Galley Proof 15/05/2020; 11:17 File: ica–1-ica200632.tex; BOKCTP/ljl p. 1 Integrated Computer-Aided Engineering -1 (2020) 1–15 1 DOI 10.3233/ICA-200632 IOS Press Deep learning-based video surveillance system managed by low cost hardware and panoramic cameras Jesus Benito-Picazo a,b,∗ , Enrique Domínguez a,b , Esteban J. Palomo a,b and Ezequiel López-Rubio a,b a Department of Computer Languages and Computer Science, University of Málaga, Málaga, Spain b Biomedic Research Institute of Málaga, Spain Abstract. The design of automated video surveillance systems often involves the detection of agents which exhibit anomalous or dangerous behavior in the scene under analysis. Models aimed to enhance the video pattern recognition abilities of the system are commonly integrated in order to increase its performance. Deep learning neural networks are found among the most popular models employed for this purpose. Nevertheless, the large computational demands of deep networks mean that exhaustive scans of the full video frame make the system perform rather poorly in terms of execution speed when implemented on low cost devices, due to the excessive computational load generated by the examination of multiple image windows. This work presents a video surveillance system aimed to detect moving objects with abnormal behavior for a panoramic 360 ◦ surveillance camera. The block of the video frame to be analyzed is determined on the basis of a probabilistic mixture distribution comprised by two mixture components. The first component is a uniform distribution, which is in charge of a blind window selection, while the second component is a mixture of kernel distributions. The kernel distributions generate windows within the video frame in the vicinity of the areas where anomalies were previously found. This contributes to obtain candidate windows for analysis which are close to the most relevant regions of the video frame, according to the past recorded activity. A Raspberry Pi microcontroller based board is employed to implement the system. This enables the design and implementation of a system with a low cost, which is nevertheless capable of performing the video analysis with a high video frame processing rate. Keywords: Foreground detection, feed forward neural network, panoramic camera, convolutional neural network 1. Introduction 1 Increasing public awareness about security issues is 2 caused by the abundance of social conflicts appearing in 3 the media. Research on video ssurveillance systems has 4 attracted more interest as a consequence of this. There- 5 fore, more reliable and accurate systems are sought. 6 The source of the data for these video surveillance sys- 7 tems is often obtained from static and pan-tilt-zoom 8 (PTZ) cameras. For example, In [1], a novel salient 9 motion detection method for non-stationary footage 10 ∗ Corresponding author: Jesus Benito-Picazo, Department of Com- puter Languages and Computer Science. University of Málaga, Bule- var Louis Pasteur, 35. 29071 Málaga, Spain. E-mail: jpicazo@lcc. uma.es. supplied by PTZ cameras is developed. [2] presents 11 a new background subtraction algorithm designed for 12 PTZ cameras capable of performing this task without 13 the need for explicit image registration, and [3] illus- 14 trates a novel method for detecting abnormal behavior 15 in crowded video scenes. The successful operation of 16 such systems depends on their capability to attain real 17 time execution, such as in [4], where a faster patch- 18 based version of Speed-Up Robust Features detector 19 (SURF), named BLS, is introduced as a saliency de- 20 tection method, or the work illustrated in [5], where a 21 tracking-by-detection system that works under impor- 22 tant computational power constraints is presented. 23 The employ of PTZ cameras is commonplace in com- 24 puter vision systems. A good example is the work pre- 25 sented in [6], where a low-power, omnidirectional track- 26 ISSN 1069-2509/20/$35.00 c  2020 – IOS Press and the author(s). All rights reserved uncorrected proof version