Frequently Asked Questions
Photomontage or orthophoto?
Photo editing programs in the realm of aerial surveys?
There are certainly a lot of attempts to use the alternative methods of merging aerial photos with full area coverage (prepared using overlapping, covering the working area with flight lines). Similarly, to panoramas, however, the quality of merged aerial photos is never going to be anywhere even close to the geometric quality (spatial accuracy) of maps, and what is also important: preparing them is a lot more time-consuming than the processing method based on beam-equalisation managed in a single block and providing a high-quality end-product.
A brief summary of our arguments:
- Georeferring the merged images is never going to result in end-products nearly as precise and reliable as in the case of an appropriately prepared orthophoto-mosaic.
– In the case of 20-30 images spatial photogrammetry provides a much faster image processing solution.
– Orthophotos are products of aerial cartography accepted all around the world, while photo montages are accepted as products in the rarest of cases.
– Nowadays software containing even the most sophisticated algorithms are available at an affordable price. Moreover, there are free solutions available as well, which, if used with an appropriate level of preparedness, may produce very good results.
– Orthophotos are easier to merge (mosaicking), as they precisely fit from the beginning
- A three-dimensional surface model is created, which constitutes a product both by itself or with the orthophoto as a texture
– Two-dimensional photogrammetric products falsely claimed to be orthophotos shall be deemed to be deceiving the Principal, which may entail serious legal consequences.
Scale
A lot of questions were asked in connection with the scale of the photogrammetric products. This is why we would primarily like to answer the questions revolving around the topic:
Instrument scale (m): The scale of the image displayed on the image sensor or film.
Image scale (M): In the case of image or the end-product the scale of the map is the quotient of the digital image and the mapped surface.
The
print scale (M) is the quotient of the actual printing size interpreted at the given image resolution (dpi, ppi) on the mapped surface.
The instrument scale can be calculated as the quotient of the value determined by the product of the field resolution and the side length of the sensor expressed in pixels and the side length of the sensor, or as the quotient of the relative flight altitude and the focal length. The recording device projects the surface detail recorded at the given moment of exposure onto a chip (film frame or sensor chip). This chip records elementary pixels the size and number of which varies in the case of each type, so sometimes recording with a higher resolution takes place on a smaller size, rectangle-shaped area than in the case of a bigger chip operating with less elementary pixels.
Therefore the instrument scale, which was an existing notion even during the film era, expresses the ratio of the image projected onto the recording plane of the instrument to the mapped surface. Thus instrument scale can be calculated as the quotient of the relative flight altitude and the camera coefficient, or as the quotient of the bandwidth and the size of the image projected onto the chip.
This, however, only provides information about the detail level of the end-product if the technical parameters of the camera are known.
The scale of the map (the real-size print scale of the end-product) can be, amongst others, calculated as the quotient of the width of the mapped ground surface and the width of the sensor chip. Therefore, it is the aspect ratio that provides exact information about the detail level of the end-product. Aspect ratio can be calculated as the quotient of the distance of two known points in the field and their distance in the image.
Print scale is the value of the scale of the map adjusted by the printing parameters. (The quotient of the field resolution and the printing size of the elementary pixel multiplied by the printing resolution expressed in dpi.)
Therefore, in the case of digital surveying the level of detail of the survey is only worth providing with scale if the value of dpi is also provided.
Theoretically, digital maps are scale-independent, the stored numeric data can be displayed in any magnification, but they become unsharp when over-magnified as compared to their original size. Therefore, the manageability of data files and the technical usability of data limits magnification or shrinking to any size.
The accuracy and density of data determine the potential uses of data.
-The accuracy of the digital data system depends on
- the accuracy of determining the points and the rate of sampling,
- the “resolution”, that is how often detail points of the landmark were recorded
The accuracy of the measurement and data density has to be consistent with each other.
The
accuracy of determining a single point does not determine the accuracy of the entire data system.
Therefore, in the case of digital maps the notion of scale has to be substituted by two ideas:
- accuracy
- data density
On the other hand, digital map objects can only be displayed on a given scale. Storage of coordinates is independent of scale, while displaying is always scale-dependant. The above guide provides help to express scale.
Can people be recognised in the orthophoto-mosaic?
In an orthophoto-mosaic with a field resolution smaller than 5mm neither the face of a person looking upwards can be identified with biometric or other direct methods, nor licence plates can be read. It is easy to understand as portraits with a resolution smaller than 0.5 cm are equivalent to one of the face blurring solutions used in crime news (Bakó and Molnár, 2012).
In special cases (colour of clothing, presumed presence of the person or a known motor vehicle, etc.) and the interdependence of several factors the presence of a given person may be indirectly assumed at even lower resolutions, but such assumptions are by no means conclusive (Florian et al 2015).
In order to unambiguously identify a person, the distance between his eyes must be mapped onto at least 200 pixels while the person is looking perpendicularly up into the camera. Thus, based on the maximum 12-cm-distance between the eyes of a healthy adult, an orthophoto with a field resolution of 0.5 cm is not only insufficient for unambiguous identification, but in case the “plane” of the face is perfectly perpendicular to the axis of the camera, then the possibility of identification is also very low. A number of examinations verify the fact that a 1 cm resolution of the object plane renders the automated or manual identification by acquaintances impossible in spite of the most up-to-date correctional or filtering algorithms, and that the actual recognition begins at a field resolution of roughly 0.5 cm, where a human face is mapped at a width of more than 32 pixels.
In case the resolution of an image makes it possible to identify persons and vehicles, and the working area may be entered also by persons not having accepted the conditions of taking and publishing photographs, their unidentifiableness has to be ensured by masking, blurring or randomizing the pixels.
It would be a mistake to limit the possibility of conducting higher resolution surveys, as nowadays several environmental and economic decision-supporting and agricultural and ecological surveys can only become feasible at even higher resolutions.
References:
G.Bakó, Zs. Molnár (2012): Face recognition software and the related sensors
G. Bakó, M. Tolnai, Á. Takács (2014): Introduction and Testing of a Monitoring and Colony-Mapping Method for Waterbird Populations That Uses High-Speed and Ultra-Detailed Aerial Remote Sensing, Sensors 2014, 14,
Hillen F., Meynberg O., Höfle B. (2015): Routing in Dense Human Crowds Using Smartphone Movement Data and Optical Aerial Imagery
Hyunju M., Hyun-Cheol C., Unsang P., Seong-Whan L., Anil K., Jaina b. (2011): Face Recognition at a Distance (Face Recognition Techniques) Part 1
Xiang X., Wanquan L., Ling L. (2014): Low Resolution Face Recognition in Surveillance Systems
Zou W.W.W., Yuen, P.C.(2012): Very Low Resolution Face Recognition Problem
What about UAV regulations?
Civil drones (Unmanned aircraft) by EASA