Merging two camera feeds in real time
Softeq developed a PoC solution that merges two camera feeds into one wide-range video and processes it in real time. Video data will be stored on the gadget’s internal SD card. If a driver gets into a car accident, the system will automatically send the video files to the cloud.
Softeq created a proof-of-concept solution for real-time video stitching
The client addressed Softeq to develop a PoC solution capable of real-time video stitching. They wanted to develop a smart driver assistance solution—a dashcam-like device installed on the windshield. The solution needed to detect accidents, check on the driver, and call emergency services if required.
The solution also had to record HD videos for car insurance claims, and store them on the device’s SD card. In the event of an accident, the system needed to automatically send the video files to the cloud.
Softeq developed a PoC solution that combines two video feeds to create a single panoramic video. Due to the apparent difference in the object position caused by different camera angles, simply merging the feeds without additional processing was not enough to create a seamless video. For this reason, we developed a Linux board support package that enables the solution to apply the necessary video corrections in real time. The solution is based on the NXP processor and supports Wi-Fi and LTE connectivity for sending data to the AWS cloud.
1. Applies synchronous exposure and white balance corrections to video footage obtained from both cameras
2. Projects the videos on a 3D sphere
3. Merges the feeds using a smooth stitching algorithm
4. Maps the merged video on to a two-dimensional plane
5. Removes the fish-eye effect using the squeezing and stretching algorithms
Softeq developed a PoC solution that helps create panoramic videos by merging two camera feeds in real time. The solution is able to cover a field of view of more than 200 degrees.
Currently, we are working on an MVP version of the driver assistance solution, which will also support Alexa. Next, we will develop a geofencing function that notifies drivers when they enter an unsafe area.
Our solution could serve as a basis providing data needed for further analysis should the client decide to create an ADAS system that helps drivers avoid on-road collisions.