Home Articles Banding and line drop errors detection in remotely sensed satellite data through...

Banding and line drop errors detection in remotely sensed satellite data through transition count technique

A. Ahmad*, M. D. Joshi**
Department of Electrical & Electronics Engineering College of Engineering, Sultan Qaboos University
P. O. Box 33, Postal code 123; Muscat, Sultanate of Oman
Tel. 00968-515327 (Off.) / 00968-514109 (Res.) Fax. 00968-513454 / 00968-513416
[email protected]

Introduction:
Satellite remote sensing is a very useful preposition for the purpose of evaluation of and management of natural earth resources. It also helps in monitoring the phenomenon of the earth surface. Since the sensed satellite, data is multi-spectral and multi-temporal in nature thus capable of providing repetitive cyclic information, which helps in determining the effects of various phenomenon on the earth’s environment. Therefore, the satellite sensing is becoming popular all over the world [1]. Every country either has established a remote sensing center or in the process of establishing of such facility.

In general, the remote sensing phenomenon depends upon sensing of the objects in the energy bands. There are number of energy bands in the energy spectrum as illustrated in the Figure 1 [2]. The energy wave radiation incident upon earth is either reflected or emitted by the earth surface. The amount of such energy depends upon internal structure and the physical properties of the surface of the targeted objects. The energy waves so reflected/emitted travel through the atmosphere and finally captured by the sensor as shown in Figure 2 [3] – [7].

The sensors are usually placed either on ground, or on an aircraft or on a satellite. However, among this satellite sensors are most popular. Remote sensing satellites are mostly function in a polar orbit at a height varying between 500 km to 900 km.

If we take the example of the pioneer satellite Landsat 1 covers about 185 km on the ground in a single scan line; and the scan line is divided in 3240 pixels [8]. Each pixel of a data is stored in eight bits thus giving 256 possible discrete levels. For the convenience of the data analysis, the data is divided in scenes where each scene consists of 2340 scan lines.

In Landsat 1 satellite, there are four energy bands, three of them are visible and one works in Infrared band region. The information is collected through six scanners individually provided for each band. Thus, there are 24 scanners are being used for collecting information data (see Figure 3).

The energy levels obtained by the scanners through the sensor are collected and then converted into electrical pulses according to the strength of the signals. Further, these electrical pulses are converted into digital forms (Digital Number or DN values). Since the data received on board satellite is enormous, due to the limited memory provided in the satellite, only a fraction of data is stored there and the rest is transmitted to the earth station for storage.

Generally, the data is stored on high-density digital tapes (HDDT) because of its huge data storing capacity. However, this data format is not compatible to the user computer (spread over to the different parts of the world). Hence, the same data is formatted on either computer compatible tapes (CCTs) or CD-ROMs.

Problem Definition:
The stored data some times incorporated with different kinds of errors. Some of these errors can be rectified whereas others cannot be tackled. Moreover, consequent upon this, there is a loss of data. When the loss of data is beyond a certain limit, then the satellite data (in scene) becomes useless for the user. Further more, if the information of the loss of the data is not known to the user at the initial stage, the utilization of the data frustrates the user from the point of view of waste of time, money and efforts at the later stage. Thus, it is essential for the supplier to identify and rectify the errors before it is being supplied to the user.

Either malfunctioning of the detector or sudden termination of the function of data collection unit in a particular energy band causes the line drop errors and banding errors. Whenever such defective data is stored, the DN values remain unchanged through out the scan line. When such data is displayed on the monitor, it gives blank lines or line of same Grey level values.

In this communication, the authors present an online scheme of identifying the line drop and banding errors at the source level. The proposed identification scheme is based on Transition Count technique [9].

Table 1: The DN values of the data frame

Detection of Errors and Transition Count Technique:
Landsat -1,2 and 3 satellites frame covers an area of 185km. x 185km. There are 2340 scan lines, each of which has 3240 pixels contained in it. Thus, a single pixel covers about 79m x 79m on ground, which corresponds to one byte of information. In the sensing system, there are six detectors in each of four energy bands. In addition, in an individual band a strip of earth of about 480m widths is viewed in one swing of scanning mirror.


Figure 1: Electromagnetic Spectrum

When one scene is displayed on a computer monitor, all the scan lines are displayed on the screen in a highly compact form giving an impression of a film negative captured from the space. However, because of defective data some times blank lines appears in between or some times in place of normal picture, number of bands are observed.

In fact, the pixel explained above is nothing but the Grey level values of the smallest picture element in the satellite data frame. Since data is recorded in eight bits (1byte), there could be 256 possibilities of data values. These data values (Digital Number, DN) are correlated with Grey level values divided in 256 gradation levels from complete white to complete black.


Figure 2: Electromagnetic Remote Sensing of Earth Resource

Considering a single line in Landsat -1 picture frame, where 3240 pixels cover 185Km. Each pixel value is representative energy level of an area of about 79m x 79m. In view of earth surface having varied topography, the energy levels received from different parts are in different quantum and hence the pixel values along a scan line are bound to be different except on uniform surface, which is an imaginary condition. Therefore, a line containing similar pixel values does not exist in nature; this is an indicator of the error.

Since satellite data is recorded in DN values thus, by comparing these values one can find information if there is any such blank line (line drop) or shaded line (banding). Technique of TC, coupled with EXOR-ing (if required) each pixel of a scan line of satellite scene with the initial pixel can provide such information.

The line drop will exist if there is TC = zero, the DN number exists at the extreme of the considered pixel levels. However, in the case of banding error the TC may have any value in between 0 to 7. Which will need that the DN number pertaining to each pixel of a scan line (3240) is compared with the DN value of first pixel. If uniform DN numbers are obtained, that could be a banding error. Transition Count (TC) technique is one of the data compression techniques, which is being used for the purpose of the detection of the fault in digital data transmission [9]. Transition count technique can be easily implemented in hardware using a few logical components of delay elements and EX-OR gates. In the TC technique the signature is the number of the counts of the transitions of 0-to-1 and 1-to-0 in the binary data stream. Thus, TC associated with a binary data sequence D = [d1, d2,….,dn-1, dn] is

TC (D) =

(1) Where å denotes ordinary arithmetic summation and Å is modulo 2 addition. Since 0£TC (D)£n-1, the response-compression circuitry consists of a transition detector and counter with [log2 n] stages. Figure 4 illustrates this concept. Let

D0 = [0 0 0 0 0 0 0 0]
D1 = [1 1 1 1 1 1 1 1]
D2 = [1 0 1 0 1 0 1 0]

Then, TC (D0) = TC (D1) = 0 and TC (D2) = 7 are respectively minimum and maximum levels of transition counts for the above given binary data stream of length 8.

Table 1 demonstrates the use of the Transition Count technique as when it is applied to a partial data frame. The selected data frame is of size 10×10 pixels where as the DN value of each pixel is represented in base 10 forms. The DN values of all the pixels of row numbers 3 and 10 of the frame (as highlighted) have minimum Transition Count for all the pixels of the row i.e. zero and thereby, declaring the line drop errors. The row number 7 has same DN values of the pixels i.e. (00011100) 2, giving the Transition Count of two for each of the pixel of the row and thus, indicates the banding error.


Figure 3: Major Components of Landsat 1-5 Multispectral Scanner

Conclusion:
The above study has found that the use of Transition Count technique can be utilized as a powerful tool for detecting line drop and banding errors in the satellite sensed data. Thus, the demonstrated technique can reduce the level of probability of errors in the data provided by the supplying agencies from its archives. Thereby, if the data is scanned using the above technique, the above-mentioned errors in the data are at once detected at the source level itself. Hence, the time, funds and efforts of the users community, which is spread over, to distant places can be saved.

References:

  1. Sivakumar, R., (1998), ‘An Integrated Multi-Source Digital Topographic database for India,’ Ph. D. Thesis, IIT Delhi (India).
  2. Curran, P. J., (1985), ‘Principles of Remote Sensing,’ (Longman Scientific and Technical), UK. p. 10.
  3. Janza, F. J., (1975), ‘Interaction Mechanism: In Reeves,’ R.G. (ed.) Manual of Remote Sensing. American Society of Photogrammetry falls Church, Virginia, and pp. 75-179.
  4. Suits, G. H., (1975), ‘The nature of electromagnetic radiation: In Reeves,’ R.G. (ed.) Manual of Remote Sensing. American Society of Phototgrammetry, Falls Church, Virginia, pp. 51 -73.
  5. Sensing (2nd Ed.). American Society of Suits, G. H., (1983), ‘The nature of electromagnetic radiation: In Colwell,’ R.N. (ed.) Manual of Remote Photogrammetry, Falls church, Virginia, pp. 37-60.
  6. Lillesand T.M. and Keifer R.W., (1979), ‘Remote Sensing and Image Interpretation,’ (Wiley), New York.
  7. Jensen, J. R., (1986), ‘Introductory Digital Image Processing, A remote sensing perspective,’ Prentice Hall, New Jersey. p. 29.
  8. NOAA, (1975), Data Users Notes. NOAA Customers Services, Mundt, Federal Building, Sioux Falls, S.D.
  9. Hayes, J.P., (1976), ‘Transition Count Testing of Combinational Logic Circuits’, IEEE Transactions Of Computers, vol. C-25, no. 6, pp. 613-6.