This Dashboard Camera Knows When You Are Talking On Phone While Driving
@satya-swaroop-YDeBJM
•
Oct 22, 2024
Oct 22, 2024
1.8K
Researchers at the Santa Catarina State University in Brazil have developed a software and hardware solution for your car that is able to detect and warn you when you are using your mobile phone behind the wheel. Lead researcher Rafael Berri and his team started by installing a camera on the dashboard of the car just behind the steering wheel. They chose that spot to set up the camera because people talking on their mobile phones while driving tend to fix their gaze straight ahead instead of scanning the road like they do normally. After installing the camera the team programmed the system to work in three steps. First, the camera locates the driver and records a cropped video of him/her just focusing on the face and the area to each side of the face. It monitors these locations to find out if you have raised your hands to bring the mobile phone to your ears.
Next the system looks for any skin pixels in the image and plots the location of these pixels. It further divides the image into areas showing the hands and faces calculating the probability of the driver talking and warns accordingly. The team put their algorithm to the test on a set of five videos shot by a dashboard camera. The camera shot videos with a resolution of 320 x 240 pixels at 15 frames per second. The videos were divided into three second ones and sent to the system for analysing. The system had an efficiency of 87.43 percent. The team warns that the system isn’t perfect and in some cases the accuracy of the algorithm drops drastically such as direct sunlight on the driver’s face.
The system however is not completely fool-proof as #-Link-Snipped-# points out that the researchers have yet not finalised a warning system and the system has no provisions to check whether the car is moving or not. It also cites privacy concerns of constant monitoring by a video camera.
Next the system looks for any skin pixels in the image and plots the location of these pixels. It further divides the image into areas showing the hands and faces calculating the probability of the driver talking and warns accordingly. The team put their algorithm to the test on a set of five videos shot by a dashboard camera. The camera shot videos with a resolution of 320 x 240 pixels at 15 frames per second. The videos were divided into three second ones and sent to the system for analysing. The system had an efficiency of 87.43 percent. The team warns that the system isn’t perfect and in some cases the accuracy of the algorithm drops drastically such as direct sunlight on the driver’s face.
The system however is not completely fool-proof as #-Link-Snipped-# points out that the researchers have yet not finalised a warning system and the system has no provisions to check whether the car is moving or not. It also cites privacy concerns of constant monitoring by a video camera.