Digital Image Processing-Introduction to Image Classification

Introduction

Classification is a simple process to categorize or sorting of data based on defined categories or criteria. Such as students in class can be categorized based on marks obtained in maths- number of student below 40%, number of students between 40-60%, number of students between 60-80% and number of students having more than 80%. So based on this criteria student can be categorized or sort in to four different class/category. This is a very simple and basic approach of classification to understand behavior of data in quantitative and qualitative both ways. 

This classification approach is used to classify satellite image. in the satellite image classification process, images are classified based on categorization and classification of pixel values present in satellite image. The pixel value present in each pixel is also called DN values (Digital Number), which is representation of reflectance values received by sensor of earth surface feature. It is already explained in previous article about the reflectance and change in its behavior with earth surface feature and incident energy

This homogeneous DN values characteristic of pixel is used for classification of remotely sensed images. By this approach pixels are assigned to different levels and each level is denoted as ‘class’. Classification process divides or categorize features into several classes based on users need and decision process for classification. 
This classification process requires expertise on elements of image interpretation or interpretation keys to identify and categorized homogeneous pixels, which represents the earth surface features. This quality of classification depends upon spectral information or in other words spectral bands present in the satellite image. 

Process of Classification

  • Classification process starts with the identification of number of classes and classification scheme such as urban, agriculture, forest, river etc. in which all the pixel will be classified. 
  • In the next stage field conduction/knowledge as well as secondary data need to acquire for classification 
  • Satellite image pre-processing, it includes geometric correction and radiometric correction as well as atmospheric corrections can be performed as per requirement 
  • Delineation and extraction of the study area using image processing tool from satellite image for the classification 
  • Image can be enhanced using various enhancement tools from basic enhancement process such as brightness and contrast adjustment to advance tools for example, linear stretching and gamma stretching etc. 
  • For image classification in image processing, usually two methods are deployed Supervised and unsupervised 
  • Supervised method-it is performed using signature collection from satellite image and training sites generation for classification 
  • Unsupervised Method- Grouping of pixels based on repetitive iterations as well as incorporation of statistical and mathematical approach 
  • After classification of satellite image post processing performed to improve quality of classification, such as recoding, reclass, sieve and filtering etc. 
  • In the last accuracy assessed for classified satellite image using accuracy assessment tool, this process performed to assess the quality of satellite image to accept the classified images.
Figure: Image Classification Process & Steps

With the above process satellite images can be classified based on adopted classification scheme and image processing software.

Comments

Popular posts from this blog

Digital Image Processing Tutorial-Image Enhancement Part-I

Introduction of GIS

About GIS OUTLOOK Webpage