Research On Aircraft Detection In High Resolution Remote Sensing Images

In the past 10 and more years,remote sensing technology has got rapid development,along with which is the explosive growth of the remote sensing data.Remote sensing images are different from the natural images,it has some strong points,such as wide range coverage,inclusion of significant information,and short observation period and so on.At present,how to make use of these remote sensing data efficiently,how to quickly analyze the remote sensing data and extract useful information has become the main bottleneck of the application of remote sensing technology.Especially after entering 21 Century,this field has been paid more and more attention in scientific research.Object detection is a foundational and challenging research topic in the field of remote sensing.It has great significance for natural resources exploration,natural disaster monitoring,military target location and so on.In today’s fast changing information war,remote sensing technology has become an importance way to obtain information,who can quickly and efficiently obtain remote sensing image information in real time,who will hold the war initiative,this is sometimes critical to the outcome of the war.With the deepening of scientific research,more and more object detection methods have been proposed.For example,Bo et al.and Xu et al.use the traditional spectral-based methods to detect aircrafts;Zhang et al.proposed another extension of the DTPBM model for the object detection of high-resolution remote sensing imagery.However,due to the complexity and variety of remote sensing images,the artificially designed shallow features do not have enough representation ability.Therefore,these methods can only be applied in simple scenario.In face of the application scenario that contains complex content and information,we still need to continue to explore.In this paper,we build a series of object detection systems to detect aircrafts in high resolution remote sensing images.Firstly,based on the traditional SIFT feature,an aircraft detection frame work based on the traditional bag-of-words model is proposed,and its performance in the actual scene is tested.After that,we proposed two neural networks based object detection framework that one is based on single deep convolution neural network,and the other one is based on a cascade architecture of convolution neural networks.And also,we test the detection performance of the two system framework of the application in the scene.