Identification And Location Method Of Sedimentation Tank In High Resolution Remote Sensing Image Based On The Improved Yolov3 Model

Ion-adsorbed rare earths are an important strategic resource in the 21 st century.They are widely used in high-precision technology fields such as equipment manufacturing,electronic information,new energy,aerospace,and weapon manufacturing.They are extremely scarce and irreplaceable.Southern Ganzhou of Jiangxi province is known as the “kingdom of rare earths”.However,driven by economic interests,there are illegal exploitation and over-excavation in the rare earth mining areas in southern Ganzhou,which not only lead to the loss of rare earth resources,but also cause serious environmental pollution,soil erosion,etc.Ecological issues.Ion-adsorbed rare earth mining areas are generally located in remote mountainous areas,with winding roads,high mountains and dense forests,and the distribution of mine sites is relatively scattered.Their distribution characteristics lead to high cost,low efficiency and difficulty in rare earth mining and monitoring.How to quickly and accurately realize the mining and monitoring of rare earth mining areas has become an urgent problem to be solved.As an indispensable container in the rare earth element extraction process,the sedimentation tank is usually distributed in the rare earth mining area exposed on the surface.The leaching liquid in the sedimentation tank can be used as a basis for judging the state of rare earth mining.Therefore,according to the state of the sedimentation pond and its spatial distribution in the process of rare earth mining,this paper uses Pleiades high-resolution satellite remote sensing image as the data source to improve YOLOv3,and build a high-resolution remote sensing image sedimentation pool identification based on improved YOLOv3 and positioning methods to provide theoretical basis and technical support for the supervision of ion-type rare earth mining.The main research contents of this paper are as follows:(1)A method for identifying and locating sedimentation ponds in rare earth mining areas based on high-resolution remote sensing images based on convolutional neural networks is proposed.This method is improved on the basis of YOLOv3,and the attention mechanism is embedded in the feature extraction network to improve the feature extraction ability of the model.And replace the target positioning loss function of YOLOv3 with CIOU Loss to improve the positioning accuracy of the model to the prediction frame.When conducting specific experiments,the Pleiades high-score image data set is used to train the model and verify it,and then compare experiments with several other mainstream convolutional neural network models;(2)An image offset segmentation method and IOMIN index are proposed.The original image is offset and segmented to obtain a data set that completely covers the detection target,and then eliminate the redundant detection frame through the IOMIN index to achieve complete detection of the sedimentation tank and solve the problem of target missed inspection;(3)Convert the model detection result from the prediction frame represented by pixel coordinates to point-like elements represented by plane coordinates through the coordinate conversion formula.Then the random forest classifier is used to classify the location points to distinguish the location points that are misclassified as dark colors and building shadows.