Multi-target Localization Based On Convolutional Neural Network

Accurate real-time positioning of multiple targets is a core issue that must be addressed in navigation,measurement and control.The way of locating the radiation source is divided into active positioning and passive positioning by whether the positioning terminal actively transmits signals.Because passive positioning is concealed,it has important strategic significance in the military field.The most widely used passive positioning is cross positioning.Direction-finding cross-positioning is to locate multiple targets by calculating the intersection of the target direction-finding lines.However,when the number of targets increases,many false points will appear.At this time,the correspondence between the observing station and the target needs to be found.Existing multi-target localization algorithms generally have problems of large computational complexity and poor real-time performance.There are many interferences in the actual environment.There is an error in the angle value received by the observatory.The anti-interference ability of the existing multi-target positioning algorithm is also weak.When the angle value has an error,the positioning result is not accurate.In view of the above problems,this paper proposes a multi-objective cross-location using convolutional neural networks.The main research contents and innovations of this paper are as follows: Convolutional neural network is used to perform cross-directional positioning of multiple targets to achieve real-time positioning of multiple targets.Moreover,in the case of an error in the angle value,high precision can be achieved by using a convolutional neural network.In this paper,the general convolutional neural network model is used for training.Although the depth of the ordinary convolutional neural network model is deeper,the accuracy will be higher,but the corresponding complexity will be larger,which will greatly increase the training time.Therefore,this paper learns from the idea of Inception model and creates a convolutional neural network model that achieves a balance between model depth and precision.Finally,the residual network and the Inception model are combined to further shorten the training time of the model.The experimental results show that compared with the existing multi-target localization algorithm,the use of convolutional neural networks to cross-target multiple targets is not onlyhighly accurate,but also highly real-time,and the anti-interference ability is also stronger.