dc.description.abstract | Due to the frequent traffic accidents, the development of vehicle safety systems becomes more and more important. Nowadays, the demands for vehicles are not only convenience and comfortability but also the needs of vehicle safety which can detect dangerous states and issue warnings for drivers. Currently, the vehicle safety systems are separately developed one after one; each system has its own detection and warning modules. If a driver uses more than one system in his vehicle, the multiple warning signals maybe confuse and interfere the driver. On the other hand, the sensor-based detection systems can only detect the short-distance dangerous states; the long-distance detection functions were seldom considered. Thus, in this study, we propose a long-distance detection and warning system and then propose a fuzzy-based fusion model to combine the multiple warning data from all considered short-distance and long-distance detection and warning systems to provide the drivers only a single significant warning signal.
The proposed fusion model integrates Land departure Warning (LDW), Forward Collision Warning (FCW), Curve Speed Warning (CSW), speeding warning, and crossroad warning systems. GPS and electronic map provide speed limit on current road and the information of forward crossroad to warn drivers of the danger. CSW reminds the forward road information based on GPS and electronic map. CSW system computes the curvature radius of 100 to 250 meters road ahead and issue warning if current speed is higher than the critical speed in such radius.
Based on the experiments of several street sceneries, it is certified that the proposed system can successfully integrates FCW and LDW with CSW, speed warning and crossroad warning via GPS and electronic map. Then these five warning data are feed into the proposed fuzzy inference system and fuzzified into fuzzy singleton. We choose max-min inference engine and check the 44 rules in the rule base to conclude all driving situations. Eventually, we use center of area defuzzifier to defuzzify the final conclusion membership function to obtain a single warning. The experimental results show that the proposed system is reasonable and useful for practical ADAS applications.
| en_US |