FLC results.