Due to the numerous societal advantages of autonomous vehicles, which include improved safety for passengers and comfort, decreased fuel expenditure, reduced travel congestion and accident, cost savings, and increased dependability, their acceptance has steadily increased in recent years. Nonetheless, there exist several functionality issues in autonomous vehicles that necessitate rectification prior to their widespread adoption on public roads. One of the most formidable challenges in developing reliable autonomous vehicles is pedestrian detection, primarily due to two factors: (i) Deformation and occlusion and (ii) the level of visual quality and spectral diversity. However, recent advancements in machine learning technology have demonstrated potential in ameliorating the ability of autonomous vehicles to accurately recognize pedestrians. In this research, a novel hybrid approach—K-Nearest Neighbour (KNN) integrated with Gaussian naive Bayes (GNB) is proposed. Autonomous cars must deal with a variety of environmental factors, such as changing weather, illumination, and pedestrian appearances. By using both the statistical modelling of GNB and the flexibility of K-NN, the hybrid technique is able to adapt to these changes. For autonomous driving to be safe, this adaptability is essential. The Waymo Open Dataset, the largest of its kind, served as the benchmark for the experimental investigation. The highly precise data that 3D LIDAR technology offers gets utilized to identify people in any lighting situation. The proposed method displays superior accuracy in comparison to conventional techniques, thereby underscoring its efficacy.