Descripció del projecte
In the last years, the automotive industry has been addressed toward technological solutions in order to increase the safety and comfort of drivers and passengers. In that sense, both forward and rear looking applications have been included into traditional vehicles, although forward looking is noticeable more developed than rear one. Nowadays, in any vehicle can be found dissimilar data inputs such as cameras, radars, or ultrasonic sensors. The information provided by those sensors can be combined in order to improve the vehicle capabilities and consequently, to provide valuable information to the driver.
This project is devoted to investigate the best way to fuse the information coming from vehicle devices, like cameras, laser scanners, radars or ultrasonic sensors. Each of these sensors can provide significant information by itself; when they are combined we obtain more information and accuracy. Consequently, a valuable source of information can be indeed analyzed to go a step forward through object detection in vehicles. In particular, the goal is to merge and process the information provided by car sensors in order to detect objects in the rear area of a vehicle, including both near and far away zones.
In order to make this possible the best of computer vision and data fusion techniques shall be combined. In particular, we plan to integrate well established practices in object detection with the information coming from car sensors. This integration shall take advantage of the capabilities of each source of data and shall provide enriched information to the decision maker methods. With this proposal, we aim to obtain more accurate detections of objects in rear view environments.
The project will be split in four main steps:
First, the State of the Art on Data Fusion applied to ADAS functions will be studied.
In parallel, a Rear Looking Database containing the ground truth of video image together with data from multiple sensors will be developed.
Then, the research on a new concept of Data Fusion applied to the object detection and/or recognition relevant for the rear looking ADAS technologies will be carried on.
And finally, the validation of the concept on the developed Database and the analysis of the results achieved within the project will be assessed.
The project will be carried in cooperation with the following two institutions:
-Computer Vision Center in the Universitat Autònoma de Barcelona.
-Ficosa ADAS S.L.U.
