Skyline Detection Key Algorithm Research

Skyline detection plays a very important role in aviation,navigation,visual navigation of drones,attitude control of micro-aircrafts,and geographical location marking.Therefore,a highly robust and accurate skyline detection algorithm is an urgent problem to be solved.This article examines several key algorithms for skyline detection.The main work of this paper is summarized as follows:(1)Considering that the starting point of the skyline is the edge position in the image,starting from edge detection theory,a skyline detection research based on fusion of edge extraction and adjacent gradient algorithm is proposed.Firstly,an image pre-processing method based on morphology is adopted.Then,the Canny detection operator is used to obtain the edge information in the image.Next,the location coordinates of the skyline can be determined according to the setting of the gradient threshold.Meanwhile,for the problem of the lack of edge in edge detection,the linear extended area search algorithm is applied to determine the location coordinates of the skyline,and a good skyline detection effect is achieved.(2)The aforementioned skyline detection algorithm is excessively dependent on the edge detection information and lacks the extraction of the overall information of the image.Therefore,a skyline detection algorithm that fuses the overall structural information of the image is proposed.Firstly,the RCA(Regional Covariance Algorithm)is used to conduct the coarse segmentation for the overall information of the image,and determining the possible intersecting area of the skyline.Next,a new method based on the combination of the global gradient mean and the local gradient mean value is proposed to determine the starting point of the sky and the non-sky regions.And then,the maximum possible horizontal line can be detected by using the AGM(Adjacent Gradient Maxima)algorithm.Finally,the skyline of the input image can be detected.This method does not need to extract the edge information of the image,which can make full use of the overall structure information of the image.The experimental results show that the algorithm has high extraction accuracy for the skyline.(3)Sparse theory is an effective machine learning method.It has been widely used in face recognition,image restoration and image classification in recent years.Therefore,this paper proposes a skyline detection algorithm based on LBP(Local Binary Pattern)and sparse representation.Firstly,the image is processed with gray scale.Then,establishing the 3*3 feature extraction region of the 3 adjacent pixels of the skyline coordinates in the training sample images.In order to get the dictionary,the LBP feature extraction is used to establish the LBP feature vector of skyline through the statistical histogram.Next,the reconstruction error is calculated by the sparse decomposition factor,and finally the skyline position is detected according to the set reconstruction error threshold.The proposed algorithm can effectively detect the pixel coordinates of the skyline in the image,which shows a good detection effect for the lack of edge of the skyline.(4)Considering the complex background,a single feature extraction has certain limitations on the skyline detection.Therefore,a skyline detection algorithm based on multiple feature extraction and edge correction is proposed.Firstly,the multi-eigenvalues of the training pixels randomly chosen in the skyline regions can be extracted by using Gabor texture information and color information.Then,the multi-eigenvalues are used to train a classifier based on Support Vector Machine(SVM)to obtain the initial position coordinates of skyline.Next,the Canny operator method is used to detect the edge of the gray image.And then correcting the position of the initial coordinate,and the coordinate position of the skyline can be finally obtained.The experimental results indicate that the proposed method can effectively detect the skyline coordinates,reduce the interference of other pixels to some extent and make the skyline more smoothly.In this paper,four skyline detection algorithms are proposed from the perspectives of the starting point of the skyline,the overall structure of the image,and feature extraction.The proposed algorithm was tested on the data sets of the Web Set in University of Nevada’s Machine Vision Laboratory.,and compared with Li et al(Edges),Gradient Info and SIFT+HOG Edges.The experimental results show that the proposed algorithm can effectively improve the extraction accuracy of the skyline can quickly and effectively detect the skyline that is not only a straight line and a diagonal line but also has a good detection effect on the skyline in a complex background.