Wyniki wyszukiwania - BazTech

1

Tea Category Identification using Computer Vision and Generalized Eigenvalue Proximal SVM

Wang S., Phillips P., Liu A., Du S.

Fundamenta Informaticae

|

2017

|

Vol. 151, nr 1/4

325--339

EN

(Objective) In order to increase classification accuracy of tea-category identification (TCI) system, this paper proposed a novel approach. (Method) The proposed methods first extracted 64 color histogram to obtain color information, and 16 wavelet packet entropy to obtain the texture information. With the aim of reducing the 80 features, principal component analysis was harnessed. The reduced features were used as input to generalized eigenvalue proximal support vector machine (GEPSVM). Winner-takes-all (WTA) was used to handle the multiclass problem. Two kernels were tested, linear kernel and Radial basis function (RBF) kernel. Ten repetitions of 10-fold stratified cross validation technique were used to estimate the out-of-sample errors. We named our method as GEPSVM + RBF + WTA and GEPSVM + WTA. (Result) The results showed that PCA reduced the 80 features to merely five with explaining 99.90% of total variance. The recall rate of GEPSVM + RBF + WTA achieved the highest overall recall rate of 97.9%. (Conclusion) This was higher than the result of GEPSVM + WTA and other five state-of-the-art algorithms: back propagation neural network, RBF support vector machine, genetic neural-network, linear discriminant analysis, and fitness-scaling chaotic artificial bee colony artificial neural network.

2

Vision based line tracking and landmark recognition

Chen R., Sugisaka M.

Systems Science

|

2008

|

Vol. 34, no 3

39-43

EN

The purpose of this research is to find an easy and effective method for the vision-based mobile robot navigation in indoor environment. In our case, the line is the floor-line with especial colour, and the landmarks are some shapes with especial colours. The color models of the floor-line and landmarks are built, in which floor-line and landmarks color samples were extracted from sample images of floor-line and landmarks taken under different lighting conditions to make color models more robust. The robot tracks the floor-line by the peak of the floor-line colour probability distributions to move, and the robot's action policy is decided when the landmarks have been seen and recognized; the landmark is recognized by its colour and shape.

3

An Image Retrieval Method Based on Color-Complexity and Spatial-Histogram Features

Wu H-C., Chang C-C.

Fundamenta Informaticae

|

2007

|

Vol. 76, nr 4

481-493

EN

This paper proposes two kinds of image features. One feature is the spatial-histogram feature. It combines the color histogram feature and the information about the dimensional position of pixels in an image to record the distribution of the pixels' colors that are present in different spatial positions within an image. The other image feature proposed in this paper is the color-complexity feature, which can be used to describe the change of pixel colors in the image. From the experimental results, ANMRR value is provided and we observe that the image retrieval system based on these two kinds of image features can provide a fairly good accuracy rate in image retrieval. Moreover, it has the capacity to tolerate errors; that is, for images that are damaged by rotation, shift, or color variant attacks, their similar image pairs can still be retrieved from the image database. Thus, the accuracy and flexibility of the image retrieval system are drastically improved.

4

An Image Retrieval System Based on the Color, Areas, and Perimeters of Objects

Wang C-L., Hwang R-H., Chan Y-K, Chen C-Y., Cheng C-C.

Fundamenta Informaticae

|

2006

|

Vol. 69, nr 3

319-330

EN

Two different objects can be generally distinguished by their colors, areas, and perimeters. This paper hence proposes an image retrieval system which uses the colors, areas, and perimeters of objects in an image as the features of the image. The system is insensitive to the shift and rotation variations of objects in images, as well as to the scale and noise variations of images. The experimental results show that this system is capable of recognizing different images very well.