Research Of Underwater Spectral Image Restoration Methods For Staring Spectral Imaging System

Underwater spectral imaging is a promising method for sub-sea object detection,geological survey,and environmental monitoring.However,it is challenging to obtain underwater spectral information.Both the energy and spectrum of light is distorted during the underwater imaging process due to wavelength-dependent attenuation by water,and sometimes the randiance of objects is merged by strong scattering light.In this paper,a set of underwater spectral imaging system is designed,and the restoration methods of underwater spectral images for the system is proposed.In the aspect of system design,we design a staring underwater spectral imaging system,which have good anti-knock performance.Underwater light source and laser ranger is also designed for imaging assists.The system can be used to obtain underwater images within 10 m with wavelengths varying from 400 nm to 700 nm.Calibrations of absolute radiance,spectral resolution and spectral sensitive limit are performed to ensure the stability of the system.In the aspect of algorithm design,we design a set of underwater spectral image processing algorithm that includes image preprocessing(i.e.image denoising and spectral image matching),radiation compensation of underwater spectral images,rectification of underwater scattered radiance,and(4)image enhancement.In the image preprocessing algorithms,denoising algorithms are designed based on median filtering and denoising neural network to promote the quality of images in weak light.Spectral image matching algorithms are proposed based on SURF operator with the aid of monochrome cameras.Radiation compensation of underwater spectral images includes the compensation algorithm of radiance based on passive light source,modeling of underwater light field and the compensation algorithm of radiance based on initiative light source,aiming at removing the influence of water attenuation.Based on underwater light transfer theory and imaging theory,underwater spectral imaging model of self-illumination objects is proposed,and radiance compensation algorithm is designed based on the model.Experiments are earied out to calibrate the model and verify the proposed method,in which a smart phone is used as light source.Considering the characteristics of light source in air and inverse-square law,model of light field in both air and water is established.Model calibration method that use the radiance of just two underwater distance on optical axis is also proposed.Comparison of estimated and measured radiance on axial and section plane are conducted,verifying the veracity of the model.On the basis of imaging model of self-illumination objects and light field,the imaging model of any underwater targets is established,and based on which algorithm for normalizing distances are presented.Underwater images taken at 8-10 m are obtained and all images are normalized to 8 m to prove the accuracy of the model.Rectification algorithms of underwater scattered radiance is used to deal with spectral distortion.Dehazing model is designed based on the reflection of reflector-board,and model-based dehazing algorithm is proposed to get precise reconstruction images.The reflectance of reconstructed images are calculated and compared with the measurements in air to show the accuracy.In the image enhancement process,scale-recurrent network is used for motion debluring.Besides,blind deblur method is used to remove the influence of defocused blur and Gaussian blur caused by forward scattering in water,thus clear and sharp underwater spectral images are achieved.As underwater spectral images are restored,precise and visual spectral signature are obtained,which can be widely used in underwater detection.