This paper introduces a reliable and accurate real time automatic questionnaire votes localization counting and statistics. The document real time reading process transforms information recorded on paper in real time into digital formats using a creative web cam and Matlab image acquisition toolbox. The snapshot of each questionnaire is then tracked for extracting the information (votes and control points marks) included in each digital form questionnaire.
The marks tracking scheme consists of four modules. The first is an image processing module. This module makes a color and shape based segmentation to find the pre-specified marks shape position (square for the control points and circles for the votes). The second one is a classification -based module to match the detected marks position to predefined training positions using the nearest neighbor classification process. The third module is a geo-referencing module to transform the detected marks coordinate system to base points coordinate system. The fourth module is a statistical based module to perform the statistics and bar graphs for the votes.
Experiments conducted with a variety of questionnaires show that our scheme can detect and track marks representing the votes robustly with high productivity, quality and speed document marks tracking and archiving services and with complete freedom from personnel management and count concerns
|
