Deep autoencoder for false positive reduction in handgun detection

Abstract

In an object detection system, the main objective during training is to maintain the detection and false positive rates under acceptable levels when the model is run over the test set. However, this typically translates into an unacceptable rate of false alarms when the system is deployed in a real surveillance scenario. To deal with this situation, which often leads to system shutdown, we propose to add a filter step to discard part of the new false positive detections that are typical of the new scenario. This step consists of a deep autoencoder trained with the false alarm detections generated after running the detector over a period of time in the new scenario. Therefore, this step will be in charge of determining whether the detection is a typical false alarm of that scenario or whether it is something anomalous for the autoencoder and, therefore, a true detection. In order to decide whether a detection must be filtered, three different approaches have been tested. The first one uses the autoencoder reconstruction error measured with the mean squared error to make the decision. The other two use the k-NN (k-nearest neighbors) and one-class SVMs (support vector machines) classifiers trained with the autoencoder vector representation. In addition, a synthetic scenario has been generated with Unreal Engine 4 to test the proposed methods in addition to a dataset with real images. The results obtained show a reduction in the number of false positives between 22.5% and 87.2% and an increase in the system’s precision of 1.2%$$-47$$ - 47 % when the autoencoder is applied.

MULTICAST: MULTI Confirmation-level Alarm SysTem based on CNN and LSTM to mitigate false alarms for handgun detection in video-surveillance

“…Several works addressed reducing the number of FP and FN in handgun detection in realistic video surveillance environments using pre-processing [16,4] and post-processing approaches [26,19]. For example, the authors in [4] presented a pre-processing technique that help improving the quality of the videos in changing illumination environments.…”

“…They showed that this technique improves the detection of knives in videos. The experimental results in [26] showed that analyzing the detected handguns using an auto-encoder improves the overall precision of the detection in videos. In [19], the authors showed that the use of binarization techniques such as OVO and OVA improves the overall performance of handguns and knives detection.…”

“…The first studies in this context improved the detection in videos by building new datasets [17,10,27,12,11]. The most recent studies applied preor post-processing techniques to further reducing the errors [16,26,4,19]. While other studies used data fusion to improve the overall performance [25,2,18,1,23].…”

Weapon Detection in Surveillance Videos Using Human Inspired Particle Swarm Optimization Algorithm and Support Vector Machine

“…In this application, the closed-circuit television is widely used for recognizing dangerous situations, where the closed-circuit television is considered as the effective operational requirement in terms of safety aspects [1][2][3]. The main purpose of closed circuit television is to provide security, crime investigation, deterrence, and reduction in insurance costs [4,5]. However, the deterrence effects of closed circuit television cameras vary from the different time periods and crime categories.…”

ViolenceNet: Dense Multi-Head Self-Attention with Bidirectional Convolutional LSTM for Detecting Violence

“…Problems arise such as lack of personnel and overlooked threats, given that after 20 min of monitoring a CCTV system, operators fail to detect objects or activities [4,5]. This calls for innovation in automated systems, optionally combined with human attention, for the detection of violent actions [6][7][8] or gun detection [9][10][11]. Safety in all areas of daily life has always been a general concern over time and in all parts of the world.…”