Informatics Publishing Limited

P. Karpagavalli ¹, V. G. Janani ¹, A. V. Ramprasad ¹

Affiliations
1 K.L.N. College of Engineering., Sivagangai, IN

Abstract

In this paper, we consider a human activity recognition approach that only requires single video example per activity. We introduce silhouette segmentation for extracting the foreground objects from background video. We used to represent a single human activity video by using bounding box representations, from that we can calculate height, width, aspect ratio, blob area, normalized bounding box specifying the sub-event’s spatial location, averaging the coordinates of all bounding boxes and Centroid, where human motion and body configuration is observed and tracked. We present an approach for measuring similarity between visual entities (images or videos) based on matching internal self-similarities .This intrinsic information is used together with statistical features and shape based approaches to recognize and classify Human activities. Spatio temporal features can be used for feature extraction of particular activity .Further; we had used K-NN, RVM, SVM classifiers in order to recognize the abnormal activity form human activity by using active learning algorithm and also measure the performance between them.

Keywords

Activity Recognition, Active Learning Algorithm, Ada Boost Classifier, K-NN Classifier, RVM Classifier, Silhouette Segmentation, SVM Classifier, Spatio-Temporal Features.

Full Text

     

P. Karpagavalli ¹, T. Saranyaa ², A. V. Ramprasad ¹

Affiliations
1 K.L.N. College of Engineering, Sivagangai, IN
2 K.L.N College of Engineering, Sivagangai, IN

Abstract

Pedestrian counting plays an important role in public safety and intelligent transportation systems. The techniques of tracking a single pedestrian becomes impractical and complicated, when the scenes are densely crowded. In this paper, first step is extracting foreground objects from the background by applying ROI mask and then the pedestrians are detected by segmentation method. Second, the large sets of features extracted are reduced by using SLFs in order to attain high accuracy for counting number of pedestrians. The bounding box representation is used for counting number of pedestrians in a crowded environment. The number of people can be estimated by using connected component labeling method. Finally, by using above features reliable people counting in crowd environment can be achieved. The proposed method is robust for illumination changes and working well in public places for safety of the people, such as railway station, hospitals, shopping malls, etc. For further implementation, consider the proposed approach on a dataset consisting of a large number of people and study how to refine the computational efficiency for large scale datasets.

Keywords

Pedestrian Counts, Region of Interest (ROI), Statistical Landscape Features (SLFs).

Full Text

     

K. Menaka ¹, B. Yogameena ², P. Karpagavalli ¹

Affiliations
1 Electronics and Communication Engineering Department, K.L.N College of Engineering, Madurai, IN
2 Electronics and Communication Engineering Department, Thiagarajar College of Engineering, Madurai, IN

Abstract

Automatic visual surveillance systems could play an important role in supporting and eventually replacing human observers. The proposed frame work contains four modules viz detection, tracking, feature extraction and action classification. The person has to be found in the image plane first, using robust background subtraction by modifying the parameters in Gaussian Mixture Model (GMM), assuming a static camera. Centroid based tracking is done consecutively. Silhouette shaper based features are extracted, including aspect ratio of minimal bounding box of human silhoutte, approximated eccentricity, normalized central moments. Human actions are classified by grouping, enhancing efficient hierarchial clustering. In this paper a multi-sample-based similarity measure has been proposed where HMM training and distance measuring are based on multiple samples. These multiple training data are acquired by a dynamic hierarchical clustering (DHC) method. By iteratively reclassifying and retraining the data groups at different clustering levels, the initial training and clustering errors due to overfitting will be sequentially corrected in later steps. Human behaviours are obtained from clustered data set and trained neural network kit to describe their appropriate classes. The results are evaluated by means of sensitivity, specificity and accuracy.

Keywords

Video Surveillance, Background Subtraction, Gaussian Mixture Model, Minimal Bounding Box, Feature Extraction, Hierarchial Clustering, Neural Network.

Full Text