Vehicle radar has significant importance in autonomous vehicles and modern intelligent transportation networks application.The range and velocity estimation with high precision is one of the challenging tasks in this modern era.This dissertation proposes some valuable improvements in the context of radar signal processing,emphasizes on range and Doppler estimation of moving targets which help to estimate the surroundings efficiently with the help of built-in radar.Firstly,orthogonal frequency division multiplexing(OFDM)waveform is used for data transmission and radar processing while target parameters are efficiently estimated theoretically.In the previous work,authors used the approximations for velocity estimation of moving targets.In this work,Doppler measurement is improved by analyzing the multivariate frequencies of all the subcarriers of received echo with extended FFT and the maximum likelihood(ML)estimation algorithm.After the implementation of ML algorithm,cramer-rao lower bound(CRLB)is utilized to analyze the performance of algorithm.For practical implementation,MR3003 automotive radar is used which follows the frequency division multiplexing(FMCW)radar signal to detect the targets.For some scenarios,radar sensor leads to the false detection while camera output could not work in harsh weather.To overcome this problem,the fusion of both the sensors are used to extract the precise information.In order to perform this task,deep learning algorithm includes convolutional neural network(CNN)is implemented for detection and identification of vehicles in optical video which further accumulate with the MR3003 radar image to identify the moving targets in both scenarios.Furthermore,the performance of the radar algorithms is analyzed on MATLAB while deep learning based CNN is implemented on python.Theoretical study and experimental results reveal that the derived methods can attain the velocity accuracy and achieve a high resolution-velocity estimation of moving targets. |