Research Of Multi-spectral Wind Turbines Target Detection Algorithm

Aiming at the practical problem of the position detection of the windmill,the multi-spectral remote sensing image target detection algorithm is researched in this paper,wind turbines are viewed as targets to be detected.Existing anomaly detection algorithm LRX and BRX have their own advantages,but poor detection performance in complex backgrounds.Current anomaly detection algorithm LRX and BRX have their own advantages,but poor detection performance under complex background and cannot get better detection result in different regions at the same time.The targets to be detected relative to the whole image area is small,less.And the location of the distribution has certain regularity.Several traditional target detecting algorithms only considers the spectral characteristics of the pixel itself,underutilized target scene characteristics and structural characteristics,so that the target wind turbines difficult to detect effectively.To solve the above problems,this paper will study the anomaly detection algorithm and the target extraction methods from the following aspects.(1)The traditional RX anomaly detection algorithm and related improved algorithm LRX and BRX are compared and analyzed.The new detection algorithm SWRX is prompted in this paper.Firstly,the original image is divided into blocks,and the sub blocks are determined by the characteristic description,and then the appropriate detection method is selected according to the characteristic description;(2)From the point of view of the scene area of the target,the ground object classification is introduced into the target detection.Firstly,the original image is divided into blocks,The traditional gray level co-occurrence matrix method is used to extract the texture features of the sub blocks,and then use quad-tree blocking method,the texture features are viewed as blocking criterion.Then,set the appropriate threshold to suppress the background area and keep the target area.(3)From the point of view of the relationship between the target point and the space position,the characteristics of the spatial structure features of a region containing targets is analyzed.The target extraction method based on spatial structure feature is proposed,the direction and distance information are added to extract the target points which meet the requirements,and the final detection results are obtained.Finally,the four regions containing many targets in the multi-spectral image are selected to carry out the experiment,the detection performance of each detection algorithm is evaluated synthetically from the probability of detection,the probability of false alarm,the probability of correct,the probability of the sum of wrong and missing and ROC curve and other evaluation indicators.The experimental results verify the feasibility and effectiveness of the proposed algorithm.