The System Design And Implementation Of Pedestrian Detection And Encrypt Privacy Protection

Pedestrian detection and recognition is a hot topic in the field of computer vision at present. It’s widely used in some practical applications, such as privacy protection, development of robots, intelligent transportation, human motion analysis and intelligent video surveillance.In recent years, various algorithms based on pedestrian detection have been proposed, but because of the non-rigid body, the complex background, the keep out between human body, lighting conditions, wearing, body size, and the influence of walking posture, make the effect of the pedestrian detection is not ideal in practical application. Currently, there is no a stable performance and general pedestrian detection algorithm.The pedestrian detection core steps consist of feature extraction, classifier training and human detection. In this paper, we choose the HOG feature, and sample pictures are from the INRIA pedestrian data sets and MIT pedestrian library, the pedestrians in two data sets are in standing posture, and form different movements, different clothes and background. First, extract the HOG feature of positive samples and negative samples; Then,these feature are input to support vector machine (SVM) for training, we can get a classifier can be used for detection; Finally, images to be detected to this classifier performs detection.At the same time, for personal privacy protection, this paper adopt the two-dimensional fractional Fourier transform(2D-Frft) to realize encryption on the detected parts, and those encrypted parts can be decrypted if necessary. Compared to the traditional Fourier transform, fractional Fourier transform can be a better expression of signal frequency characteristics, and is very sensitive to the order time. Fractional Fourier transform is widely used in image encryption system. Two-dimensional fractional Fourier transform has increased the range of keys, to further improved the security of the system. In this paper, first of all to extracted the human body part that has been detected. Then the extracted images are encrypted by using fractional Fourier transform with certain order respectively in x and y directions. In order to decryption, we can use the fractional Fourier transform inverse with the same order.At last, based on the research of pedestrian detection and encryption technology, this paper designed a human-computer interaction interface, integrated the pedestrian detection and encryption function, make the operation more intuitive and simple.