Design and implementation of a Pedestrian recognition algorithm using trilinear interpolation based on HOG-UDP

The need for greater interest in and the study of the pedestrian protection system is growing in the effort to minimize the human and material damages from traffic accidents and to improve traffic safety. Among the numerous methods for distinguishing key characteristics, particularly among those used for detecting pedestrians, the most frequently used method is the one that recognizes key characteristics using Histogram of Oriented Gradient (HOG). The method that relies on HOG for detecting pedestrians, however, requires a massive amount of calculations and is currently seeing a falling recognition rate owing to its unnecessary dimensions. As such, an algorithm with an improved pedestrian recognition rate was designed and realized in this study by reducing the number of dimensions for the characteristic vectors through the minimization of the unnecessary cell histogram calculations and the adoption of the UDP dimension reduction method. The developed algorithm was confirmed from such assessment that the algorithm demonstrated an about 11% processing speed improvement compared to HOG, the existing pedestrian recognition algorithm, and an about 3.98% recognition rate improvement and a 31% calculation speed improvement.