Development of Fast Vehicle Recognition Argorithm in a Farming Environment

Abstract

In order to recognize the position of a vehicle in a farming envionment using image sensor, it is important to identify a marker on a moving vehicle in the image. The marker is extracted from the image by the difference in the characteristics of the color detected by image processing. The author tried to extract the marker pixels from an image by using chromaticity of pixels which is the ratio of R, G, B in each pixel. The image sensor recognized a marker set at 50m away from a CCD camera with the threshold of one chromaticity, but miss-recognition occurred with some images. Therefore, another method was developed using probability density function (PDF) whose variables are two chromaticities of a pixel. There was no miss-recognition of a marker with this method. and recognition was achieved at intervals of 0.53 second with PDF-table which was used as alternative to calculations by PDF.