In this paper

we propose a coarse-to-fine pedestrian detection based on feature transformation and SVM with a monocular moving camera.In this method

a coarse pedestrian detector is learnt by Look-Up Table (LUT) Gentle AdaBoost cascade.Then each stage classifier in coarse detector is taken as a feature

and a fine pedestrian detector based on those features is learnt with SVM from those training data which pass through the coarse pedestrian detector.The detection results by this detector are refined by temporal analysis using the color and spatial information of each detection results to improved pedestrian detection rate and decrease the false alarm rate.Experimental results show our method high performance.