A Study Of Shallow Seafloor Relief Measurement Using Optical Satellite Images

Water depth and seafloor relief in shallow sea are very important hydrological information,which play important roles in such aspects as maritime navigation,marine environment management and marine resources development and utilization.With the rapid development of remote sensing,extracting water depth and seafloor relief in shallow sea from satellite imagery is becoming a current research focus.First,this paper discusses bathymetric inversion algorithms that use multi-spectral satellite images.After introducing the mechanism of bathymetric inversion algorithms based on multi-spectral remote sensing images,four bathymetric inverse algorithms were reviewed,which are one-band linear regression model,two-band-ratio linear regression model,multi-band-combination linear regression model and BP-ANN model.Then these four bathymetric inverse algorithms were experimentally compared using Worldview-2 multi-spectral images and the data of true water depth,which acquired from the same area and at the same period of time.The results show that multi-band-combination model and BP-ANN model are better than one-band model and two-band-ratio model,and can be used to extract more accurate water depth from satellite multi-spectral images.Second,this paper presents an improved BP-ANN algorithm in order to overcome the shortcomings of the traditional BP-ANN algorithm,i.e.slower convergence,more sensitive to its initial weights and thresholds in learning phases.The theory of particle swarm optimization(PSO)was used to optimize the weights and thresholds of ANN in the process of learning.The results show that the improved BP-ANN algorithm has faster convergence speed and better generalization ability,and it is not sensitive to the initial weights and thresholds.The results also show that more accurate results can be acquired using the improved BP-ANN algorithm than using the traditional one.And last,this paper presents a more rigorous algorithm for the correction of refraction effects in two-media stereo photogrammetry.In this algorithm,the mid-point of the shortest line segment joining two aerial corresponding rays of a point on an underwater object is used as a photogrammetric intersection point,so that the intersection point can be determined and the refraction correction formula from the intersection point to the true object point can be strictly derived whether the two rays are intersecting or not.The negative effect on the refraction correction was analyzed,which caused by that the two rays are non-intersecting.The positional relationship between the intersection point and the true point was also studied.The formulas regarding water depth and geodetic coordinates of points on an underwater object were deduced,that is often known as correction of refraction effects.After that,the algorithm was tested by two experiments using the data of WorldView-2.The results show that the algorithm is suitable whether or not the two aerial corresponding rays of an underwater object point are intersecting,and it can greatly improve the measurement accuracy of underwater object's elevation.