Lolaark Vision Applications
Many imaging operations are performed in bad visibility conditions, leading to poor scene perception by humans and machines. Some common examples are:
Object detection and scene perception in autonomous vehicles and remotely operated vehicles.
- Underwater inspections of ships, dams, bridges, etc. by commercial divers.
- Underwater operations using marine robots and submersibles.
- Performing security operations using surveillance cameras on the ground or airborne.
- Biomedical imaging applications such as endoscopy and confocal microscopy.
- Night vision camera enhancements. Indeed, those cameras are volatile to humid conditions.
- Military and law enforcement operations
All of these applications may suffer poor visibility due to low illumination, haze, fog, precipitation, or water turbidity or smoke.
Camera sensors are unable to capture images with sufficient contrast in such conditions, which makes object recognition and scene perception very difficult and often inaccurate for both humans and machines. These operations are required to be performed in real-time, and bad decisions may cause accidents leading to loss of lives and properties.
The common problem in all the above cases is to find computationally efficient and fast algorithms to enhance videos or images in real-time to improve visibility so that the required task can be performed with much higher accuracy.
We have a novel solution for you
Our team has invented new mathematical models and computational algorithms which can enhance visibility in real-time. Our technology is specially crafted keeping real-time applications in mind. We have a unique method where we first split our camera images into color and intensity information. Both these informations are separately enhanced and then combined, to achieve much farther visibility.
All these operations are performed practically in real-time. Our technology is highly parallelizable, hence we can implement it on programmable Graphics processing units, such as NVIDIA graphics cards. We have developed working prototypes that can be deployed on simple gaming PCs, with NVIDIA GTX 1060 (or equivalent).
Our solution to the pain problem is that you simply integrate our technology as a real-time pre-processing step in your system. This will significantly improve the accuracy of your system.
Our current areas of expertise
Though our technology is useful in many areas of computer vision, robotics, and pattern recognition, our current operations are limited to four key areas.