Boundary Detection in Images in Intelligent Systems of Autonomous Vehicles Using Wavelet Transform

Abstract

“Smart city” is one of the concepts within which modern scientific and technical areas are actively developing. This concept defines the tasks associated with the organization of various infrastructure objects' continuous monitoring to optimally divide resources and ensure security. One of these tasks is to ensure road safety involving autonomous vehicles. The solution to this problem involves various technologies, including computer vision technologies. One of the approaches in this area is based on machine learning methods. These methods consider objects models in images represented as attribute vectors. Boundary attributes are often used here. The formation of these attribute comes down to two steps ─ the boundaries detection and their description in the form of descriptors. This paper describes an approach to the objects' boundaries detection in images in intelligent systems of autonomous vehicles, which is based on the use of wavelet transform. The method is based on determining the significance of the brightness change magnitude at some point at a certain level of the wavelet decomposition. For that, it is necessary to evaluate the contribution to the total image energy of the detailed coefficients corresponding to this point. The method determines the sequential refinement of boundaries, which is as follows: as the brightnesses of the original image's copies at different levels are interconnected, we assume that the boundary points at different levels correspond to each other. The proposed method is simple to implement, has a relatively high speed and the ability to flexibly configure for real operating conditions.