We now come to the last week of our course. In it, we will look at some final topics in radar, including some innovative applications and say something about how different remote sensing imaging data types can be used together. First though, we look at distortions in radar images. The first form of distortion is shadowing. While not exactly a distortion, it does affect interpretation. As with optical images, shadows are caused in radar when signal cannot be received from behind an object. Unlike optical image shadows, however, shadowing in radar imaging is absolute. For nadir viewing optical sensors, it is possible sometimes to detect measurable signals from shadow zones because of atmospheric scattering of incident radiation into the shadowed regions at fairly short optical wavelengths. Radar shadowing is likely to be most severe in the far range and for larger angles of incidence. Whereas it is often non existent for smaller incidence angles. That can be appreciated by looking at the diagram on this slide. The first real geometric distortion we encounter, is that to do with the changing range resolution across this width. It is called near range compressional distortion. Recall that the ground grains resolution is best at far range and poorest at near range. Thus, the near range pixels cover a larger region of the surface than the far range pixels, and yet both are made the same size in the display product as illustrated in the diagram here. As a result, near range features are compressed into smaller than realistic display pixels. We see the impact of that on the next two slides. Consider a region on the ground in which there is a square of grid like features, such as field boundaries. Within each of the square cells, there could be many pixels. Imagine also that there are some diagonal lines as shown. They could be roads connecting across field corners. The right hand figure shows how that region on the ground will appear in recorded and displayed radar imagery. Not only do the near range features appear compressed, but linear features at angles to the flight line appear curved. The combined effect is as if the image were rolled over on the near swath side. The image at the top of this slide shows the effect again, but with a real image recorded by an aircraft radar. The bottom image has been corrected. We now look at the distortion peculiar to radar imaging. It arises because objects are delineated in the range direction by differences in time delay. As a result of that, how would a tall tower appear in the range direction in a radar image? The radar echo from the top of the tower arrives back at the radar before that from the base because it travels a shorter two way path. That causes the tower to lie over towards the radar on the image. To emphasize that, we draw concentric circles from the radar. All points lying on one of those circles, will be at the same slant ranged from the radar, and we'll create echoes with the same time delay. By projecting the circle which just touches the top of the tower onto the ground, we see that the tower top and indeed the whole tower superimposed on ground features closer to the radar, than the base of the tower. This effect is referred to as layover. By comparison, in optical imagery, vertical objects appear to lie away from the imaging device, since they are superimposed on features further from the device than the base. We now look at relief displacement, which is similar in origin to layover. Instead of a tower though, consider a vertical feature with some horizontal dimension, such as the model mountain shown in the diagram. Using the same principle of concentric circles to project the vertical relief onto the horizontal ground plane, several effects are evident. The front slope is foreshortened, the back slope is lengthened. Together, these effects suggest that the top of the mountain is displaced towards the radar set. In ground range format, they give the effect that the mountain is lying over towards the radar. If you go back to the Santa Barbara image in the last lecture, that relief distortion is quite evident in the 20 degree Seasat image. If we knew the local height, then the degree of relief displacement can be calculated. In principle, therefore, the availability of a digital terrain map for the region should allow relief displacement distortion to be corrected. Not only is the relief displaced, but the brightness of an image is modulated by topography, particularly in mountainous regions. On front slopes, the local angle of incidence will be smaller than expected, and thus the slopes will appear brighter. On back slopes, the angle of incidence will be larger than expected, making them darker than would otherwise be the case. This slide shows an example of relief displacement in a mountainous region, again with 20 degrees Seasat radar imagery. Slopes facing the radar illumination direction appear bright, while those away from the illumination appear darker. There is rather severe terrain distortion evident in the small circle, most easily seen by comparing the top image with the optical image at the bottom. In summary, because the formation of a radar image depends on resolving time delays in the range direction, several unusual geometric effects happen, which are not apparent in optical imagery. First, the recorded image appears compressed at near range, as against far range for optical imagery. Secondly, tall structures lay over towards the radar. Thirdly, relief is displaced towards the radar. Remember also, the displayed brightness of an image is modified by relief, according to the scattering characteristics of a surface at different angles of incidence. Finally, if mountainous terrain where covered by volume scatterer, such as shrubland at C band, relief modulation of brightness would be less evident. The first three questions here should help consolidate your understanding of terrain distortions in radar imagery.
