The paper considers the method for describing the contour of a moving object using descriptors. They were calculated taking into account the fact that they were one-dimensional, which would increase object classification speed, and invariant to parallel transport, rotation and scaling. Amplitude spectra of the Fourier transform were chosen as descriptors. Then etalon contours for comparison and classification were determined. The etalon contours are based on the terrain, the observation time for the objects and other conditions which are typical for this type of site supervision. Selection of etalon contours allowed clarifying the essential features of a particular class and starting to compare descriptions and to classify contours with their help. The authors have defined two ways of descriptor comparison: by a correlation coefficient and by lambda distances and considered the methods of implementation and compared the main advantages of the methods. The method of comparison using lambda distances was selected experimentally. The method has contributed to the most clear and qualitative separation of moving objects into classes. Then the threshold conditions were developed. They allow comparing more accurately the input contours and etalons. The conditions were divided into two types. This helped to solve real problems in moving objects classification. Using the information on contour description and the way of its comparison the authors developed the algorithm of classifying contours of moving objects.