AbstractThe vehicle operation condition for a certain kind of vehicle is used to describe the relationship between vehicle speed and time in a specific traffic environment. The experimental group uses the acquisition device dash cam to record the driving data of a light vehicle in Shanghai for research to quickly and accurately build the operation condition of the vehicle. The information of three different periods of time for three consecutive times is also collected with the sampling frequency of 1 Hz. We use smooth processing, elimination, attribution, and sliding window method to preprocess the huge data of vehicle operating condition for studying the driving condition of the automobile. A total of 724 time discontinuities are divided into three types for different processing, smooth processing with smooth function of moving average filter to eliminate the abnormal value of speed, and idle processing with long-term parking. Thereafter, the method of sliding window is used to process the idle speed data. Then, the kinematic segments are cut, and the feature values are extracted. Finally, the dimension of 13 features is reduced to four principal components. On the basis of three methods, such as contour coefficient, K-means clustering analysis method is used to continuously verify and divide the dataset into three categories: namely, stop and go, high-speed driving, and low-speed driving, which are consistent with the reality. According to the time proportion of each category, the most representative 15 segments are selected to construct the 1254 s long steam. The error rate of the result is less than 10%. The construction model of the vehicle operation condition is very detailed for the data-processing process of the vehicle operation condition, and the model constantly seeks optimization on the basis of the initial clustering center. This work provides a reference for the establishment of the vehicle performance research in line with the traffic and actual operation conditions of specific cities in China.