Photo Interpretation and Remote Sensing

After reading this article you will learn about:- 1. Analogue and Digital Sensors and Systems 2. Analogue and Digital Methods 3. Digital Image Analysis vs. Visual Interpretation 4. Advantages and Disadvantages of Analogue and Digital Methods 5. Pre-Processing of Digital Remotely Sensed Images.

Photo interpretation and Remote Sensing Methodology can be defined as:

The dialectic and interdisciplinary integration of personal experience, reasoning, specific scientific knowledge and expertise and ground truth, in order to investigate, qualitatively and quantitatively, the objects, facts characteristics, phenomena and patterns of the natural and socio economic reality, as well as of their multidimensional relations, interactions and interdependencies, and their change trends in time, by using their remotely sensed imagery.

Analogue and Digital Sensors and Systems:

Artificial system of acquisition of analogue and digital remote sensing imagery based on the recording of the reflected electromagnetic radiation.

Artificial system of remote sensing on the basis of the recording of the emitted (thermal) electromagnetic radiation.

Active system of acquisition of remote sensing data on the basis of the transmittance of artificial EMR and the recording of the radiation backscattered from the earth.

Analogue and Digital Methods:

When processing and analysing remotely sensed images, manual and digital methods are usually combined. In manual image processing and analysis, most of the fundamental photo-recognition elements of visual interpretation, as well as their appropriate combinations, are used. Manual interpretation is often limited to analysing only a single channel of data or a single image or stereoscopic model at a time.

A human interpreter can only detect and evaluate noticeable differences in the imagery. Furthermore, he cannot carry out repeatable interpretation work. Digital image processing and analysis requires a computer system, with the appropriate hardware and software to process the remotely sensed data, which is recorded in digital format.

Digital Image Analysis vs. Visual Interpretation:

(a) Manual (visual) approach to image interpretation:

Remote sensing data can be basically seen as wavelength intensity information, which needs to be decoded before the message can be fully understood. This decoding process is analogous to the interpretation of the remotely sensed imagery, which relies on our knowledge of the properties of electromagnetic radiation.

In order to extract meaningful information out of these data the image interpreter has to exercise his judgment, his scientific knowledge, his general knowledge of the phenomena as well as his experience, so that he will be able to make truthful assumptions about the object/feature under investigation.

The first stage of image interpretation is known as detection. The detection stage is naturally followed by the recognition and identification stage in which the image interpreter has to exercise general, local, as well as specific levels of reference to allocate objects into known categories.

The general level is the interpreter’s knowledge of the phenomena and processes to be interpreted, the local level is the interpreter’s intimacy with his own local environment, and the specific level is the interpreter’s deeper understanding of the processes and phenomena that he wants to interpret.

In recognition and identification, the non-geometric image characteristics of tone or colour, texture, pattern, shape, shadow, size and location normally give clues. The result of identification is a list of objects and features in the area. These form the basis of delineation of areas having homogeneous patterns and characteristics.

This is the analysis stage. Each delineated area has to be classified through a process of induction (general inference from particular cases) and deduction (particular inference from general observations). Accuracy is then controlled by field checks. The final stage of the interpretation is classification, producing spatial data which can be displayed as maps, or incorporated into a Geographic Information System.

(b) Computer-Assisted Approach:

The manual approach suffers from its inability to deal quickly with a large quantity of image data. The development of computers led to the invention of “digital” methods of image interpretation. Image interpretation is basically classificatory process identification and recognition can be treated in mathematical terms, provided that image data in digital form are available.

However, the computer cannot replace the knowledge, experience, intelligence or understanding of the human-image interpreter. Various techniques such as Artificial Intelligence and image Understanding, developed by scientists in order to improve the results of computer assisted image interpretation are being studied.

The computer—assisted approach is usually based on:

i. Statistical and syntactical pattern recognition,

ii. The decision theoretic approach, and

iii. Symbolic reasoning.

Therefore, visual interpretation of digital imagery provided by remote sensing platforms does not allow full exploitation of the data provided. A human can only visually interpret 3 layers of remotely sensed information at a time.

Further, our visual acuity does not allow us to identify all spectral differences in imagery. Machine processing of imagery allows for the quantitative analysis of all spectral bands in imagery simultaneously, and is able to detect subtle differences that we cannot.

Advantages and Disadvantages of Analogue and Digital Methods:

(a) Analogue Methods:

1. Characteristics and Advantages:

i. Traditional: intuitive.

ii. Simple, inexpensive equipment.

iii. Uses brightness and spatial content of the image.

iv. Usually single channel data or three channels at most

v. Subjective, concrete, qualitative.

2. Disadvantages:

The human interpreter understands much easier remote sensing data from the visible part of the spectrum and can observe and analyse only one image at a time.

Therefore, the manual approach of image, interpretation is recommended:

i. When the area under investigation (study) is rather small.

ii. When the recognition and identification of an object/earth feature is necessary for the study of a specific problem/phenomenon.

iii. When the “spectral signatures” of the object/earth feature under investigation are rather confusing and cannot be easily distinguished.

iv. When high spectral or spatial resolution is required in order to be able to distinguish the characteristics needed in our study.

(b) Digital Methods:

i. Recent: require specialised training

ii. Complex, expensive equipment.

iii. Rely chiefly upon brightness and spectral content, limited spatial.

iv. Frequent use of data from several channels.

v. Objective, abstract, quantitative.

1. Advantages:

i. Cost-effective for large geographic areas

ii. Cost-effective for repetitive interpretations

iii. Cost-effective for standard image formats

iv. Consistent results and simultaneous interpretations of several channels

v. Complex interpretation algorithms possible

vi. Speed may be an advantage

vii. Explore alternatives

viii. Compatible with other digital data

2. Disadvantages:

i. Expensive for small areas

ii. Expensive for one-time interpretations

iii. Start-up costs may be high

iv. Requires elaborate, single-purpose equipment

v. Accuracy may be difficult to evaluate

vi. Requires standard image formats

vii. Data may be expensive, or not available

viii. Pre-processing may be required

ix. May require large support staff.

Therefore, the computer-assisted approach is recommended:

i. When the area of concern is large.

ii. When we are not interested in the detailed detection of objects/earth features.

iii. When the area of concern is homogeneous and therefore, the “spectral signatures” of object/earth features are easily distinguished.

iv. When high spectral and spatial resolution are not particularly important for the results of the study.

Pre-Processing of Digital Remotely Sensed Images:

During the digital remotely sensed images’ acquisition, some “internal” or “external” errors degrade their quality and consequently the accuracy, the completeness and the reliability of the photo-interpretation analysis that follows. “Internal” errors are due to the remote sensots/systems, are systematic and may be determined by in-flight/before-flight calibration measurements.

“External errors are due to many different platform perturbations that influence the modulation of remotely sensed image characteristics. These non-systematic errors can be determined by relating points on the ground (known by their geodetic co-ordinates), which are mathematically correlated, with their well-defined positions on the remotely sensed images.

The basic errors during the remotely sensed images’ acquisition that degrade the images’ quality and set certain restrictions to the possibilities of the Photo-interpretation-Remote Sensing Methodology are radiometric and geometric ones.

So, before the photo-interpretation analysis and, more generally, before the remotely sensed images’ exploitation, either only radiometric corrections, or only geometric ones or, and most commonly, both radiometric and geometric corrections are required.

(a) Radiometric Corrections:

The most common radiometric corrections of remotely sensed images acquired by multi-spectral scanners are:

i. The restoration of the 6th line dropout (“loss of data”),

ii. The restoration of the 6th line caused by the abnormality of the signal (“striping”)

iii. The restoration of the horizontal lines parts displacement,

iv. The restoration of the atmospheric diffusion and absorption results, which can eliminate the possibility of data acquisition from the explored region (for some areas of the electromagnetic spectrum). It has to be noted that the images acquired with wavelength λ> 0.7 mm (infra-red), are practically clear of the atmospheric diffusion effects.

(b) Geometric Corrections:

During the acquisition of digital multi-spectral remotely sensed images, some systematic and non-systematic errors influence the images’ geometrical quality.

For systematic errors, relevant geometric corrections could be performed depending on:

i. The movement’s and the orbit’s elements of the carried platform or the remote system,

ii. The element concerning the basic characteristics, the calibration measurements and the distortions of this particular remote system.

For non-systematic errors, such as:

i. Errors resulting from the accidental altitude divergence of the remote system’s orbit (these errors cause a change of the remotely sensed images scale), and

ii. Errors resulting from accidental divergences of the remote system axis, from the specified placed of reference, is required the knowledge of the geodetic co-ordinates of some relating points on the ground, which simultaneously, are well-defined on the remotely sensed images and consequently their image co-ordinates can be measured in pixels lines and columns.