On The Method Of Vehicle Detection Based On Machine Vision And Millimeter Wave Radar Fusion

In recent years,the technology of assisted driving and unmanned driving has developed rapidly,vehicle environment perception is one of the important research orientation.The multi-sensor fusion method can effectively make up for the deficiency of single sensor detection,improve the safety of driving.The front vehicle detection method based on the fusion of vision and millimeter wave radar is presented to realize a comprehensive and intuitive detection of the front vehicles.The main work of the thesis is as follows:1.The forward vehicle real-time detection based on machine vision is realized by the improved Adaboost algorithm.Firstly,the traditional Adaboost algorithm is improved,according to the training error number and classifier error rate of the samples,the weights of the training samples are constrained to avoid the over-adaptation of the algorithm;Then the cascades classifier is trained according to the improved Adaboost algorithm and Harr rectangle feature;Finally,the performance of the classifier is tested,and the results show that the improved algorithm has a higher detection rate.2.Millimeter wave radar is used to detect the vehicle targets in front.Firstly,the radar detection data is preprocessed to obtain the information of the forward targets;then the vehicle target is selected according to the data characteristics,and the empty targets,low risk targets and interference targets in the data are filtered out;Finally,the selected effective targets are verified and processed according to the predicted value and measured value,and the effective targets are visualized.3.The fusion model of machine vision and millimeter wave radar is established and the model includes space fusion and time fusion.Firstly,according to the principle of camera imaging,the conversion model of radar data to image pixels is established to realize the spatial fusion of the two sensor data;Then the time fusion method of the sensor data is determined according to the working frequency of the two sensors;Finally,the parameters of the fusion model are calibrated by using Zhang Zhengyou calibration method and four-point calibration method4.The machine vision and millimeter wave radar fusion detection platform are built.Firstly the system hardware is selected and installed,and the software system is designed to realize the fusion model and detection method.Then the real car experiments are carried out in real driving environment,the experimental results show that the fusion method proposed in this thesis has high accuracy,and compared with the single sensor detection system,the fusion detection system has higher detection rate and stability,and has good real-time performance.