Vehicle Radar Target Detection Technology Based On Deep Learning

As the automobile industry rapidly develops and innovation in automobile intelli-gence continues to transform the field,cars have become the primary mode of transporta-tion for both people and goods.Vehicle-mounted millimeter-wave radar has become a research hotspot in the field of vehicle-mounted sensors due to its all-weather and high reliability target detection characteristics.On-board radar target detection is a key tech-nology in intelligent vehicle systems.Obtaining target information on the road ahead through radar sensors can improve vehicle driving safety.This thises focuses on the research on the target detection technology of vehicle radar,analyzes the applicable scenarios and advantages and disadvantages of different types of radar constant false alarm detection methods,and adopts the radar target detection method based on deep learning to solve the problem of low probability of weak target detection and high false alarm rate in strong clutter background.The specific research contents are as follows:First of all,the vehicle-mounted millimeter-wave radar system is analyzed and stud-ied,and the LFM continuous-wave millimeter-wave radar-related signal processing algo-rithm and process are deduced and analyzed to achieve the acquisition of information such as the distance,speed,and angle of the target.Data support is provided.Secondly,the radar clutter statistical model and constant false alarm detection algo-rithm are studied.The theoretical analysis and simulation verification of different clutter models are carried out,and the effectiveness and adaptability of different constant false alarm algorithms are verified by simulation under different complex backgrounds,and the performance degradation caused by model mismatch in conventional radar target detection algorithms is proposed And other issues.Then,aiming at the low target detection probability and high false alarm rate in the strong clutter background,a target detection algorithm based on conditional generative ad-versarial network is proposed.The network structure and loss function of the generative adversarial network are studied and analyzed,and based on the principle of conditional generative adversarial network,the simulated range-Doppler spectrum is used as the train-ing data set to train the network to realize the clutter suppression and noise reduction of the detection samples.Under the constraint of false alarm probability of 10^-7,the detection probability can reach 98%.Finally,aiming at the problem of actual roadside green belt and tall building inter-ference and the problem of low-speed weak target detection,the target detection method based on the YOLO network model is studied.Analyze and improve the YOLOv5 net-work structure and loss function,and use Rep VGGBlock and EIOU loss to optimize the model,so that the improved model has better adaptability and generalization ability to radar data under the condition of lower computational load.The experimental data set in the actual road scene was used for comparative experimental analysis.The results show that the improved algorithm in this thises can effectively complete the vehicle radar target detection task,which further verifies the feasibility of applying the deep learning method to the field of radar target detection.It provides a new idea for the follow-up vehicle radar target detection method based on deep learning.