Innovation on Geolocation and Pattern Recognition for Paddy Growth Stages Reporting in Indonesia

3rd International Conference of Indonesia Society for Remote Sensing

地球科学;计算机科学

Agustan^1 ; Yulianto, Swasetyo^1 ; Sumargana, Lena^1 ; Sadmono, Heri^1 ; Alhasanah, Fauziah^1

Center of Technology for Regional Resources, Agency for the Assessment and Application of Technology (PTPSW - BPP Teknologi), Jakarta, Indonesia^1

关键词: Automatic pattern recognition;    Location based;    Pre-processing;    Reporting systems;    Sampling method;    Statistical approach;    Stratified random sampling;    Temporal resolution;   

Others  :  https://iopscience.iop.org/article/10.1088/1755-1315/165/1/012001/pdf
DOI  :  10.1088/1755-1315/165/1/012001

学科分类：计算机科学（综合）

来源: IOP

【 摘 要 】

Information on paddy growth stages is important for rice yield prediction. By utilizing statistical approach such as Area Frame Sampling method, data on paddy growth stages in the certain area may be used to estimate its harvest potential. The method of area frame sampling is based on segment observations determined from a stratified random sampling. Normally, information of paddy growth stages is obtained from terrestrial and remote sensing method. Since the life cycle of paddy is around 120 days, temporal resolution for observation becomes the main consideration. There is a need to develop a robust location-based reporting system for paddy growth stage. This system is designed by controlling the observer to report paddy condition on a determined location, called segment; the observer must report from the center of each segment. This paper discusses the innovation on the use of mobile phone for geolocation and pattern recognition to collect paddy growth stage data. The GPS on the mobile phone is explored for geolocation whereas the camera on the mobile phone is utilized to capture the paddy images. This information is then sent to the server for automatic pattern recognition. The statistical method with pre-processing, feature extraction, classification, feature selection and learning were applied on pattern recognition. Testing on geolocation was conducted in the Java area since May 2017 and installed for 2,356 observers. It is found that 73% of observers successfully reported the paddy growth stage from the center of each segment, which is locked in 10-meter accuracy, and 27% reported not from the center of segment due to field conditions.It is also found that a combination of GPS and mobile network or assisted-GPS can speed up the positioning.