Air defense and anti-missile operation usually demands the EO device to quickly detect the sea-skimming missile in a given area. Because the small sea-skimming target usually features low brightness, small size and susceptible to interference from the sea horizon and sea surface clutter. Therefore, a quick search and threat evaluation algorithm is proposed against sea-skimming small targets. First, statistic row mean value and gradient are worked out, and the sea horizon is fitted by means of least square criterion. The background shall be suppressed by using morphological factor and the image shall be binarized by using self-adaptive threshold. Finally, statistic suspicious target features are carried out for threat evaluation of these suspicious targets, sea clutter and sea horizon interference are eliminated and the high priority threat is picked up.
Stereo vision matching is to search the corresponding relation of one spatial object observed from different angles in the projected image, and obtain the parallax image depending on the deviation (parallax is the geometric distance between different points projected from the same spatial point on different image. In a parallel binocular stereo vision system, two cameras have the same focal length and parallel imaging planes, so there is no rotation and scale conversion for images obtained by two cameras. SIFT binocular stereo vision system does not need multi-scale conversion and coordinate axis rotation, therefore the algorithm is simplified and the fault tolerance for SIFT target matching is maintained.
The target identification of the sea battlefield is the prerequisite for the judgment of the enemy in the modern naval battle. In this paper, a collaborative identification method based on convolution neural network is proposed to identify the typical targets of sea battlefields. Different from the traditional single-input/single-output identification method, the proposed method constructs a multi-input/single-output co-identification architecture based on optimized convolution neural network and weighted D-S evidence theory. The simulation results show that