Human anomaly detection is one of the most promising areas of research in the recent past. Automatic detection of multi-class human anomalies will facilitate in understanding more complex actions and its variations. Most of the conventional multi-class anomaly detection models are independent of noise elimination and feature segmentation due to large number of feature space and training images. Also, these models use limited segmented features for multi-class anomaly detection process. As the number of human anomaly classes are increasing, it is difficult to find the multi-class anomaly due to high computational memory and time. In order to improve the multi-class human anomaly detection process, a hybrid multiple feature extraction measures are proposed to find the essential multiple features in the motion vectors for classification problem. In this work, a hybrid convolution neural network framework is optimized by using the non-linear SVM classification problem. Experimental outcomes proved that the proposed multi-level segmentation-based classification model has better human anomaly detection rate than the traditional multi-class segmentation models.

Anomaly detection, feature extraction, classification algorithm, foreground objects, background objects.