In order to study the operation characteristics of vehicle deceleration at the crossroads, a drone was used to collect the video of road intersections in a mountainous city in this paper. Firstly, the Lucas-Kanade and background difference algorithms were developed to extract the vehicle data in the videos through the cloud platform DataFromsky AI. Secondly, the vehicle speed variation characteristics and the distribution law of parking points are analyzed according to different slope types during parking deceleration. The results show that in different slope sections, the closer the vehicle is to the stop line, the greater the initial velocity of the area and the higher the dispersion degree. The velocity at the stop line has a strong positive correlation with the stop distance. The vehicle section's initial speed, deceleration rate, and velocity decrease in the uphill section are smaller than those in the downhill section. The closer the distance between the vehicle and the stop line, the more concentrated the distribution of the stopping position. The location of vehicle parking spots in the downhill section is more scattered, the distribution interval is more significant, and the parking distance gradually increases. The results of this study can provide theoretical support and data support for the calibration and correction of simulation model parameters of road intersections in mountain cities.