Determining the Dynamic Characteristics of Moving Vehicles Using Image Blur Analysis

Abstract

A method to recognize vehicles on an image and to determine their dynamic characteristics is considered. Control of timely determined characteristics will allow detecting violations of vehicles, including excessive vehicle weight, which leads to increased wear and damage to the roadway. For object detection the convolutional neural network is used. Hough and Canny detectors are used to determine object boundaries and wheel axles. The used vehicle model with two axles in detail is described. After detecting and determining the number of wheel axles of the vehicle, a model of vertical oscillations of the vehicle is selected by the passive method of image analysis, and based on the obtained geometric and kinematic parameters, the dynamic characteristics of the vehicle are determined.