| The fundamental methods of linear object recognition and tracking are introduced in this paper. As to recognition, many kinds of algorithms for line detection are deeply researched, and a new line detection algorithm based on fake direction is proposed. By convolution with edge image and 3*3 arithmetic operators which we designed, realizes the effective grouping for edge pels, obtaining the fake direction, so that randomized Hough transform can be implemented to detect straight line among groups in the same fake direction. As the result indicated, the proposed approach effectively reduce the compute time. Meanwhile, concerning with the basic thought of Freeman chain code, we fetch out detailed test of the grouping validity.In the aspect of tracking, starting with similarity measures and template update, lots of similarity measures are discussed, considering their advantage and disadvantage, according to the practical image serial characters, center object-weight method and edge pixels-weight method are proposed. As for the tracking affiliation effect, template updating strategies are discussed. Then combined with matching credit and center object increase, template content updating method and template size altering method are proposed. |
