

Precision agriculture enables the dynamic use of technologies to boost up crop production, increase crop yield and at the same time decrease the input variables. Yield monitoring and estimation are considered to be the major steps to implement site-specific crop management or precision farming. Unmanned aerial vehicle (UAV) imaging based automatic rice yield estimation in different growing stages, could be a solution. In this work, we proposed a color-based segmentation algorithm that used Lab color space and k-mean clustering techniques to detect rice grain panicles area. Segmentation of grain from the image background was carried out in two steps. In the first step, a filter was applied to RGB (red, green and blue) images to remove noise and the image converted to Lab color space. The k-means clustering was applied to organize all colors contained in both a and b layers. The pixels were clustered based on their color and special features. In the second step, the variation between colors was measured and labelling of pixels was completed by cluster index. The clustering index was joined to a specific region and the images segmented using color information. The proposed method showed that rice grain can be segmented and that rice grains panicles numbers and area can be estimated from UAV images. The comparison provided significant results. The correlation between the ground truth measure and proposed method for rice grains panicles number and area were found to be around 0.931 and 0.842 respectively.