ITC analysis of aerial images


Image needs for computer image analysis

One common mistake is to make the assumption that because imagery is acquired by a digital sensor it is automatically suitable for the production of forest inventories via semiautomated computer image analysis approaches. It is important to point out that the image requirements of computer image analysis can be quite different from that of softcopy interpretation. Obviously, this must be taken into account when specifying the deliverables for an aerial acquisition mission, as the typical products will have been designed for the interpretation realm. The differences are significant on many levels (see Table 2). At a first level, soft copy interpretation typically requires stereo pairs with a good base line and sizeable image overlaps, while orthophotos or long georeferenced image strips are generally preferred for computer image analysis. Generally, the soft copy image interpretation realm favours good image and cursor handling capabilities, faster image loading and roaming speed, and a facility of movement from one stereo pair to another without loosing context and position; while computer image analysis needs "sine qua non" to be able to insure radiometric consistency throughout the area to be analysed.

With careful planning and good discussions with the aerial data providers, it is possible from a single mission to order data in all of the formats necessary for both soft-copy interpretation and computer image analysis, at very little incremental cost.

Table 2. Typical image needs for soft copy interpretation and for computer image analysis.
Soft-Copy Interpretation Computer Image Analysis
Color-balanced stereo pairs (with a good base) with a color-balanced orthoimage mosaic as reference Separate sizeable Ortho images or long georeferenced image strips (no artificial stretches or color balancing)
Pan-sharpened (RGB,CIR, RGBI) are preferred Can benefit from separate panchromatic and prefers as many bands as possible
8 bit images are sufficient and preferred Images reflecting the capabilities of the sensor (8, 12, 14, or 16 bits of radiometry, in 8 or 16 bit image files)
Images are preferred compressed, with a pyramid of resolutions for faster loading and fast scale changes Only lossless compression tolerated. Some additional pyramids for visualization as a convenience
Sizeable ground area covered (for context) with large image overlap for stereo viewing Long strips will improve BRDF corrections and make ITC analysis more efficient (with overlap to concentrate on ±15° off-nadir)

 


Project status

  • On-going