Road safety is a global challenge, with most traffic accidents being attributed to human errors and recklessness, such as distractions and bad decision-making. ADAS play a crucial role in reducing traffic accidents caused by these mistakes, improving road safety. Computer Vision and Artificial Intelligence are essential for perception in mobile robots and autonomous vehicles, enabling them to analyze and interpret their surroundings. This paper presents a systematic review of Computer Vision and Artificial Intelligence techniques for object detection and safe navigation zone identification in Autonomous Vehicles and ADAS. The analysis compares the selected state-of-the-art works, highlighting their methods, accuracy, strengths, and limitations. In addition, the review compares multiple benchmark datasets employed across the studies, analyzing their characteristics and application scope in perception tasks. The results show advancements and considerable accuracy in object detection and safe navigation zone identification using semantic segmentation and bounding box methods, together with monocular depth estimation and multitask models that increase 3D perception and efficiency. Some approaches are able to run on embedded devices and achieve considerable FPS rate, allowing soft real-time performance. However, challenges like high computational costs and adverse weather conditions remain, showing that sensor fusion along with future research is necessary to improve robustness.