Study On Forewarning Information System Of Human-centered Vehicle Active Safety

The researching on technology of vehicle active safety is in the ascendant now. The R&D of technology of vehicle active safety, such as collision avoidance, to reduce the driver's burden or misjudgment is important to improve the traffic safety. In despite of the large devotion of energy on automatic driving and the result of the researching, there is no one accurate model to describe the driving process veritably and completely, because the driving process is a high intelligentized process. So, it's clear that forewarning is the effective technical means to improve the driving safety. The genesis and development of vehicle active safety technology, the idea of human-centered vehicle active safety and the states of arts of vehicle active safety technology are introduced briefly in this dissertation. In allusion to the researching status, one idea is presented that all the vehicle active safety technology should be human-centered, i.e. driver's safety-centered, driver's perceive identity-centered, driver's operation identity-centered. The characteristic and flow of information and the three phases of driving, i.e. perception, decision-making, operation, are analyzed in this dissertation. A simple model of vehicle-driver-environment is presented based on analyzing the reciprocity of vehicle, driver and environment in the closed loop. The functional structure of human-centered vehicle active safety forewarning information system is formed and the model of forewarning oriented human-centered vehicle-driver-environment is presented to meet the need of safety forewarning. After analyzing the traditional model of human information perception and processing based on attention single resource theory, one conceive of dynamic resource and static resource of attention is formed based on the attention multi-resource theory. The distributing models of dynamic resource and static resource of attention are presented, and based on these models, a new model of human or driver information perception and processing is presented. In allusion to the new model's functional phases, some models are formed, i.e., the model of matching degree of character indexes based on fuzzy measurable function, the priority weight distributing model of information character indexes based on HAP, the information sort classifying model based on subjection degree, the information importance exponent based on information sort subjection degree, sort priority weight and human orientation coefficient. The model of information selection based on the importance exponent is presented. Based on analyzing the characteristics and shortages of traditional decision-making models, a new decision-making model with self-learning mechanism is presented aiming at the basic characteristics of driving task. Some models or scheme are presented for this new model of decision-making, i.e. the feature matching model based on evidence fusion, correspondence analysis model based on weighted Euclidean distance, driving state safety evaluation model based on fuzzy center of gravity, refusal coefficient to realize the selection of driving operation scheme, the self-learning model based on mapping transposition. Driver's safety consciousness measure is introduced to this dissertation according to the role of safety consciousness in the driving task. A statistic model of safety consciousness measure is presented. Driver's safety consciousness will be weighted to audit the driving habits real-time by the model. A model of fatigue accumulation based on fatigue element is presented and the feature fatigue in the phases of perception, decision-making and operation is analyzed with the response curve of unit ramp function. An algorithm of fatigue analyzing based on safety consciousness measure is presented. The human-centered forewarning information system for vehicle active safety is introduced from three aspects, i.e. information capture, risk evaluation, structure of system hardware. Information capture is introduced from road information, host vehicle, obstacle information. An improved optimal threshold algorithm for image sequences is presented to realize the lane segment. An algorithm of image speed sensor based on lane state periodic is presented to measure the speed of host vehicle real-time and exactly. A concept of gray level counts of unit road surface pixel is inducted and based on it, an algorithm of obstacle location is designed. The algorithm of monocular measurement is adopted in this dissertation and the filter of distance data is realized with the theory of entropy. Based on information capture and analyzing a classic driving state, an algorithm of risk evaluation or decision-making based on minimal safety distance is presented.