The Road Type Recognition System Of Automotive Proving Ground Based On Multi-Source Information Fusion

The automotive test is an important step of evaluation and verification of the vehicle's overall performance, reliability and quality, which is the key method in accelerating automotive R&D and improving the product quality. There is clear regulation of the road mileage distribution for the testing vehicle in the specification of automotive reliability test which the testing vehicle must follow to drive strictly for receiving sufficient impact excitation. It ensures the accuracy of later reliability analysis. It is the proper methodology to guarantee the automotive testing following the road mileage distribution that is real-time monitor of the category of road and related mileage. And the accurate recognition of road type is the vital technology to realize the accurate monitor of automotive testing. Therefore the research of the recognition system of automotive proving ground road type reflects significance to assurance of the normalization and accuracy. The traditional recognition method based on the vehicle's DGPS position matching with proving ground GIS information has the defect of accuracy decline due to signal shielding. Multi-resource information fusion technology improves the accuracy of measurement and error tolerance via comprehensive processing of multi-sensor information and promotes the accuracy and reliability to higher level. The main research of the thesis is shown below:1. The features analysis of the roads of automotive proving ground is implemented. And the whole system architecture of road type recognition system based on multi-resource fusion is designed.2. The road type recognition method based on road excitation is researched. A road excitation model based on vibration acceleration values of multi-node and vehicle velocity fusion is set up via building quarter vehicle suspension system model and analyzing the measurement theory of automotive dynamics and kinematics. The road excitation feature extracting method based on spatial domain analysis and PSD is researched, which shows better effect and provide a technical fundamental for high accurate recognition of road type.3. Feature extraction methodologies of image and position information are research respectively. The color and texture features of road image are extracted via histogram matching algorithm and SURF algorithm. And the Matching method of vehicle's DGPS information and proving ground GIS data is introduced to support the accurate road type recognition. The Bayes classifier based on image features and ANN classifier based on road excitation features are designed respectively. And the results of each classifiers and DGPS matching are merged by GAR to realize road type recognition system based multi-resource information fusion, which improve accuracy of recognition result4. The hardware and software design of the entire system has been realized. We finished the functional verification and comprehensive test for the system. Finally, the system is proved to recognize the road type of proving ground accurately and stably.