Recognition Of Cave Targets Based On Computer Vision

The cave targets include tunnels,tunnels,military caves and so on,which have high recognition value in construction,geographic exploration,military technology and other fields.Recently,computer vision technology has developed quickly.Target recognition based on computer vision technology has a good effect and low cost.In this paper,for the problems of few automatic recognition methods for cave targets recognition,and most of the recognition methods are about multi-sensor fusion,and have difficulty and high cost,proposed and designed cave targets recognition methods based on multi-feature,deep learning,and infrared and visible image fusion.This study proposed a multi-feature based method,which recognized the uncoverd and parrtly coverd caves.In this method,HOG features are used to pre-screen images;then an image local adaptive threshold generation algorithm is proposed according to the gray level features,which can segment and extract suspected objects in images;finally,the shape features are used to distinguish the suspected contours in images and complete the recognition.The recognition accuracy of this method is higher,but the rate of missed detection is higher,and the recognition effect decreases when the image noise is complex.This study proposed a method of object recognition based on deep learning,which has higher recognition accuracy than the method based on multi-feature.Traditional deep learning methods need massive data.This method combines convolution neural network and meta-learning to establish meta-convolution network,which can be trained with small samples to get the model.Then it combines lifelong learning with multi-feature based method to design expert model,so that the model can be continuously updated.This method can maintain high recognition accuracy for the uncoverd and partial coverd caves when the image has multiple noises.This study designed method of infrared and visible image fusion.The method collects infrared images of fully coverd and night caves,analyses infrared features,determines temperature threshold,identifies fully coverd and night caves,and then uses image fusion method based on VGG network to fuse the recognition results(visible image).The method utilizes both visible and infrared images to realize all-weather recognition of cavse.Finally,this paper establishes the cave datastes,and designs the cave recognition system based on infrared and visible image fusion,completes the relevant experiments and analysis.The results show that the method has high accuracy and good stability.