Measurement of Vehicle Parking Deviation Angle Based on Machine Vision

ABSTRACT

In order to improve the detection accuracy and increase the automation level of some items in vehicle detection, a method of computing vehicle parking deviation angle based on machine vision was proposed. The feature points of left and right vehicle images were detected by SIFT (Scale Invariant Feature Transform) algorithm. Euclidean distance was used as the similarity measurement factor, the feature points matching algorithm based on bidirectional searching was put forward. The matched points of left image was divided into two sets, on the left and right half planes. The feature vectors on right half was corrected symmetrically, and the symmetric points were detected. By camera calibration, the threedimensional coordinates of midpoints were obtained and the vehicle longitudinal axis was received, finally the vehicle parking deviation angle was analyzed. The novel contribution of this method was that the measurement is based on feature points detection and symmetric feature points matching. The experiment results showed that this method effectively detected the parking deviation angle, which could be applied in vehicle detection and provided reference to correct the measurement data of vehicle testing.

Keywords: