Search

Image analysis using the Artificial Intelligence (AI) is now in trend. There are a lot of data in the image format. Using image analysis many useful information can be extracted. In this period where many people are having smart phones, they are clicking a lot of images. Sometimes they have forgot about the images. But when they want to access the image they have clicked few days before, then they have to search them manually. Search image manually is a very tedious task. Second there are different types of disease diagnostic tools produce medical images. In these images detecting a disease is required expert knowledge. And many times experts also can do a mistake. But in the artificial intelligence these problem could be rectified. Using Deep-Convolutional Neural Network(DCNN) of AI, enables the world of image analysis for us.

for the validation aspect the estimation of accuracy of classifiers is significant factor of analyzing the result of classifiers. If the pixel found in single class. So we prefer the hard classification. But satellite data does not find in single class. The pixel of satellite data belongs more than one class. So in that case we prefer the soft classification. The classification is only one parameter and other parameter is assessment of accuracy and we know without assessment of accuracy classification of classifiers is incomplete. There are several method to find out the the accuracy of classifiers like MLC, FERM, RMSE, and Entropy. Accuracy results are generated for soft classified data set from AWiFS, LISS-III and LISS-IV sensor. The output results from tool used to investigation effect of sampling on assessed accuracy

Remote sensing is a popular and fast way to acquire information on geographic surface. The multispectral images are main source of information on geo-surface. Classification is used to obtain this data and constructing maps. In general, the classification is crisp and assigns a single class to each image elements or pixel, known as hard classification. However, there is another consequence that some pixels may not purely belongs to single class. This problem treated as mixed pixel problem. Mixed pixels are big issue in term of hard or static classification. Due to hard classification the significant amount of result is lost. but we can bring under control this problem to using soft classification. Classification is not completed without assessment of accuracy. Accuracy assessment of classification play vital role in term of classification and it tells amount of accuracy for the classification process. Accuracy assessment is an important issue for validation of image classification. There are several tools that provide accuracy analysis using confusion matrix, Kappa coefficients. This assessment is based on the rule of single class per pixel known as hard classification. Further for Mixed pixels in classification handled by soft classification i.e. assigning membership of each class to that mixed pixel and accuracy assessment is also differ from hard classification as each pixel have membership proportion for every class which involved in classification schema. Further assessing accuracy also depends upon how samples have been taken from census of classified output. This paper study is to focus on comparative study in soft classification approaches and accuracy assessment with reference data and without reference data pixels.

In this dissertation work, router has been designed for providing high throughput by overcoming disadvantages of switch. The router provides high speed dynamic arbitration compared to traditional switches and makes it fair at the cost of extra energy consumption. This work mainly includes the performance evaluation of mesh based interconnection networks under self-similar traffic condition. Network on Chip (NoC) is emerging as an efficient communication backbone for next generation Multi-Processor System on Chip (MPSoC). The NoC architecture is data packet based communication network on a single chip. The NoC architecture is m*n mesh of switches & resources are placed on the slots formed by switches. Recourses can be processor core a DSP core, an FPGA block, memory block of any kind such as RAM, ROM or any other intellectual property block. The resources are connected by the Interconnect IPs. The main component of IIP is network interface and the switch. Each switch is connected to the one recourses and our neighboring to the one resources and each resource is concept of region which occupies an area of any number o resources and switches.