ADAPTIVE SWITCHER FOR DAY AND NIGHT PEDESTRIAN DETECTION IN AUTONOMOUS VEHICLE AND PEDESTRIAN DETECTION APPARATUS USING THEREOF

摘要

The present invention relates to an adaptive switcher for day and night pedestrian detection in an autonomous vehicle, and more specifically, as an adaptive switcher for day and night pedestrian detection in an autonomous vehicle, using daytime learning data and nighttime learning data to communicate with the daytime environment and It is learned to adaptively determine the night environment, and it is characterized in that the day model or the night model for pedestrian detection is selected according to the day environment or the night environment determined from the input image using the learning result. In addition, the present invention relates to a pedestrian detection device using an adaptive switcher for day and night pedestrian detection in an autonomous vehicle, and more specifically, as a pedestrian detection device, learning using a daytime model learned using daytime learning data and nighttime learning data a pedestrian detector including a night model; and an adaptive switcher that analyzes an input image and selects any one of the daytime model and the nighttime model, wherein the pedestrian detector uses the daytime model or the nighttime model according to the selection of the adaptive switcher. It is characterized by its configuration to detect . According to the adaptive switcher for day and night pedestrian detection in the autonomous vehicle proposed in the present invention and the pedestrian detection device using the same, the adaptive switcher analyzes the input image and automatically selects the day model and the night model according to the input image. Pedestrian detection performance by switching to a model with high pedestrian detection accuracy can improve In addition, according to the adaptive switcher for day and night pedestrian detection in the autonomous vehicle proposed in the present invention and the pedestrian detection device using the same, by configuring the adaptive switcher as a simple CNN model, the It can handle the process of switching to the optimal model without significantly affecting the pedestrian detection speed.