Image Processing using Artificial Neural Network
Image processing using Artificial Neural Network applies for Geo Technics, civil engineering, mechanics, industrial surveillance, defense department, automatics and transport fields. Image pre processing, date reduction, segmentation and recognition are the process of managing images with ANN.
An image can be represented as a matrix, each element of the matrix containing color information for a pixel. The matrix is used as input data into the neuronal network.
The small dimensions of the images, to easily and quickly help learning, establish the size of the vector and the number of input vectors.The transfer function used is a sinusoidal function.
Artificial Neural Network
The image is a function defined on a spatial domain, it has a limited scale of numeric values (natural numbers, real numbers or complex numbers), values which can be used to form a matrix.
The matrix is the type of data depending on that certain images are divided into images of intensity scale and indexed (each component being a unique number, a scalar).
Scalar image intensity is defined as an image where each pixel value is set to have a measure of luminous intensity. Scalar indexed image is an image in which the value of a pixel is an index where information can be associated with the color of the pixel in question.
An image can be represented as a matrix Mm × n, each element of the array containing information of color for a pixel.
Each color can be represented as a combination of three basic colors red, green and blue. The array is used as input to the neural networks that are aimed at identifying images or grading.
Every neuron input unit indicates color information in the image, and each output neuron links to an image. All images will be scaled to the same size (width and height) and small to be easy and quick to learn. On the sizes of the images shall be determined on the size of the input vector and the number of neurons. The transfer function for this type of problem is called sigmoid function. The rate of learning has values in the range [0.1] and the error it is recommended to have less than 0.1.
Different processes for Processing of images is as follows:
1. Image pre processing is an operation which shows a picture with the same dimensions as the original image.The objective of images pre processing with ANN consists in improving,restoring or rebuilding images.
2. Data reduction or feature extraction process is of extracting number of features smaller than that of number of pixels from input window.
3. Segmentation is a division of an image into regions.
4. Recognition is one of the process which determines objects in an image and their classification.
The various domains in which Image Processing is used are :
- Industrial inspection (quality control) in order to detect the defective products in the production of steel, textiles, fruits, vegetables, plants and food.
- Medicine for detection of tumors and the establishment of a medical diagnosis.
- Service of documents, namely automatic processing of forms, sorting emails, the possibility of learning a handwritten text, etc.
- Identification and authentication for registration number recognition,fingerprint analysis in order to identify persons.
- Optimization problems
- Geo technical engineering, in order to classify the hazardous areas with possible landslides, to determine the characteristics of the soils.
- Civil engineering, for the study of the rubberized concrete homogeneity, to identify the fissures/cracks in different structures.
For a further information on Image Processing using artificial neural network please contact us.