How VE Helped Create a Gunshot Locator
External Hardware
A gunshot locator is a system that uses acoustic, vibration, optical, or maybe other types of sensors, or a combination of such sensors, to detect and communicate the position of gunshots or other weapon firings. It helps:
Brains Behind the Idea
An ex-policeman with knowledge of the American Emergency Response System had the idea that would become Beverly Security. James Oliver, a former law enforcement officer, created the Absolute System- Gunshot Locator, to promptly notify authorities of a shooting.
"We are a better country than having kids getting killed. I think we should do something about that.”
Organized Crimes Division 30 years of experience
When he retired, James started working on an idea of a product that could detect guns early and help in the rapid response of law enforcement agencies to minimize casualties. Further in stressful situations, there is always the danger of the perpetrator getting away. He wanted to help the enforcement agencies effectively identify the perpetrator.
How Does It Work?
Reducing response time to a few seconds
The Absolute System is an indoor gunshot detection system. In the event of an emergency, the system alerts the monitoring company’s monitoring center which is a typical call center equipped with video monitoring. The center immediately takes control of the surveillance cameras installed on the premises.
Since the Absolute System sends the GPS location, the monitoring center specialists can view the maps and also provide the geographical location details to the authorities.
Breaking Down the Design Process
Used Mask R-CNN for Object detection
R-CNN is short for “Region-based Convolutional Neural Networks”. These are a family of machine learning models for computer vision and specifically object detection. The main idea is composed of two steps. First, using selective search, it identifies a manageable number of bounding-box object region candidates. And then it extracts CNN features from each region independently for classification.
Mask R-CNN extends R-CNN to pixel-level image segmentation. The key point is to decouple the classification and the pixel-level mask prediction tasks. The mask branch is a small fully-connected network applied to each RoI, predicting a segmentation mask in a pixel-to-pixel manner.
Implementing the Understanding
We started by downloading images of guns and swords from the open image dataset and annotating them using the VIA tool.
We made a copy of the samples/balloon directory in Mask_RCNN folder and created a samples/guns_and_swords directory where we continued our work.
Once we were done with setting up the coding environment, we moved on to the next step of visualizing the masks and images.
To train the Mask R-CNN model, on the Guns and Swords dataset, we initialized our model with COCO weights and visualized the losses using Tensorboard.
Gunshot Detection Sensors Design
Understanding the application of sensors
A gunshot detection system is typically comprised of one or more sensors (optical and/or acoustic) that accurately and reliably detect live ammunition (gunshots) being discharged.
Acoustic sensors detect the specific acoustic signature. To ensure it is a gunshot, it also records other sounds, including human voices, that occur within the vicinity of its microphones during a suspected shooting incident.
Infrared sensors detect what’s known as muzzle flash – the visible light created when a shot is fired that is caused by the combustion of gunpowder mixing with the ambient air.
Room for Design Challenges
Every hurdle is a chance to improve
Issue: US police doesn't allow and accept the automated calling.
Solution: External call centers were engaged in the process of raising alarms and calls.
Issue: US discontinued the use of 2G technology in the year 2018.
Solution: The system was updated to the newer version that was 4G enabled.
Issue: During the pandemic, the US witnessed a shortage of 4G SIMs.
Solution: The launch had to be pushed further to ensure that compatible chips/SIMs are available.
Solution to Results
Attaining results beyond expectations
Immediate and precise gunshot detection
No wired data connection, only a high/low voltage power supply
Universal integration with building security system
Automatic notification to 911 and building security
How the Detection System Works?
Meet the Man of the Hour
Brains behind this operation
Venugopal has over 25 years of experience in the industry. At Virtual Employee, he has worked on 4 IOT projects based on PIC32, nRF52, ESP32, and NXP1769 microcontrollers.
- Embedded systems
- IoT
- Telecommunications PSTN/VOIP & Networking
- Code generation
- Knowledge Cloud platform
- Micro-services
- Information Display Systems
