satellite image interpretation · 2019-12-07 · image interpretation strategy visual image...

SATELLITE IMAGE INTERPRETATION

Course: Introduction to RS & DIPUCC:620124

Zahid KhalilContact:

Zahidkhalil.rao@gmail.com

2

The concepts of image interpretation

Image interpretation: the process of identifying objects orconditions in images and determining their meaning orsignificance.

The interpreter’s task: use scientific tools andmethodology to arrive at objective findings.

Geographical knowledge is needed to relate the visiblecharacteristics on the image to the real-world geographicalfeatures, even though some of these features may not bephysically visible.

Types of Image Interpretation

▪ Spatial interpretation

▪ Spectral interpretation

3

Spectral vs Spatial

Spectral pattern recognition

Family of classification procedures that utilize pixel by pixel spectral information as the basis for automated land cover classification.

Spatial pattern recognition

Categorization of image pixels on the basis of their spatial relationship with pixels surrounding them.

5

The ground view

Ground view of Mt. Everest, the highest spot on earth.

7

Visualization – Orthographic View

Orthographic view of Mt. Everest. The photo was taken by

astronauts on the International Space Station in 1993.

8

Spatial interpretation

Spatial interpretation means identifyinggeographical features using spatialcharacteristics of objects shown on images.

The most important tasks for spatialinterpretation is to establish interpretation keys,i.e. identifying the typical spatial and spectralpatterns of known geographical features.

Basic elements of image interpretation

X, Y location

size

shape

shadow

Tone

colour

Texture

Pattern

Height & Depth

Site, Situation & Association

Elements of image interpretation

9

ND GIS Users Workshop Bismarck, ND October 24-26, 2005

Shape

▪ Many natural and human-made features have uniqueshapes.

▪ Often used are adjectives like linear, curvilinear, circular,elliptical, radial, square, rectangular, triangular, hexagonal,star, elongated, and amorphous.

10

11

Shape

Centre-pivot irrigation system in Morrow County, Oregon. Most of the fields are planted with wheat.

Alluvial fans at the north of Turpan Depression, Xinjiang, China.

12

Shape

Alluvial fans at the north of Turpan Depression, Xinjiang, China.

Ground view of alluvial fans at the north of Turpan Depression, Xinjiang, China.

13

Shape

Alluvial fans along the east side of Death Valley, California. Alluvial fans can be easily recognised by their fan shape and adjacency to mountain fronts.

15

Ground view of the Great Pyramids, Egypt.

Some man-made features have unique shapes.

Shape

ND GIS Users Workshop Bismarck, ND October 24-26, 2005

Shadow

▪ Shadow – usually a visual obstacle for image interpretation

▪ Gives height information about towers, tall buildings

▪ Shadow reduction is of concern in remote sensing

Hobject

SShadow

tanH

S =

16

17

Shadow

Shadow shown with low sun angle is the key to the interpretation of shape of Mt. Everest.

ND GIS Users Workshop Bismarck, ND

October 24-26, 2005

Tone and Color

▪ A band of EMR recorded by a remotesensing instrument can be displayed onan image in shades of gray ranging fromblack to white.

▪ These shades are called “tones”, and canbe qualitatively referred to as dark,light, or intermediate (humans can see40-50 tones).

▪ Tone is related to the amount of lightreflected from the scene in a specificwavelength interval (band).

18

19

Tone and colour are used to identify agricultural fields. The fields with crops or harvested are clearly separated by their tones and colours. Also note the tone difference shown on the bare fields indicating different soil moisture contents.

Tone or Color

ND GIS Users Workshop Bismarck, ND

October 24-26, 2005

Texture

▪ Characteristics placement &arrangement of repetitions of tone orcolour in an image

▪ Visual impression of roughness orsmoothness of an image region

▪ Texture refers to the arrangement oftone or color in an image.

20

21

Texture

Textures associated with forest, pasture and cropland. The colour photograph shows the strong contrast in texture between forest (dark and coarse), pasture (light and smooth and farmland (light and smooth with regular road and drainage network).

ND GIS Users Workshop Bismarck,

ND October 24-26, 2005

Pattern

▪ Pattern is the spatial arrangement of objects on thelandscape.

▪ General descriptions include random and systematic;natural and human-made.

▪ More specific descriptions include circular, oval,curvilinear, linear, radiating, rectangular, etc.

22

23

The contrast between sand dunes (left) and loess (right) landscapes found near Yulin, Shannxi, northern China. The mobile sand dunes are well recognised by their repeated patterns, while the high density of gullies of the loess landscape suggests severe soil erosion and mass movement.

Pattern

25

Loess Landscape

The high density of gullies is the key for image interpretation of the loess landscape.

ND GIS Users Workshop Bismarck, ND October 24-26, 2005

Height and Depth

▪ As discussed, shadows can often offerclues to the height of objects.

▪ In turn, relative heights can be used tointerpret objects.

▪ In a similar fashion, relative depths canoften be interpreted.

▪ Descriptions include tall, intermediate,and short; deep, intermediate, andshallow.

26

ND GIS Users Workshop Bismarck, ND October 24-26, 2005

Association

▪ This is very important when trying to interpret an object or activity.

▪ Association refers to the fact that certain features and activities are almost always related to the presence of certain other features and activities.

27

Image interpretation strategy

▪ Visual image interpretation : Process of indentifying what we see on the images and communicate the information obtained from these images to others for evaluating its significance

▪ Includes relative locations and extents

▪ Use of data products like Satellite single band imageries, FCC for performing image interpretations to extract thematic information for subsequent input to GIS

28

Spectral interpretation key built-up area

Built-up area is mainly covered by buildings, mixed with small proportion of vegetation and other cover types. The spectral characteristic is quite unique, often with ‘darkened’ concrete spectral signals, that is related to the shadow of buildings. The pattern of road network is also a helpful key

29

Spectral interpretation key – bare Land

The land under construction or, in a general term, “disturbed ground” is characterized by those where the native vegetation cover has been completely removed, showing strong reflectance in all bands.

30

Spectral interpretation key –water body

When the water is deep enough (i.e. no reflectance from the bottom of the water)and clean enough (i.e. not containing too much sediment), this cover type can be clearly identified. When they are shallow and muddy, they create significant confusion in the auto-classification processing.

31

Temporal interpretation and Land cover change

Using multi-temporal images to identify the change and movement.The temporal interpretation is based on spatial and spectralinterpretation. It adds the temporal dimension for imageinterpretation

The changing desert multispectral images from 1987– 1998, Fukang county, Xinjiang, China. The source images are Landsat TM presented in band4 (red), 3 (green),2 (blue) colour composite.

32

Image Acquisition DATE/TIME Considerations

▪ This usually affects the spectral characteristics that change with the season, or time of the day. E.g. shadow effects E.g. seasonal change of crops shows very different spectral interpretation keys.

▪ Image acquisition date/time largely influences the interpretation of the image.

33

Seasonal Appearance

Rice croplands requires the land be periodically sub-merged with water and dried in winter season. The significant cover change associated with the agricultural growing season makes this class unique in its properties and usage. On the winter image the rice crops were harvested and fields were drained, thus the paddy fields were shown as bare soils thus readily distinguishable from their surroundings.

34

Landscape features recognized

▪ Land cover

▪ Vegetation

▪ Soil

▪ Built-up areas

▪ Water bodies

▪ Glaciers and snow

▪ Geological structure

35

Land cover - vegetation

Vegetation is always the key for image interpretation (Bosten Lake, Xinjiang, China)

36

Land cover - soil

Soil properties often have to be interpreted using indirect information, such as vegetation and landforms. The image shows the marginal area between irrigated farmlands (deep sandy-loamy soil) and stony soil on alluvial fans near Yinchuan City of Ningxia, China

37

Land cover – built-up areas

Built-up areas of Beijing City. The built-up areas area can be easily identified by its road network, concrete surface materials and clustered spatial distribution.

38

Land cover – water bodies

Water bodies (river sand fishponds) are significant features in the Pearl River Delta (PRD). They are clearly distinguished in the Near Infrared band(Band 4) of the Landsat TM image

39

Glaciers and snow

Glaciers around Mt. Everest are highly visible by their contrast in colour and shapes to the surrounding snow and vegetated covers

40

Image Information Extraction

▪ Image arithmetic

▪ Indices

▪ Image classification

▪ Supervised classification

▪ Unsupervised classification

41

Questions & Discussion