Research And Implementation Of Desktop Hyperspectral Push-broom Imaging And Quality Evaluation Methods

The hyperspectral imaging system collects radiated and reflected electromagnetic signals of targets by detecting objects,forming hyperspectral images.Hyperspectral images have the optical characteristics and spectral recognition ability,which can effectively classify,recognize and quantitatively extract substances.Among them,the light and small hyperspectral imaging system has become the future development trend of spectral imaging system due to its features of convenient installation,small volume,light weight,low cost of research and easy carrying.Therefore,it is necessary to develop light and small spectral imaging systems and evaluate their imaging quality.This paper which is focused on the development status of hyperspectral push-scan imaging systems at home and abroad,studies and designs the desktop hyperspectral push-scan imaging system.At the same time,this paper realizes the automatic focus of the imaging system and image quality evaluation of this system.The main works of our paper are as follows:1.The control and calibration methods of the desktop hyperspectral push-sweep imaging system are designed.By controlling the push-sweep platform and hyperspectral camera,the push-sweep imaging system is realized.At the same time,wavelength calibration and reflectance correction of hyperspectral images are carried out for the imaging system.Experimental analysis shows that the quality and resolution of hyperspectral images acquired by the push-scan imaging system are good.2.An automatic focusing method of the hyperspectral camera based on spectral quality evaluation is proposed.This method constructs a spectral quality evaluation function based on quaternion wavelet transform.By calculating the quality evaluation value of spectrum under different defocusing states,the camera is guided to automatically focus.Experimental analysis shows that this method has high focusing accuracy and only uses single-line spectral data for focusing,which is suitable for push-sweep hyperspectral camera.3.Based on the quality evaluation of hyperspectral image statistical characteristics and spectral signal measurement,a spectral signal quality evaluation method based on lifting wavelet transform is designed.Besides,by establishing the dataset of hyperspectral image quality evaluation,a hyperspectral image quality evaluation method based on deep convolution neural network learning is designed.Experimental analysis shows that two evaluation methods designed in this paper can evaluate the quality of hyperspectral images accurately and have good subjective consistency.4.A desktop push-scan hyperspectral imaging software system based on C++ and QT is developed.This software system has five functional modules,including image controlling module,camera focusing module,hyperspectral data acquisition and processing module,wavelength calibration module and imaging quality evaluation module,which can realize spectral data acquisition,camera automatic focus,wavelength calibration and hyperspectral image quality evaluation.