Today's wheelchair technology is evolving, with options ranging from manual wheelchairs, electric wheelchairs to the latest smart wheelchairs. However, wheelchairs on the market today still lack of the resources to facilitate advanced security for their users, particularly from the risk of injury or death caused by user neglect to differences in the rough and uneven texture surface structure and distinctions in road height levels when crossed by wheelchairs. Which would be at risk of accidents such as the existence of descending stairs. The victim may sustain physically visible physiological effects because of the accident, such as abrasions, bruises, tears, fractures, head injuries, and even death in fatal cases. The goal of this research is to enhance the security and safety of smart wheelchairs by developing autonomous controls with the extraction of gray level cooccurrence matrix (GLCM) texture features and the Naive Bayes Classification in anticipation of irregular road conditions and the existence of levels that are at risk of endangering wheelchair users based on camera input. According to the results of 100 experiments using 2-dimensional imagery data tested at d = 1, 2, 3, 4 and θ = 0°, 45°, 90°, 135° values, the resulting d = 1 and θ = 0° levels have the highest accuracy of 87% when classifying images of descending stairs and floors.