Font Size: a A A

Research On Method Of Target-background Contrast Enhancement Based On Polarization Vision Of Mantis Shrimp

Posted on:2022-02-09Degree:MasterType:Thesis
Country:ChinaCandidate:B B ZhongFull Text:PDF
GTID:2480306560455434Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
The polarization-sensitive creatures can detect prey or habitat by detecting polarization light,and its polarization vision mechanism has important reference significance for the modern technology of polarization information acquisition and processing.The mantis shrimp,as a typical representative,its complex compound eye structure makes it possess a variety of possible polarization information processing systems.Its rich triaxial eye movements have potential adaptive mechanisms.And it has both polarization vision and color vision.From the perspective of information processing,we explore the mechanism behind these biological behaviors,and are inspired by these to design a series of bionic methods.In experiments,such as the scene where the color of the target and the color of the background are similar,the method in this thesis can effectively enhance the contrast between the target and the background,and has potential application prospects in the fields of camouflage removal and target detection.The main research work is as follows:(1)Based on the study of the structure of the compound eye of the mantis shrimp,the polarization information processing mechanism is studied,and several multi-channel polarization distance models are designed and discussed to enhance the contrast between the target and the background.In the simulation experiment and the actual measurement experiment,the image processed by the four-channel polarization distance model has the best effect.Its gray scale contrast(GSC),signal to clutter ratio(SCR),and Fisher distance(FD)are 3.11 times,6.5 times,and 5.53 times as compared to those of the degree of polarization image respectively.For the model,a polarization-vision neural network is simulated with four-orientation receptor information as input,and the network connections are realized in a cascaded order.(2)Based on the study of the eye movement of the mantis shrimp,this thesis studies the mechanism by which this behavior can optimize polarization information,and designs a bionic model that can adaptively enhance contrast vision.In this model,a pixel array is used to simulate a compound eye array,and the angle of polarization(Ao P)is used as an adjustment mechanism.The polarization information is pre-processed by adjusting the direction of the photosensitive axis point-to-point.The experimental results show that the model can effectively enhance the contrast between the target and the background of the Ao P image.The three indicators are respectively 12.8 times(GSC),7.4 times(SCR),and 8.6 times(FD)as compared to those of the original Ao P image.(3)Based on the previous research work,an adaptive contrast enhancement method based on polarization vision is designed.The polarization images in the four directions are optimized by the adaptive model and used as the input of the four-channel polarization distance model.In the experimental results,the three contrast indicators of this method increase by 18.34%(GSC),362.00%(SCR),and 32.00%(FD)as compared to those of the four-channel polarization distance model respectively.It shows that this method further improves the significance of the target in the image.(4)Based on two kinds of vision of the mantis shrimp,a contrast enhancement method based on polarization vision and color vision is designed.By fusing the result of the new four-channel polarization distance with the chrominance component of the HSI color space,a new color image is obtained.Experimental results show that this method enhances the details of edges and other details based on the research content(3),and restores the color information of the target area to a certain extent.
Keywords/Search Tags:Biological polarization vision, Polarization contrast, Object detection, Eye movements, Adaptive mechanism
PDF Full Text Request
Related items