Binary particle swarm optimization algorithm for human pedestrian recognition

YANG Ying1, LIU Weiguo2, WANG Youcai1
(1.College of Mechanical Engineering and Automation, Northeastern University, Shenyang 110819, China;
2.Zhejiang Key Laboratory of Automobile Safety Technology, Hangzhou 311228, China)

Abstract: Aiming at cars on the road in front of pedestrian safety issues, binary particle swarm optimization(BPSO) was used to detect, in order to ensure the safety of pedestrians.Firstly,the collected pedestrian sample images on the road changed into feature vectors by twodimensional discrete cosine transform(DCT).The description of the pedestrian was converted to a small number of data points from the image space to the frequency domain space.Use the symmetry of the DCT algorithm,decompress image, get pedestrian image feature vector.Secondly,the features were selected by the BPSO algorithm from the feature vectors,from the the space pedestrian frequency domain,extracting valuable feature subset,in order to get the most representative features of pedestrian,completing the pedestrian detection.The results show that BPSO algorithm in the case of the small sample size,both in realtime detect the correct rate or detection are superior to traditional support vector machine(SVM) algorithm. The results indicate that this research can be efficient and fast detect pedestrians, provide important basis for vehicle active safety technologies, to reduce the number of traffic accidents,that is of great significance.
Key words: pedestrian detection; binary particle swarm optimization(BPSO); discrete cosine transform(DCT); support vector machine(SVM); feature selection