Rice Crop Phenology Using Texture Analysis on Time-series Images Obtained from Still Camera

In this study, an algorithm is proposed to determine the duration in a rice crop cycle based on texture analysis. During an observation period in 2013, daily images were acquired from a still camera installed at a paddy field. Given a set of time-series images, the texture analysis is used to classify different stages of the rice growing. Regarding the hypothesis, the rice crop cycle can be separated into three paddy stages; starting (coarse texture), midpoint (fine texture) and ending (no texture). The texture analysis is based on Gray Level Co-occurrence Matrix (GLCM) with multi-distances. The proposed diagram can be described as follows. Initially, the paddy region is selected as an area of interest (AOI). The selected AOI is computed for a vegetation index, Excessive Green (ExG), and enhanced by using histogram equalization method. Then, the enhanced image is used for generating co-occurrence matrix in order to describe the texture feature. Statistical property (contrast) is analyzed and measured on the co-occurrence matrix for paddy stage classification. The results show that our proposed diagram can be efficiently used to determine the duration of paddy stages. To obtain more efficient results, our perspective work will consider other features such as color, edge, shape, etc.