METHODOLOGIES AND APPLICATION Intelligent video analysis for enhanced pedestrian detection by hybrid metaheuristic approach K. R. Sri Preethaa 1 • A. Sabari 2 Ó Springer-Verlag GmbH Germany, part of Springer Nature 2020 Abstract Intelligent video analytics for pedestrian detection plays a vital role for enhanced and effective surveillance system. Since smart city projects are gaining momentum in most of the countries nowadays, enhanced pedestrian detection plays a vital role in the field of security and surveillance. Various classification models were in existence for detecting the pedestrians which suffers from variety of challenges like illumination, pedestrian outfits, gestures, occlusion, lighting, etc., that affects the accuracy of detection. A strong feature vector describing the pedestrian is developed to enhance the accuracy of detection. In this paper, a novel hybrid metaheuristic pedestrian detection (HMPD) approach is proposed to enhance the accuracy of the classifier. HMPD extracts the working principles of support vector machine and genetic algorithm. The proposed model is trained using a set of human and non-human images. The accuracy of the proposed model is tested with benchmarking video data available at VISOR repository. The result clearly shows that HMPD approach produces the maximum accuracy than any traditional approaches. HMPD approach can further be applied in other domains for enhanced security and surveillance. Keywords Pedestrian detection  HOG filter  Video analytics  Metaheuristic 1 Introduction Intelligent video analytics provides invaluable, real-time information for developing event-based video surveillance applications. It gives video-based data about incidents in progress to law enforcement or first responders that helps in enhancing the physical security and ensures public safety. In recent days, video surveillance has wide variety of applications in terms of security enhancement and tamper detection. Intelligent video analytics provides real- time alerting system about an incident in progress for predefined behaviors of people, vehicles and objects. The pedestrian detection is once such area where it helps in detecting the occurrence and activities of human through images from any live stream video. Pedestrian detection can be customized to meet specific needs and operational processes for an environment. It supports the unique requirements specific to alert responses and forensic evidence. Intelligent video analytics has advanced open, standard-based, extensible rich environ- ment to detect, classify and index algorithms that help the user ‘‘mine’’ the index of events for a variety of criteria. In today’s digital world, where the video streams from CCTV cameras play a vital role in surveillance and it is mandatory for improving security and safety in every field. Pedestrian activity detection and storing information are required in area like banks, ATM machines, shopping malls, jewelry shops and most of the other places where an anomalous or abnormal activity analyses might be required. Pedestrian detection has variety of challenges ranging from the different possibility of articulations, variety of human features, variety of appearance, presence of addi- tional occluding accessories, different styles of clothing and frequent occlusion between pedestrians. Despite the challenges, numerous approaches have been proposed on Communicated by V. Loia. & K. R. Sri Preethaa sripreethaakr@gmail.com A. Sabari drsabaria@gmail.com 1 Department of CSE, KPR Institute of Engineering and Technology, Coimbatore, India 2 Department of IT, K S Rangasamy College of Technology, Tiruchengode, India 123 Soft Computing https://doi.org/10.1007/s00500-020-04674-5