MULTIPLE OBJECT TRACKING BASED ON PENALTY GRAPH FOR AVOIDING HIJACKING PROBLEM OF TRACKERS

摘要

PURPOSE: A multi image object tracking program based on a penalty point graph for a hijacking problem of a tracker is provided to add a penalty point, which is in inverse proportion to a distance of a neighbor image object, to a weighted value of each particle, thereby preventing a hijacking problem for object tracking. CONSTITUTION: A multi image object tracking program extracts an image object from a current image. The multi image object tracking program obtains location information of the extracted image object(S2000). The multi image object tracking program labels a new object by a new identifier(S3100). The multi image object tracking program generates a penalty point graph(S4000). The multi image object tracking program predicts a particle group of the each image object(S5000). The multi image object tracking program calculates particle weighted values of predicted particles(S6000). The multi image object tracking program assigns a penalty point to the weighted values(S7000). [Reference numerals] (AA) Start; (BB) End; (S1000) Inputting an image; (S2000) Extracting an image object and obtaining location information; (S3000) Location information; (S3100) Adding to a tracing list; (S4000) Generating a penalty point graph; (S5000) Predicting the particle group of each image object; (S6000) Calculating particle weighted values; (S7000) Applying a penalty point to the particle weighted values; (S8000) Removing particles with '0' weighted values; (S9000) Updating the location of each image object and the penalty point graph; (S9100) Resampling the particle group; (S9200) Is a next image input ?