Research On Remote Sensing Target Detection Technology Based On EEG And Eye Movement

Detecting targets in remote sensing images has a wide application in both military and civil fields.For military purposes,it is mainly used for searching and detecting military targets,dynamically analyzing and monitoring battlefield environment.For civil purpose,it is applied to many industries,such as meteorology,agriculture and geography.The traditional computer vision based methods for target detection in remote sensing images have some achievements for specific targets.However,these methods were highly dependent on the prior information and the specific image target,and were difficult to detect uncertain targets,small targets,camouflage targets and incomplete targets in remote sensing images.The human brain is considered to be the most powerful system for visual information processing,with a capacity to catch the sensitive key information from a scene in only a few hundred milliseconds.Therefore,information processing and interaction system,which combines the intelligent perception ability of human brain for complex environment and the powerful computing ability of computers,can integrate the advantages of machine and human brain in one.Hence,it will be possible to solve the bottleneck problem of the machine learning method based on formal expression system in dealing with the complex problems in the real world and provide a new technique and way for target detection in remote sensing images.Aiming at the actual needs of target detection in remote sensing images,target detection technology for remote sensing images is studied based on multi-modal cognitive neural signals in this thesis.We mainly focus on feature extraction of Electroencephalography(EEG)and eye movement(EM)related with target detection and the classifying models.In order to further improve the detection accuracy,fusion methods for EEG and EM are also studied.The main work is as follows:1.The existing single trial event-related EEG signal analysis methods mainly built a unified spatial filter,without considering the spatial differences among a variety of Event Related Potentials(ERPs).Hence,this thesis proposed a remote sensing image targets detection method based on the combination of multiple ERP spatio-temporal features.In this method,the ERP components with significant differences induced by the remote sensing images containing the target or the non-target were extracted through quantifying entire waveforms.And then the spatial and temporal characteristics of various ERP components which extracted by a group of spatial filters are fused by the principal component analysis method.Finally,the object recognition model is constructed using the logical regression method to realize the fast detection of the weak and small targets in the remote sensing images of the complex background of large field of vision.The results show that using the method in this thesis,there are statistical significant increase in target detection accuracy without increasing the time consumption compared with the traditional methods through making full use of the temporal and spatial features of multiple ERPs.2.Aiming at the problem that the existing eye area-based region of interest analysis methods do not determine whether the area contains targets or not,a remote sensing image target detection and positioning method using attention to different fixation regions,which based on the eye movement parameter index analysis for image target location.The method designed an area clustering algorithm based on spatio-temporal constraints to extract the area of attention of the subjects at first,and then it judged whether the target is included in the region by the attention degree of the region,hence recognizing and locating the targets in remote sensing images.The results show that the method presented in the thesis realize the detection and location of remote-sensing image targets in complex backgrounds through combining eye-moving region-of-interest analysis technology with cognitive analysis technology.3.Research on the method for detecting and locating targets in remote sensing images through the fusion of EEG and EM.The eye movement signal cannot accurately determine the cognitive activity in the fixation process because of the limited information,and the EEG signal cannot locate the region of interest(ROI)that the human eyes are fixing.Aiming these problems,the thesis uses a ROI analysis technology based on eye movement to construct a set of candidate regions that may contain targets on one hand and fuses EEG and EM related to target recognition to select regions containing targets on the other hand.Experiment result shows that this method improves the accuracy and stability of remote sensing image target detection through the fusion of multi-mode neural features.