Surface Displacement Measurement from Remote Sensing Images. Olivier Cavalie
Читать онлайн книгу.
pixel, but entails lower values with multi-looking (see Chapter 3). It varies in range along the swath, and the specification must deliver the worst case: in general, the best values are in the middle of the swath where the antenna gain is maximum. This parameter affects measurements for ground motion: if its value is too poor, then some surfaces that have low backscattering coefficients will not be properly estimated, for instance asphalted roads, tarmacs and even sand in the desert.
Polarization: The design of the antenna sub-system determines what polarization should be implemented. For ground displacement purposes, as the signal-to-noise ratio (SNR) is more favorable, co-polarized data are mainly used, and it is difficult to say whether using HH and VV for offset tracking or InSAR techniques are more advantageous. The use of dual polarization, HH+HV or VV+VH, or quad polarization (HH+HV+VV+VH) is more relevant for other remote sensing applications, such as classification, forest extents and heights, maritime surveillance, pollution at sea or other change detection characterizations. Furthermore, the use of quad polarization generally reduces the swath and azimuth resolution by a factor of two, and thus the size of the archive in this mode is less important.
1.1.3. Parameters specific to optical missions
Pushbroom: A pushbroom camera consists of an optical system projecting an image onto a linear array of sensors. Usually, a focal plane is composed of several time delay integration (TDI) image sensors, mounted in a staggered configuration. The image is directly built at the sensor level. Charge-coupled device (CCD) sensors are used where ultra-low noise is preferred, and now complementary metal oxide semiconductor (CMOS) matrix detectors are increasingly used.
Stereo, tri-stereo and more: To build a disparity map (see Chapter 2), at least two stereo images from separate view angles are necessary. It is possible to combine more images (tri-stereo or even more) to build a more precise map. With four images, all faces of buildings may be seen, and it is possible to build a digital elevation model and associated pixel values. The different images over a geographical point are taken from different incidence angles. If they are taken with the same satellite, then they will be asynchronous. If they are taken with a constellation of satellites on the same orbit, then they may be synchronous, such as with the Co3D system. Asynchronous images imply two acquisition instants. If the delay is a few seconds, mobile elements (such as terrestrial vehicles, bots, planes) or clouds will alter the raw displacement measure. If the delay is a few weeks or months, some buildings may appear or disappear. Evolution of agricultural landscape will also alter the match between the images.
Agile satellites: The agility of satellites is a key function. Agile satellites are able to do attitude maneuvers and focus on one scene. This increases the revisit frequency over the same region, and so reduces the time interval between two acquisitions. Such satellites are able to point towards a geographical point thanks to steering mirrors or by changing their whole attitude (yaw, pitch and roll control).
Base-to-height (B/H) ratio: The shift between two images creates a stereoscopic parallax in one direction (see Chapter 2). The stereoscopic angle (also called the base-to-height or B/H ratio) is the ratio between:
– the distance between the two viewing points (base – B);
– and the distance to the observed scene (height – H).
In the older generation of satellites, the B/H ratio was directly fixed with the instrument characteristics. For example, the SPOT-5 satellite had high-resolution HRS instruments to provide large-area along-track stereoscopic images (forward and backward of the satellite). With agile satellites, it is possible to choose the best view conditions and define the choice of B/H ratio. A large B/H ratio favors the observation of different points of view of the scene (e.g. two faces of buildings) but creates constraints on the disparity computation, as some elements of one scene are not seen in the other one. With a lower B/H ratio, the delay between two images is reduced and a scene can be observed under nearly the same conditions, but it adds constraints to the precision of the disparity computation algorithms. In the case of SPOT-5, the shift between panchromatic and XS detectors (19.5 mm in the focal plane) creates a slight stereoscopic angle that allows a stereo reconstruction using the P+XS correlation. In the case of Pléiades, the B/H ratio is chosen by the user.
The key optical parameters of satellite systems are explained in the following:
Spatial resolution: Spatial resolution is a measure of the smallest angular or linear separation between two objects or two pixels on the ground. It is usually expressed in radians or meters. Spatial resolution decreases as the viewing angle increases. Spatial resolutions are, in general, given at the nadir of the satellite. For example, for Ikonos, the spatial resolution is equal to 0.82 m at the nadir and 1 m at 26° off-nadir.
Table 1.4. Examples of satellite missions and B/H ratios
| Satellite | Mission characteristics | B/H ratio |
| SPOT-5 | Steering mirror (MCV). Stereoscopic angle up to 54 deg | 1.02 |
| SPOT-5 | HRS instrument: angle 40 deg, delay 90 s | 0.8 |
| Co3D | Synchronous acquisitions | 0.20–0.30 |
| SPOT-5 | PAN–XS stereoscopic angle: delay 2.25 s | 0.018 |
| Pléiades | Satellite agility – stereo or tri-stereo | 0.10–0.20 |
| WorldView-3 | Satellite agility – stereo or tri-stereo | B/H = 0.20–0.40 |
Spectral resolution or spectral bands: A ground feature or target is defined by its reflectance and spectral emissivity curve. Its response depends on the wavelength ranges. Different classes of objects in an image can thus be distinguished by comparing their responses over distinct wavelength ranges. A large spectral band captures more signal energy and thus enhances the SNR of the image. A panchromatic band defines a wide spectral band, with high SNR and high resolution. In contrast, a multispectral band (XS) captures specific wavelength ranges separated by filters or instruments sensitive to particular wavelengths: the received signal energy decreases as the bandwidth decreases, and therefore the satellite system is designed with a lower resolution to maintain an acceptable SNR. For a given satellite system, regardless of the spectral bands, disparity estimations are performed with the bands that have the highest spatial resolution and the best SNR. Spectral bands may be separated into several families or groups: thermal infrared (TIR) and medium-wavelength infrared (MWIR) depend not only on solar illumination but also on emission from the imaged objects themselves.
Swath: Swath is the geographical width of the image, which is linked to the optical instrument and the number of detectors in the sensor.
Viewing angle/incidence angle: The viewing angle (satellite point of view) is the angle between the nadir point and the angle towards the terrestrial point. The incidence angle (region point of view) is the angle between the vertical to a point and the direction of the satellite. Thanks to steering mirrors or the agility of the whole satellite, it is possible to modify acquisition angle configurations. A low incidence angle (region near the satellite nadir) means that, for example, only roofs can be seen.