Underwater Target Rapid Detection Method Based On Hyperspectral Technology

Compared to ordinary RGB imagery,hyperspectral imagery has very high spectral resolution with abundant spectral information.Hyperspectral imagery has the unique characteristic of acquiring spectral and spatial information simultaneously.Compared with the traditional target detection algorithm based on spatial information such as color feature,shape feature and texture feature,the hyperspectral target detection algorithm can detect the target according to the difference in spectral dimension information between the target and the background.Therefore,hyperspectral technology can be advantageous in the sense of a poor underwater imaging environment,complex background,or protective mechanism of aquatic organisms in underwater target detection.Due to high data redundancy,slow imaging speed,and long processing of hyperspectral imagery,a direct use of hyperspectral images in detecting targets cannot meet the needs of rapid detection of underwater targets.This paper mainly studies the rapid detection method of underwater targets based on hyperspectral technology.On the premise of retaining the target spectral characteristic information and maintaining good detection results,band selection is the main data dimensionality reduction method to achieve rapid detection of underwater targets.The specific research contents are as follows:Firstly,aiming at the problems of high redundancy of hyperspectral data,slow imaging and processing speed,a fast detection method for underwater hyperspectral targets based on band selection is proposed:first,band selection is performed for the target to be detected,and applications are selected the best band subset for detection;then,the underwater band imaging system composed of only the best band subset is used to collect the best band image of the underwater target;finally,the constrained energy minimization algorithm is used for detection.Secondly,for the selection of the optimal band subset,the minimum variance criterion,OIF index method and clustering algorithm are effectively combined to propose a Constrained-Target Optimal Index Factor Band Selection(CTOIFBS)algorithm.The algorithm can select the best band subset for the target to be tested,and experiment on HYDICE data and real underwater data.The results show that the detection result using the band subset selected by CTOIFBS algorithm is better than other traditional band selection algorithms.Thirdly,in view of the optimal band subset imaging process of underwater targets,this paper design an underwater spectral imaging system.The experiments prove that the data collected by the USI system and the data collected by the ordinary hyperspectral imager have considerable features of expression of the ground features.The system can be applied to the rapid collection and detection of underwater target spectral information.