| With the increasing number of cars in recent years,car safety has become the focus of attention.Millimeter wave radar is insensitive to light and weather conditions and can directly measure information such as distance,angle and radial velocity of the target.It is an important tool to detect other vehicles,pedestrians and the road environment.Initially,the function of millimeter wave radar in the vehicle sector focused on distance warning and target avoidance,but as the technology evolved,the applications of radar expanded to include adaptive cruise control（ACC）,automatic emergency braking（AEB）,blind spot detection（BSD）or lane change assist（LCA）,and so on.A precondition for millimeter-wave radar to achieve these functions is the ability to accurately track and detect targets.However,the road environment is complex and millimeter-wave radar is susceptible to other interfering factors.Due to the increase in the number of automotive millimeter wave radars,mutual interference between radars is becoming a serious problem.Interference can lead to a loss of targets or inaccurate detection results.Therefore,the interference suppression method based on the singular spectrum analysis is investigated in this thesis.In addition,due to the low measurement accuracy of the sensor and the influence of noise,millimeter-wave radar may cause errors in target tracking,and sometimes even cause target loss.Therefore,a study on the tracking method of millimeter wave radar is conducted in this thesis.The major work of this thesis is as follows:1.The thesis analyzes the effect of millimeter wave radar mutual interference on distance detection and angle estimation,and then modeled the interference.Simulation results for mutual interference show that mutual interference leads to an increase in the noise floor and a decrease in the signal-to-noise ratio during distance detection.The increased noise floor makes it difficult to detect the target by the constant false alarm rate（CFAR）.In terms of angle estimation,the degradation of signal-to-noise ratio makes the multiple signal classification（MUSIC）based on subspace techniques unable to achieve accurate estimation of the target angle.Then,the application of traditional Kalman filtering and particle filtering in the target tracking process and its limitations are analyzed.Kalman filtering is only applicable to linear models,while in the millimeter wave radar tracking process,a nonlinear conversion process from polar to cartesian coordinates is involved.Particle filtering is applicable to nonlinear systems,but it suffers from the particle degeneracy problem.As the number of filtering iterations increases,the particle weights tend to end up concentrated in a few samples,while most of the particles have small weights,which results in the waste of a large number of particles,in turn affects the tracking performance of particle filtering.2.An interference suppression method based on singular spectrum analysis is proposed.Since the conventional singular spectrum analysis method cannot select the desired signal components for reconstruction in the presence of interference,an adaptive segmentation and template matching based on SSA（ASTM-SSA）is proposed according to the characteristics of mutual interference in millimeter wave radar.In the adaptive segmentation,the change point of interference is determined by change point detection,and then the IF signal is divided into interference and interference-free segments based on the location of change point.In the template matching,the two segments are decomposed,and in the interference-free segment,the desired component is determined by the difference of adjacent singular value.The desire component is used as a template to match with each component of the interfering segment,and the component with the highest correlation is selected to reconstruct the signal.Simulations and experiments show that the proposed method can effectively suppress mutual interference between millimeter wave radars and improve the target signal-to-noise ratio.Both the target distance and angle can be detected accurately.3.A target tracking method based on improved particle filtering is proposed.During target tracking,the target trajectory deviates from the actual trajectory,and even the target is lost in a short time.Due to the particle degradation and particle depletion of traditional particle filter（PF）,an unscented particle filter（UPF）based on the combination of particle filtering and unscented kalman filtering（UKF）is proposed for target tracking.The unscented particle filter calculates the mean and variance of the particles by the traceless Kalman filter,and then uses this mean and variance for sampling.In this process,the mean and variance are calculated using the new observation information,which makes the sampled particles closer to the posterior distribution of the target.In the case of target loss,the historical state and motion model are used for target trajectory prediction,and the prediction process retains the historical state information as well as incorporates part of the predicted state information.Simulation and experimental results show that the unscented particle filter is better than the traditional particle filter and unscented Kalman filter in terms of target tracking accuracy and target loss prediction. |
