| 摘要 |
<p>The present invention relates to a method for recognizing pedestrians based on a far infrared camera at night which receives a thermal image of pedestrians from a far infrared camera, sets a candidate group using a difference of Gaussian (DoG) filter having strong resistance to image noise, and accurately recognizes the pedestrians using a sorter according to pedestrian behavior characteristics. According to the present invention, the method for recognizing pedestrians based on a far infrared camera at night comprises the following steps of: receiving a thermal image of pedestrians; obtaining a first response value from each pixel in the inputted image using a DoG filter; measuring a variance of each pixel in the inputted image; generating a binary image by binarizing a second response value of each pixel calculated by multiplying the first response value and the variance; generating pedestrian candidate group clusters in the binary image; recognizing pedestrians by classifying the pedestrian candidate group clusters by one or more sorters learned according to pedestrian behavior characteristics; and calculating pedestrian locations.</p> |
