Studies on Various Type of Human Detection Algorithms for Multiple and Occluded Persons in Static Images M. Shanmugasundaram 1 , N. Shanmuga Vadivu 2 1 Dept. of CS, College of Computer Science & Information Systems, JAZAN University, Kingdom of Saudi Arabia. 2 Dept of ECE, RVS College of Engineering and Technology, Coimbatore, Tamil Nadu, India. Email: shanmugavadivun@gmail.com Abstract: Detecting and tracking human in still or video images provides a promising technology development and solution to many real world problems. Moreover, detecting human may be the frst step to put forward the next logical steps for many applications. But, it is a challenging task due pose, dresses, color and occlusion. This paper proposes a study of human detection in static images in different view. In the literature, numerous works had been proposed to detect a single human in an image. So, the survey has been conducted for detection of multiple humans without occlusion, detection of multiple human with occlusion and human detection in fused image. Due to the diffculties found during the process of human detection such as occlusion and shadow, people in group, main focus has been given to multiple-human detection. Keywords: Human detection, Pose, Occlusion, fusion, Machine learning, Object detection, Feature extraction I. IntroductIon The process of object detection and tracking in static and video images leads to lot of research interest in certain areas such as surveillance, medical and security. Detecting human object provides a lot of exposure in many felds. It is a next logical step after the successful development of face detection algorithms. It plays a vital role in providing solutions to many research requirements and is very useful in pedestrian detection and surveillance. Over a period of decades, algorithm developed for human detection has been getting more popular and more focus has been given to the enhancement of existing work in the literature [24][33][36]. Initially, interest was given to single human and it has been exponentially increasing to multiple human objects. Even though most of the research is going towards detecting human, the researchers feel more diffculties to detect and recognize human object due to its wide variability in appearance. Moreover, some other criteria adds more challenges like clothing, different articulations, textures and styles, variations Article can be accessed online at http://www.publishingindia.com in illumination and clutter backgrounds [4][8]. In the literature, algorithms are developed to detect a single human [2], multiple human [35] and occluded human. Recently detecting the human object from fused image gets more popular and number of applications uses different fusion techniques [36][39][40]. Technically, algorithms for human detection are organized into two parts: 1. Feature Extraction and 2. Object Classifcation. Initially, the process of extracting features includes some preprocessing steps such as noise removal and segmentation. Classifcation mainly involves identifying the type of objects. Machine learning is one of the widely used concepts to classify the features. By using this, a detector used to recognize only human object is constructeld from a large number of training examples. Scholars have mainly been using two such a machine learning algorithms which are SVM [5][1][2] and Adaboost [11][8]. Object detection can also be done using edge detectors such as Sobel, canny etc. Section-II gives an overview description of the proposed techniques. Analysis of multiple human detection is shown in section-III. Section-IV provides a detailed study of multiple humans with occluded condition is discussed. Finally section-V deals with detections in fused image. II. classIfIcatIon overvIew The survey conducted to this issues is classifed into single human detection, multiple, occluded and fusion based human detection. Suffcient work had been carried out to detect single human object successfully. So, this paper does not focus much on this. Section-III gives the various techniques used for detecting multiple human in a single image. Several humans appeared in a single image with occluded situation is given in section-IV. Finally, recovering human objects in a fused image which is combined two images of the same scene captured by two different devices is presented in section-V. Techniques given in this paper is organized as shown in Fig. 1. Journal of Applied Information Science Volume 5 Issue 2 2017 ISSN.: 2321-6115