| In recent years,deep learning algorithm has been widely used in image classification,target detection,image recognition,remote sensing image change detection and other fields.Change detection of remote sensing images can obtain the information of land resources utilization and change,plant,house,mountain,ocean,etc.It can be used to detect natural disasters in remote sensing images such as earthquakes,landslides,tsunamis,dynamic information change detection in military battlefield,damage assessment of various military equipment and change detection of military deployment.High-resolution remote sensing images contain abundant information of geographic features.Traditional change detection algorithms are mainly based on pixel level.They are too sensitive to the processing of high-resolution geographic features,which results in high false alarm rate of change detection.Moreover,traditional algorithms are time-consuming and require a lot of manual participation.Aiming at the problems of traditional change detection algorithms for high-resolution remote sensing images,a change detection system based on C-Yolov3 model for high-resolution remote sensing images is developed in this paper.According to the characteristics of high resolution remote sensing image,the gray difference method,gray ratio method,canonical correlation analysis method and PCA algorithm are analyzed and implemented.The Yolov3 model is analyzed.The C-Yolov3 model of high resolution remote sensing image change detection is designed and implemented by integrating traditional algorithm with deep learning Yolov3 model and combining Regional correlation algorithm.The overall accuracy,detection rate and Kappa coefficient of change detection are improved.Through experiments and comparisons,combined with the texture,structure and characteristics of the objects,the system can recommend the detection algorithm based on image content,give the flow detection steps,and use the way of the website to unify the operation through the server. |
