Autonomous ships provide the sufficient autonomy and safety level if they able to robustly detect the interest objects in their perception system. Therefore, this part of project tackles research challenges in relation to object detection in autonomous vehicles and presents novel solutions for addressing these challenges.
We evaluated all solutions on a real marine dataset. The dataset was collected in the Finnish Archipelago, which consists largely of maritime vessels in opensea landscapes. This sensor system includes RGB (visible spectrum), thermal infrared (IR) camera arrays, 3d LiDAR and radar.
In summary, the following research objectives and contributions have been delineated:
# Object Detection based on Multi-sensor Proposal Fusion in Maritime Environment [j]:
# Object Detection based on Multi-sensor Proposal Fusion in Maritime Environment [1]:
Most of state-of-the-art object detectors employ object proposals methods for guiding the search for object instances across images. These methods can improve detection accuracy by extracting reliable proposals that contain objects of interest. Moreover, they can considerably reduce computation compared with a dense detection approach such as sliding window by avoiding exhaustive sliding window search across images. We proposed an effective object detection framework based on proposal fusion of multiple sensors such as infrared camera, RGB cameras, radar and LiDAR. such as infrared camera, RGB cameras, radar and LiDAR. Our framework (Fig.1) first applies the Selective Search (SS) method on RGB image data to extract possible candidate proposals which likely contain the objects of interest. Then it uses the information from other sensors in order to reduce the number of generated proposals by SS and find more dense proposals. Finally, the class of objects within the final proposals are identified by Convolutional Neural Network (CNN) as a main architecture of deep learning. Experimental results on real dataset demonstrate that our framework can precisely detect meaningful object regions using a smaller number of proposals than other object proposals methods.
Fig. 1. Overview of the proposed framework. Initial proposals with 933 candidates are first generated by SS and are then filtered using proposal fusion of multiple sensors. After that, the final proposals are classified using CNN[j].
\ No newline at end of file
Fig. 1. Overview of the proposed framework. Initial proposals with 933 candidates are first generated by SS and are then filtered using proposal fusion of multiple sensors. After that, the final proposals are classified using CNN[1].
# References
1. F. Farahnakian, M.Haghbayan, J. Poikonen, M. Laurinen, P. Nevalainen and J. Heikkonen, “Object Detection based on Multi-sensor Proposal Fusion in Maritime Environment”, The 17th IEEE International Conference on Machine Learning and Applications (ICMLA), 2018, US.
