Sandau R. Digital Airborne Camera: Introduction and Technology

Подождите немного. Документ загружается.

274 6 Data Processing and Archiving

difference in the “sensor model”, the mathematical representation of the sensor, the

basic functionality of which consists of transforming the coordinates between the

image and the three-dimensional model or ground system. Although the classical

sensor model of the aerial ﬁlm camera can be used in the case of array cam-

eras, including images composed from multi-frame systems, an expanded version is

required for line cameras, because here a raw image strip consists of a sequence of

single-line individual images with a common interior orientation, but with a varying

exterior orientation.

The principle of the workﬂow used in processing imagery from today’s

high-resolution airborne cameras is illustrated in Figs. 6-1, 6-2, 6-3 and 6-4.

Combinations of the steps shown and modiﬁed sequences can be used too. The

ﬁrst phase, which includes the processing necessary to go from the images captured

in the camera’s mass memory on board the aircraft to the raw data that enters the

photogrammetric process, consists of three steps (Fig. 6-1):

1. The process begins with downloading data f rom the camera’s disk memory and

setting up a project structure.

2. This step is imperative for line cameras and optional for array cameras: determi-

nation of a preliminary exterior orientation with GPS/IMU data from the aircraft

and data from a GPS base station. If the project requirements are less stringent,

this also provides the ﬁnal exterior orientation.

3. For cameras without real-time corrections during imaging (such as today’s

high-resolution area array cameras), it is necessary to perform DSNU and

Downloading

data

GPS

base station

Determination

of position

and attitude

Geo-

referencing

Preliminary

orientation

data

Sensor

calibration

GPS/IMU

Data

Image

data

DSNU/PRNU-

correction

Interpolation

of defective

pixels

Source data

Camera

mass

memory

Optional with

area array camera

Area array camera

Line cameras

Fig. 6-1 Processing chain for digital airborne cameras: the ﬁrst phase, from downloading the

images captured in the camera’s mass memory on board the aircraft to the raw data the enters the

photogrammetric process (described in Steps 1–3 in the text)

6 Data Processing and Archiving 275

Source data

Rectification

through

plane

projection

Preliminary

orientation

data

Imagedata

Determination

of tie points

Image

mosaic

structure

Line cameras

Single-frame

area array cameras

Pan-

sharpening

for RGB

Geometrically

pre-corrected

data

(Stereo-viewable and

matchable)

Only in special cases

Multi-frame

area array cameras

Fig. 6-2 Processing chain for digital airborne cameras: the second phase, geometric pre-

processing of the raw data (described in Steps 4 and 5 in the text)

Geometrically

pre-corrected

data

Preliminary

orientation

data

Image data

Always in the case

of line cameras

optional

Completely

pre-corrected,

oriented

image set

(Stereo-viewable

and machable)

Useful in the case

of area array cameras

Bundle

adjustment

Coordinates

of pass points

Matching of

tie points

Measurement

of pass points

Atmospheric

and BRDF

correction

Fig. 6-3 Processing chain for digital airborne cameras: the third phase, ﬁnal image pre-processing

and triangulation to obtain ﬁnal orientation data and coordinates of pass points for use in further

photogrammetric operations (described in Steps 6 and 7 in the text)

PRNU correction of image data. For array cameras only, there is usually some

interpolation of defective pixels.

4. The second phase includes geometric pre-processing of the raw data (Fig. 6-2).

This involves two further steps:

276 6 Data Processing and Archiving

Final

orientation

data

Image data

Sensor model

structure

Completely

pre-corrected,

oriented

image set

Digital

workstation

Photogrammetric and

remote sensing

products:

DSM, DTM,

orthophoto,

object extraction,

3D visualisation

...

Reflectance

Classification

Vegetation analysis

...

External

data sources

Fig. 6-4 Processing chain for digital airborne cameras: the fourth and ﬁnal phase involves pho-

togrammetric processing of the reﬁned image data and the results of triangulation to produce the

deliverables from the project (described in Step 8 in the text)

(a) In line cameras, images are corrected by projecting them on to a plane.

For multi-frame cameras, it is necessary to create the composite image as

a mosaic of the component images by locating homologous points in the

overlap zones and correcting the images accordingly.

(b) If the camera generates colour images at reduced resolution compared to the

panchromatic, a pan-sharpening process is performed on the colour images.

5. The third phase may involve some ﬁnal image pre-processing but its main pur-

pose is triangulation in order to obtain ﬁnal orientation data and coordinates of

pass points for use in further photogrammetric operations (Fig. 6-3). There are

two steps:

(a) Optional for all camera types: Radiometric correction of images with respect

to atmospheric and BRDF inﬂuences.

(b) Triangulation of a set of images comprising automatic determination and

measurement of tie points and bundle adjustment. The result is precise exte-

rior orientation of the projection centres of all images involved, optionally

supplemented with interior orientation variables (self-calibration). Special

characteristic for line cameras: since an estimation of the projection centres

of all individual lines would be underdetermined, orientation is performed

using selected ﬂight path points, the spacing of which is selected such that an

interpolation with the aid of the GPS/IMU result can be carried out between

them.

6 Data Processing and Archiving 277

6. The fourth and ﬁnal phase involves photogrammetric processing of the reﬁned

image data and the results of triangulation to produce the deliverables from the

project (Fig. 6-4):

7. Further processing steps up to the ﬁnal product are similar to those used for

scanned ﬁlm aerial photographs, provided that the post-processing software

provides a suitable sensor model.

The vertical bars at the ends of the component images are often also referred

to as processing levels or data levels (for example, data product level 0, 1, ...).

There is no standard deﬁnitions of such data products, however. The data products

of such interfaces can often be imported to other processing platforms for further

processing.

With the greater dynamic range of digital imaging systems, which manifestly

exceeds the 8 bits that can be acquired with ﬁlm, it is advisable to design the entire

processing chain for a greater bit depth. For reasons of processing speed, only 16-bit

depth per image channel is appropriate for today’s computer systems.

The situation is different with regard to end products. For reproduction in printed

form or on screen, 8 bits per pixel per channel are usually sufﬁcient and often spec-

iﬁed. Here, the data volume can be further reduced by using common compression

methods, especially JPEG. Compression ratios of 3:1 can be achieved without per-

ceptible loss of image quality or changes in geometry (Tempelmann et al., 1995),

and even substantially higher ratios are possible with colour images.

Chapter 7

Examples of Large-Scale Digital Airborne

Cameras

In the preceding chapters, we have investigated and explained all known theories

and technologies which need to be well understood in order to develop and produce

digital cameras, and to market and maintain them on a worldwide scale. The rea-

sons for the transition from analogue to digital airborne cameras are explained in

Chapter 1, especially such aspects as

• areas of application in photogrammetry and remote sensing,

• complementarity of airborne and spaceborne sensor systems,

• matrix and line sensor concepts.

All the information given in Chapter 1 and the succeeding chapters leaves open

the question of which basic concept (matrix or line sensor) and which implementa-

tion variant are to be used. This decision is to be according to the main application

areas addressed and the available technologies. This chapter gives selected exam-

ples of different implementations. As explained in Section 1.5, this book focuses its

attention on large format cameras. The three examples selected – ADS40/ADS80,

DMC and UltraCam – are representative of the variations available on the market

today. These cameras may be brieﬂy as follows:

• The ADS40/ADS80 is a multiple-line sensor, which generates continuous image

strips for all channels (pixel carpet type).

• The DMC generates high-resolution panchromatic images from four images with

slightly divergent optical axes (butterﬂy type).

• In the UltraCam, the high resolution image is composed in patchwork manner

from nine matrices acquired by four cameras (patchwork type)

All three cameras are being successfully marketed worldwide.

7.1 The ADS40 System: A Multiple-Line Sensor

for Photogrammetry and Remote Sensing

7.1.1 Introduction

The realization of the project ADS40, which is considered the 1st generation,

spanned a time period of 20 years the phases of which can be summarized in the

following manner:

279

R. Sandau (ed.), Digital Airborne Camera, DOI 10.1007/978-1-4020-8878-0_7,



Springer Science+Business Media B.V. 2010

280 7 Examples of Large-Scale Digital Airborne Cameras

• Formulating a vision

• Pre-projects and feasibility studies

• Meetings with inventors, developers, scientists and component providers

• Observing and assessing the market

• Initiating the innovation process

• Formulating the business plan and the product requirements

The development of components and prototypes took place in the period from

1997 till 2001. The digital airborne sensor ADS40 was presented to the world map-

ping community on the occasion of the XIX ISPRS Congress in Amsterdam in July

2001. This sensor was developed jointly between DLR (German Aerospace Center)

in Berlin and Leica Geosystems (previously LH Systems) in Switzerland. The air-

borne digital sensor ADS40 is part of the ﬁrst commercially available digital large

format airborne survey system. The optical performance characteristics based on

a resolution of over 130 lp/mm and a FoV (Field of view) of 64

◦

result in excel-

lent area coverage performance similar to that of the known aerial ﬁlm cameras,

which produce aerial photographs in the format 23 × 23 cm (9



×9



). This ADS40

Airborne Digital Sensor System is an innovation in the world of photogrammetry,

because it also fulﬁlls the multispectral requirements in the ﬁeld of remote sens-

ing. The narrow bands of the ﬁlters, the linear sensitivity curve of the CCD lines

and their absolute spectroradiometric calibration made the transition possible from

a photographic sensor to a complete image measuring device. The data generated

by the CCD lines is radiometrically corrected in real-time before it is stored on

the mass memory. The line-perspective imagery with uniform illumination in ﬂight

direction allows for automatic reduction of disturbing illumination and reﬂectance

effects. The tuning of geometric and radiometric resolution opens new possibilities

in the ﬁeld of research and development. Through the integration of a GNSS sen-

sor and an inertial measuring unit (IMU) into the system exact geo-referencing and

co-registration of individual multi-spectral bands is possible for the ﬁrst time.

7.1.1.1 2nd Generation ADS40

In 2007 two additional sensor heads the SH51 and SH52 were added to the ADS40

System. An innovative, patented beamsplitter called Tetrachroid with dichroitic ﬁl-

ters allows the co-registered data acquisition of high resolution imagery across all

bands at equal GSD. The ADS40 is the only large format airborne digital camera

that can make this claim. This eliminates the need for often time consuming pro-

cessing steps such as pan-sharpening and the creation of virtual images, which are

an integral part of multispectral image generation with large format digital array

cameras. Digital imagery generated with the ADS second or third generation does

therefore no longer show color fringes along edges.

7.1.1.2 3rd Generation ADS80

Technical improvements already introduced, tried and tested with the 2nd genera-

tion Sensor Head are maintained in the third generation. The thermal and pressure

7.1 The ADS40 System 281

stabilized lens, the focal plate and the IMU, which with the help of a specially

designed table has been brought into the optical axis of the sensor, are tightly inte-

grated into a quasi-monolithic block that guarantees highest (geometric) stability.

This extremely stable design allows support and exchange of various IMU types

and the application of a simpliﬁed sensor model with signiﬁcantly less local distor-

tion parameters. Survey ﬂights undertaken at two different altitudes and following a

simple pattern can be used for system calibration through bundle block adjustment.

Independent research by Passini and Jacobsen (2008) reveals that the geometrical

accuracy and the use of pushbroom data in photogrammetry is at par if not better

than that of digital frame cameras.

7.1.1.3 Advantages Which Speak for the Line Sensor Technology

The experience with line sensors in satellites which are circling the Earth and Mars

has proven that data quantities can be kept to an optimal level whilst providing

best stereo observation quality. Equally line sensor technology has proven that it

is possible to produce sensors, which can capture multiple spectral ranges with

high geometric and radiometric resolution. Furthermore the simple, exact geometry

of a line sensor combined with the exact spatial trajectory and absolute orienta-

tion thereof given by the GNSS/IMU system provides distortion-free image strips.

Geometric distortion caused by patching together surface array images from various

cameras into one large virtual image is avoided.

Other advantages of line sensor cameras are:

• Same resolution at the instant of image capture in all bands

• Complete spectral separation in all bands

• Parallel line perspective images with superior color continuity in large mosaics

• Line sensors have no defective pixels

• The Base/Height ratio is superior to that of rectangular digital frame

cameras

• No shutters are required, which in multi lens systems need accurate synchroniza-

tion and can fail

• One high-quality, temperature and pressure controlled lens with a single focal

plate maintains calibration permanently

7.1.1.4 Economic Aspects and Competitive Systems

Contrary to satellite-borne cameras the return on investment of airborne cameras

for earth observation depends largely on ﬂight management and ﬂight economy

because they are not ﬁnanced by government organizations nor do survey companies

have free access to the required airspace. For these reasons a large effort was also

necessary to develop a reliable ﬂight management and camera control system as

well a ﬂight planning and ﬂight evaluation system. To compete favorably against

the existing ﬁlm based aerial cameras development priority was also given to the

following:

282 7 Examples of Large-Scale Digital Airborne Cameras

• Area coverage performance equal to the ﬁlm-based aerial cameras

• Savings in ﬂight cost by capturing all spectral bands simultaneously with better

radiometric resolution than ﬁlm or surface array cameras

• Stereo viewable image strips with a choice of various stereo angles from the same

ﬂight

• Providing simultaneously new applications in airborne remote sensing

• Faster data processing based on the simple line sensor geometry

7.1.1.5 The Path to the Ideal Lens System for Airborne Digital Cameras

Choosing the line sensor concept for the CCD sensors provided many advantages:

• All CCD sensors can be placed on a s ingle focal plate and all spectral bands can

be captured through only one high quality lens system

• Line sensors can be used which provide a 2–3x wider swath than surface arrays

• Economical, defective-free CCD sensors could be used

• In ﬂight direction all bands produce 100% overlapping image strips

• Image strips are homogeneously illuminated due to constant line viewing angles

• Better image quality because the extreme edges of the ﬁeld of view of the optics

are not used

• It is possible to use staggered CCD lines to increase the image resolution and

decrease aliasing

For the optical designers the decision to use line sensors represented a major

opportunity, to fulﬁll different further user demands. The narrow bands in the

spectral range useful for the applications of remote sensing and the demand for a

continuous spectral response (central wavelength and range) across the entire length

of the CCD line led to the demand for a telecentric lens on the imaging side of

the optics (see Fig. 7.1-1).

For photogrammetric applications the focal length and the place of registration

of each pixel must remain stable on the focal plane, also under strongly changing

12:57:05

adc proto t1

Scale: 0.50 BRN 22-Jan-00

50.00 MM

Fig. 7.1-1 Telecentric optics DO64 with passive temperature compensation

7.1 The ADS40 System 283

conditions of temperature and pressure during the ﬂight. The ADS40 optics are

stabilized by means of a unique mechanical compensation system that adjusts for

variations in temperature.

The operating range for optics and electronics are shown in Fig. 7.1-2.

Fig. 7.1-2 Operating range for optics and electronics in the range of the atmosphere accessible to

aircraft

7.1.1.6 The Deﬁnition of the Filters

The ADS40 and ADS80 have special, narrow-band ﬁlters for the spectral regions

red, green, blue and near infrared (see Fig. 7.1-3). Thus the ADS40 sensor is a

useful tool for remote sensing beyond traditional applications for photogrammetry.

All methods and possibilities of image analysis, e.g. for automatic interpretation and

classiﬁcation of image content, can be applied to the ADS40 and ADS80 images.

The narrowness of the spectral bands and the steep ﬂanks of the ﬁlters are

ensured with interference ﬁlters. A special algorithm, developed by the University

of Illmenau, Germany permits a conversion of the RGB and Pan images into a true

color image.

A complete novelty compared to all line sensor cameras developed in former

times is the development of special beam splitters the so-called Trichroids in the ﬁrst

generation of the ADS40 and the Tetrachroids in the 2nd generation ADS40 and 3rd

generation ADS80. The location in the spectrum is realized by means of dichroitic

ﬁlters. These differ from the classical beam splitters, which divide the entire irra-

diated energy onto several detectors. Instead dichroitic ﬁlters provide a spectral

separation and allocation. The entire energy contained in the spectral channels is

led with little loss at the reﬂection surfaces to the detectors. The Trichoid produces

284 7 Examples of Large-Scale Digital Airborne Cameras

Fig. 7.1-3 Deﬁnition of the

narrow band ﬁlters of the

ADS40 and ADS80 Sensor

CCD Triplet

Lens

Trichroid

Ground pixel

reflectance

Fig. 7.1-4 Cross-section

through the Trichroid energy

beam splitter (1st Generation

ADS40)

co-registered RGB multispectral channels (see Fig. 7.1-4) and the Tetrachroid even

co-registered RGBN channels (see Fig. 7.1-5).

7.1.1.7 The Choices for the Focal Plate Module

As a CCD provider the company ATMEL (previously known as Thomson) was cho-

sen, a company that also provides very similar CCD lines for the SPOT 5 satellites