This letter first demonstrates the color features of ambers and then a method to estimate their colors. The RGB (Red, Green, and Blue) color space and the HSV (Hue, Saturation, and Value) color space are used for data analysis. Four kinds of ambers: three of different colors (yellow, brown, and black) and an amber with an outer cover (AOC) created by solid resin or amber dirt are examined. The proposed method consists of masking, detection of AOC, and an estimation of amber colors. Experimental results for 185 images show that the proposed method can estimate amber colors with an accuracy of 90.3%. It is also shown that the yellow color is perfectly distinguished, and that the estimation accuracy of the three colors is 98.0%.

No abstract

No abstract

This study aims to develop a method that can detect the areas of Japanese texts that are difficult to read on a PC screen by analyzing eye movements. An effective method for detecting the areas of text that are difficult to read is to create a heatmap and consider the areas where the eye stays for longer than a fixation duration threshold (threshold). Usually, the threshold is determined through a 50 ms step search. However, since the eye stay time (fixation duration) for difficult-to-read text is longer than the fixation duration when reading smoothly, it is expected that steps that last longer than 50 ms are more effective at detecting the threshold. Furthermore, there are individual differences in the speed at which we comprehend text. In this study, we proposed a method that considers both the efficient search and individual differences among each subject. The experimental results showed that the threshold determined by the proposed method is more useful than the thresholds of 1000 and 800 ms that were used in a previous study, as well as the usual fixation duration of 250 ms.

In order to manage river bank, we propose a method to classify vegetation in river band by Support Vector Machine (SVM) on the basis of texture information. The method comprises three steps: detection of vegetation, initial classification by SVM, and reclassification to reduce noises. The results for 38 images obtained by the Akita Office of River and National Highway show that the proposed method using both texture information and RGB components can accurately classify grasses and harmful vegetation in comparison with a case using the RGB original components.