Research On LSS-target(the Low Altitude,Slow Speed And Small Target) In Complex Background

The LSS(Low,Small and Slow)target is a collective term for the small aviation equipment that flies below the height of 2 kilometers and has a speed of less than 50 kilometers per hour.Multi-rotor drones and remote-controlled aircraft models have the characteristics of small size,light weight,and easy to modify.If the LSS aircraft is equipped with cameras or explosives and is used in improper ways,it may cause grave threats to public safety and air defense.Therefore,research on the LSS target detection technology such as the subject of anti-UAV group has become a hot topic.Based on a thorough investigation on the current status of research,we will analyze the imaging characteristics,noise characteristics,the LSS target characteristics and background characteristics in complex backgrounds and propose a set of LSS target detection methods with adaptive self-adaptation in complex space environments.This paper proposes different LSS target detection algorithms for visible and infrared images.For complex scenes with visible light cameras:(1)Aiming at the conditions of strong light or backlight,a background modeling algorithm based on uneven lighting correction is proposed.The system introduces a two-dimensional gamma function to adaptively suppress uneven lighting images and extract color features.Texture features are extracted by introducing extended-scale local invariant operators.Then it is cascaded into the ViBe+ background model to achieve the effective detection of the LSS targets.(2)Aiming at complex dynamic scenes,an object detection algorithm based on visual saliency is proposed.This algorithm makes full use of the color salient features of the LSS targets to improve the contrast,and proposes morphological difference to select the target seed points and improve the scan line filling algorithm to achieve the target selection.In this paper,seven groups of video sequences with complex sky backgrounds are selected to verify the algorithm.The results show that the algorithm proposed in this paper can accurately detect the LSS targets in complicated environments.For complex scenes with infrared cameras:(1)Aiming at infrared images with non-uniform noise,an LSS target detection algorithm based on TCAIE-LGM smoothing is proposed.The algorithm calculates the texture complexity and twodimensional information entropy of the image as the control parameters,and employs the adaptive L0 gradient minimum smoothing to remove fringe noise and suppress highfrequency details of the image.After that,a double Gaussian difference operator(DoG)is introduced into the multi-frame model based on pixels to achieve effective segmentation of targets and complex backgrounds.In order to verify the effectiveness of the algorithm,three sets of videos(from the VOT-LTIR 2015 database,OTCBVS database and Terravic Motion IR database)are used to test the algorithm in the research of this article.The results show that the algorithm proposed in this paper can reduce the false alarm rate and complete the detection with higher accuracy.(2)For infrared images of non-stationary complex scenes,this paper proposes a fast detection algorithm for the LSS targets.Cumulative histogram can well characterize the gray scale distribution of the image.The algorithm uses a double Gaussian function to fit its histogram,and uses the maximum likelihood estimation method to remove clutter and isolated noise points in the image.The region growth based on the four-condition constraint demonstrated in the paper can effectively obtain the complete target,and the constructed confidence function can greatly improve the accuracy of target judgment.According to experimental verification,the algorithm takes an average time of 0.085 s per frame to detect an unmanned aerial vehicle with a resolution of 640 × 512 at a distance of 2km in an upper computer environment with a main frequency of 3.2Ghz,an 8-core CPU,and 8G memory,which meets engineering requirements.For complex scenes in a multi-source sensor environment: An LSS target detection algorithm based on the fusion of infrared and visible light images is proposed.The algorithm locates the ROI region according to the weighted movement of the image and the information entropy theory,and fuses the infrared image and the visible light image to create a prior.Then targets are extracted based on local background subtraction.This algorithm can work in all-weather environment.Compared with the traditional singlesensor low-altitude slow-speed small target detection technology,it has obvious advantages in accuracy.The public data set verifies that the accuracy of the proposed algorithm is higher than 95%,which is improved by more than 10% in contrast with the traditional single sensor detection algorithm.