Currency Sorting Algorithm And Abandoned Objects Detecting In Surveillance Area

With the development of technology,bank business is developing towards comprehensive automation and using currency sorter will be an inexorable trend.Meanwhile,intelligent monitoring constitutes the foundation of safe city's construction.According to the above,this thesis mainly studies two aspects:the classification of new and used bills and those torn or not in cash sorting;the event of abandoned objects in surveillance area.The thesis has obtained the following results:1.A currency sorting algorithm based on image features and machine learning is put forward.In classification of new and used bills,the blank area contrast and the brightness of the whole bill are selected as features.Classification and regression methods are adopted respectively.These two methods both show good performance in effectiveness and stability test,especially the SVM classification.Both algorithms have high computing efficiency to meet the needs of bill sorting system.In the tear detection,the Canny operator is chosen to detect edge.According to the length in vertical direction,whether the bill is torn or not can be determined.The algorithm can suppress noise and accurately locate the position.Experiment shows that the method is able to effectively detect conventional tear.2.A detection algorithm for abandoned objects is proposed.Firstly,a simplified background model is adopted to obtain moving object.This algorithm realizes a tracking method using the nearest neighbor matching,and obtains the static foreground area through the analysis of trajectory.Finally,the interference of pedestrians is eliminated by calculating the track's starting and end point.Moreover,homography transform is adopted to further reduce false detection.This algorithm has realized 0 missed detection,0 false detection on PETS2006 standard database.