|
Target detection and tracking technology in infrared image is an important part of modern
military defense system. Infrared dim-point targets detection and recognition under complex
background is a difficulty and important strategic value and challenging research topic. The main
objects that carrier-borne infrared vigilance system detected are sea-skimming aircrafts and
missiles. Due to the characteristics of wide field of view of vigilance system, the target is usually
under the sea clutter. Detection and recognition of the target will be taken great difficulties .There
are some traditional point target detection algorithms, such as adaptive background prediction
detecting method. When background has dispersion-decreasing structure, the traditional target
detection algorithms would be more useful. But when the background has large gray gradient,
such as sea-sky-line, sea waves etc .The bigger false-alarm rate will be taken in these local area .It
could not obtain satisfactory results. Because dim-point target itself does not have obvious
geometry or texture feature ,in our opinion , from the perspective of mathematics, the detection of
dim-point targets in image is about singular function analysis .And from the perspective image
processing analysis , the judgment of isolated singularity in the image is key problem. The
foregoing points for dim-point targets detection, its essence is a separation of target and
background of different singularity characteristics .The image from infrared sensor usually
accompanied by different kinds of noise. These external noises could be caused by the
complicated background or from the sensor itself. The noise might affect target detection and
tracking. Therefore, the purpose of the image preprocessing is to reduce the effects from noise,
also to raise the SNR of image, and to increase the contrast of target and background. According
to the low sea-skimming infrared flying small target characteristics , the median filter is used to eliminate noise, improve signal-to-noise ratio, then the multi-point multi-storey vertical Sobel
algorithm will be used to detect the sea-sky-line ,so that we can segment sea and sky in the image.
Finally using centroid tracking method to capture and trace target. This method has been
successfully used to trace target under the sea-sky complex background.
|
