The increase in the crime rate has led to the deployment of surveillance systems to different places such as airports, malls and schools in order to prevent crime. However, inaccurate detection of anomalous behavioural patterns due to false detection (false alarm) errors is the main challenge that affects the performance of surveillance systems. Although several interesting techniques have been reported in literature for the detection of anomalous behaviour using semi-supervised and supervised techniques, these still require improvement in terms of accuracy and minimizing of false detection errors in crowded environments. This paper presents a performance analysis of the convolutional neural network with long short-term memory (CNN-LSTM) in surveillance systems. The image features are extracted from the image frame sequences using the CNN, while the LSTM uses the gate mechanism to keep vital information for remembrance. The results are compared with existing detection models, including the mixture of probabilistic principal analysis, motion deep net, social force and dictionary-based models. Experiments are done on the University of California San Diego dataset using the proposed anomalous behavioural pattern detection system. The results obtained show that the proposed system outperforms the others mentioned with 86% accuracy.

In this paper, geometrical analysis is used to extract novel straight line segment features from the wings around the peaks of the Hough Transform (HT). Based on these features, a practical segment detection method is proposed which has the ability to determine complete straight line segment parameters including the location of the center, length, slope and the Euclidean distance to the origin. The proposed method does not rely on edge point verification in the image space, i.e., the complete set of segment features are determined only using the information embodied in the HT data. The proposed method can distinguish between highly collinear straight line segments. Segment detection is robust to disturbing edge points, especially ones collinear with the object. A predefined collinear segment resolution that provides a theoretical criterion to determine straight line contiguity is derived. Image processing and analysis experiments show consistent robust performance.

The distance between a straight line and a straight line segment in the image space is proposed in this paper. Based on this distance, the neighborhood of a straight line segment is defined and mapped into the parameter space to obtain the parameter space neighborhood of the straight line segment. The neighborhood mapping between the image space and parameter space is a one to one reversible map. The mapped region in the parameter space is analytically derived and it is proved that it can be efficiently approximated by a quadrangle. The proposed straight line segment neighborhood technique for the HT outperforms conventional straight line neighborhood methods currently used with existing HT variations. In contrast to the straight line neighborhoods used in existing HT variations, the proposed straight line segment neighborhood has several advantages including: 1) the detection error of the proposed neighborhood is not affected by the length of the straight line segments; 2) a precision requirement in the image space described using the proposed distance can be explicitly resolved using the proposed formulation; 3) the proposed neighborhood has the ability to distinguish between segments belonging to the same straight line. A variety of experiments are executed to demonstrate that the proposed neighborhood has a variety of interesting properties of high practical value.