Research On Object Recognition Algorithm Based On Contour Feature Description

Object recognition is an important field of contemporary computer science research,and it has developed into an independent discipline.In recent years,this discipline has developed rapidly,and it has a wide range of applications,almost covering all fields.From the field of Astronautics to biological science,information science,resources and environment science,astronomy,physics,industry,agriculture,national defense,education,art and other fields and industries.It has been widely used in national defense economy,national defense construction social security and social development,and it also has a profound impact on the whole society.In this thesis,contour features are used as recognition features.The methods of contour acquisition,contour partition,evaluation and merging,and shape context recognition based on boundary markers are studied.The main contents of the thesis are as followsThe contour acquisition algorithm is researched.A good edge detection algorithm is very important for object recognition.For this reason the effect of the edge extraction algorithm is compared and extracted.and the experimental results are verified.Then,the mathematical morphology method is used to process the result of edge extraction to eliminate unnecessary fragments,and the shape map of the target is obtained.Finally,the object contour is extractedThe object recognition algorithm based on significant contour feature segments is researched.In order to obtain the contour segment with significant features,the thesis first formulates the contour division scheme,then proposes three parameters to evaluate the value of the contour segments,and finally merges the contour segments that do not meet the evaluation criteria.Experiments on the images of MPEG-7 library prove that the evaluation and merging of contour fragments can ensure that the contour segments have significant features and it has a good effect on object recognitionThe object recognition based on the shape of the boundary marker is researched The shape context algorithm itself does not have rotation invariance.In order to recognize rotating object,a combination of boundary marking and shape context is proposed.Firstly,this thesis compares several methods of boundary marking,then calculates the center of gravity of the object and then studies the invariance of translation,rotation and scaling of boundary marking.Finally,experiments are carried out using boundary marking and shape context algorithm to verify that this method can be used to recognize rotating objects.The theoretical method studied in this thesis is applied to vehicle type recognition.On the one hand,it verifies the feasibility of the theoretical results on the other hand,it also realizes the recognition of some types of vehicles.Firstly,the model library is built by using the method proposed in this thesis.Then,the feature fragments are extracted from the vehicle's contour,and the contour fragments are marked by the boundary marking method.Finally,the shape context algorithm is used to recognize the contour fragments,and the recognition results are obtained.The proposed method is verified in the validity and applicability.