The Research Of Change Detection In Image Pairs

Over the decades,image processing has been widely used in various engineering areas.As the applications of image processing have been extended,the original techniques and corresponding algorithm have met many limitations when solving the new problems.Therefore,it' s crucial to develop image processing,with the aim of applying the techniques in practical engineering problems,which would possibly be beneficial to everyday life.This is also the topic of this image era.This thesis is focused on several algorithmic methods and their applications in detection of changes in imaging.This thesis first introduced several common image registration algorithms,which is important to reduce the interference for the next step.The shortcomings of each method were also summarized.More specifically,I studied the registration algorithm based on Harris corner and compared two existing Harris corner filtering methods.Based on these two methods,I proposed a new corner filtering method based on the ratio of two values.The results showed that this method was better when compared to the other existing methods.This thesis also studied on the algorithm of detection for image change.This method is mainly based on the difference in image and the correlation factors in the LBP histogram.I also proposed an important feature in this detection method which is not sensitive to light.This feature would allow the detection based on the improved Haar-like feature and random forest.Last but not the least,this thesis also examined the fusion method based on the Markov random field.The similarity between the image segmentation and the image detection was discussed.I proposed a new fusion method of detection for image change based on Markov random field.This method could combine the results from different detection algorithms,which shows better results than any single detection algorithm.Based on all the work above,a new detection system was designed and compared to the other two detection methods.