Vehicle detection with image conversion into point representation
Zbigniew Czapla · 2016
The paper presents a novel image data-based method of vehicle detection. A sequence of input images consists of consecutive frames taken from the video-stream. The video stream is obtained from the camera placed over a road. Images from the sequence of input images are converted into point representation separately. Conversion of images into point representation is performed using analysis of small image gradients. Point representation of images consists of edge points whose layout is in accordance with edge of objects contained in images before conversion. For all input images the same detection field is defined. The state of the detection field is determined on the basis of sums of the edge points calculated within the detection field. A vehicle driving through the detection field changes its state. Detection of vehicles is carried out by analysis of the state of the detection field. The experimental results are provided.