Research On Reasoning Method Of SAR Mission Failure Based On Improved Deep Forest

Synthetic aperture radar(SAR)in geological research,environmental monitoring,weather forecasting,disaster early warning,resource exploration and military reconnaissance and other fields is widely used,and become the main equipment of military reconnaissance,airborne SAR in the process of reconnaissance missions,component failures and parameter errors,such as environmental noise could lead to its reconnaissance mission failure.Therefore,how to effectively improve the real-time and accuracy of the reasoning of SAR mission failure causes is of great research value for the system to quickly troubleshoot the failure and improve the success rate of the mission.This paper takes a certain type of airborne SAR equipment as the research object.Based on the imaging data obtained by SAR in different terrains,different parameters and different environments,the causes of mission failure radar are inferred on the basis of deep forest algorithm.The main research contents are as follows:(1)In order to improve the reasoning speed of mission failure causes,this paper proposes a feature extraction algorithm of SAR image based on rotation invariant LBP gray co-occurrence matrix.This method only need scan a gray image to build a rotation invariant LBP gray level co-occurrence matrix,reducing the number of the image scanning and image information loss,which improves the gray level co-occurrence matrix exists in the process of extraction of SAR image texture feature of image gray level information loss,rotation invariance and properties,such as computational complexity.The proposed algorithm has been verified by experiments on nearly 3,000 SAR image datasets in multiple scenes.The experimental results show that the proposed algorithm can reduce the time consumption by about 25% while ensuring the accuracy of reasoning,and greatly improve the reasoning speed of the cause of task failure.(2)In order to further improve the accuracy of reasoning for mission failure causes,this paper proposed a reasoning method for mission failure causes of SAR based on improved deep forest by adding fitting quality characteristics and increasing the diversity of basic learners in the deep forest cascade layer.In order to reduce the loss of important feature information,the contribution degree of the feature is ranked in each level of linkage layer,and the features with higher importance are added to the input of the next level as fitting quality features.In addition,the diversity of basic learners in each layer is increased by extending the basic learners from the original two to four,thus improving the overall performance of the deep forest integration model.The proposed algorithm has been verified on nearly 3,000 SAR image datasets containing 29 types of failure causes.The experimental results show that the accuracy of the proposed algorithm can reach94.3%,which is better than the traditional methods such as deep forest and random forest.