Research On Technology Of Feature Extraction And Target Location From Images Of Extended Targets

Technology of feature extraction and tracking of images has a wide range of applications in civil and national defense.So feature extraction and target positioning has been a focus and hot technology at home and abroad.Space debris has severe interference with the normal operation of spacecraft.Adaptive optics can detect and locate space debris based on ground-based telescopes.Therefore,it is necessary to study the feature extraction and positioning technology of the extended target image of adaptive optics.The research goal of this paper is to explore several image processing algorithms,which are suitable for extracting the motion trajectory of the extended target's feature points from the sequence image obtained by adaptive optics system,and then carry out simulation verification.This paper combines the traditional feature extraction algorithm with the new target tracking algorithm proposed in recent years,and studies and explores several issues in this field.The main research content of the full text includes:1)This article describes in detail the target feature extraction algorithms commonly used extended.The Research progress at home and abroad of SIFT algorithm,HOG algorithm,color name feature algorithm are analyzed and compared deeply.2)The phase correlation algorithm based on feature extraction is studied.This paper proposed a method for positioning extended targets in sequence images by extracting multiple SIFT features in an extended target image.The distribution of SIFT feature points obtained from the extended target image is not stable.Therefore,this paper expands a single feature point into a certain size of feature block,and it is easier to obtain the position change of the feature block in the sequence image.In order to improve the positioning accuracy of the feature block,the phase correlation function was interpolated to obtain the sub-pixel positioning accuracy.Due to the large error of motion trajectory extraction of a single feature block on an extended target image,this paper improves the extraction accuracy of the motion trajectory by calculating the average of multiple feature blocks.The moving image sequence of typical extended target is produced by the motion of the portrait's translation,rotation,and scaling.The simulation experiment of feature extraction of sequence extended target image is performed,and the effectiveness of the proposed algorithm is verified.For the adaptive optical extended target video sequence,the target is located by the phase correlation algorithm based on SIFT features,and the target trajectory is obtained.3)The article studied the improved kernel correlation filtering(KCF)algorithm which used multi-channel feature of HOG feature and color name feature.Using three typical video sets of human and animal movements in complex backgrounds,KCF algorithm was used to extract the motion trajectory of extended target.The performance of the algorithm was evaluated using the center position error,success rate,and distance accuracy.Comparing the simulation results of three typical video sets,it shows that the multi-features algorithm based on KCF can effectively improve the overall tracking ability.The research in this article lays a foundation for the follow-up technology of motion target detection and tracking based on feature extraction of sequence extended target image.