Browse Prior Art Database

Gun-shot Detection for Wearable Camera Devices

IP.com Disclosure Number: IPCOM000247214D
Original Publication Date: 2016-Aug-17
Included in the Prior Art Database: 2016-Aug-17
Document File: 5 page(s) / 891K

Publishing Venue

Motorola

Related People

Kuan Heng Lee: AUTHOR [+6]

Abstract

Wearable camera device (aka body-worn camera) has basic functionality of capturing image, video for court evidence purpose. Using wearable camera device to detect a gun-fire from the video analytics can be an attractive feature for public safety. Here, we propose a video analytics method to detect a rear gun-fire from the police officer, and it gives promising performance.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 38% of the total text.

Gun-shot Detection for Wearable Camera Devices

By Kuan Heng Lee, Prebesh Pavitran, Tih Huang Yeoh

Eugene Chin, Mern Keat Lee and Ahmad Hambaly Alfy

Motorola Solutions, Inc.

 

 

ABSTRACT

Wearable camera device (aka body-worn camera) has basic functionality of capturing image, video for court evidence purpose. Using wearable camera device to detect a gun-fire from the video analytics can be an attractive feature for public safety. Here, we propose a video analytics method to detect a rear gun-fire from the police officer, and it gives promising performance.

PROBLEM

Acoustic has been widely used to detect and locate a gun-fire. However, it is still unable to identify the person that fired it with timestamp. On the other hand, gun equipped with sensor can be used to detect a gun-fire officer but it comes with extra hardware and implementation cost.

SOLUTION

Here we propose a real-time gun-shot detection from the wearable camera based on kickback/recoil detection. Basically, the kickback is the backward momentum of a gun when it is discharged. This kickback momentum can be accurately detected using motion trajectory and blurry calculation. With the timestamp logged, it is then possible to determine the exact timing of a gunshot released by a police officer. The method of detecting the kickback is given in this paper.

OPERATION

Figure 1: Gun-shot detection system for wearable camera devices

In Figure 1, it shows the high-level block diagram of a gun-shot detection system for wearable camera devices. The system consists of camera sensors to support both bright (i.e. RGB) and low-light conditions (such as IR, thermal imaging). A switch mechanism is provided to adapt the right sensor according to the light conditions using the LUX estimator. This ensures that the algorithm can be applied into both bright and dark scenes.

The video-based gun-shot detection is the main engine to detect kickback from the police officer. Once the kickback is detected by the video frame, the acoustic can be used as a complement for the detection (as a gun blast video can be detected earlier than sound).  

Figure 2 depicts the basic flow of the gun-shot detection system for wearable camera device. The video-based gun-shot detection will be triggered after a gun-drawn event (i.e. a gun is drawn out from the holster).

Figure 2:  Gun-shot detection system flow chart for wearable camera device

1.    Video-based gun-shot detection

Figure 3 shows the video-based gun-shot detection in details. It consists of two-hand grip detection, region-of-interest (ROI) and kickback detection.

Figure 3: Video-based gun-shot detection

1.1.  Two-hand grip detection

Depending on the chest position of the body-worn camera, the two-hand grip may look differently as shown below:

 

(a)   Lower chest case: “V” shape

 

(b)   Upper chest case: Reverse “V” shape

Figure 4: Two-hand grip with V-shaped arms

The algorithm detects the two-hand grip gesture by searching the two upper arms in either of “V”...