Rice panicle length measuring system based on dual-camera imaging

Abstract

The length of the rice panicle determines the number of grains it can hold, and consequently rice yield; it is therefore one of the most important traits assessed in yield-related research. However, the conventional method of measuring panicle length is still a manual process that is inconsistent, subjective and slow. In this study, a novel prototype, dubbed “Smart-PL”, was developed for the automatic measurement of rice panicle length based on dual-camera imaging. Cameras with a long-focus lens and a short-focus lens were utilized to capture both a detailed image and a complete image of the rice panicle, respectively. Specific image processing algorithms were exploited, to analyze the neck image for neck identification and the whole-panicle image for path extraction. Subsequently, co-registration was used to identify the neck location in the whole-panicle image, and a resampling method was used to search for the path points between the panicle neck and the tip. Finally, the panicle length was calculated as the sum of the distances between each adjacent path point. To evaluate the accuracy of this prototype, six batches of rice panicles were tested. The results showed that the mean absolute percentage error (MAPE) for the system was about 1.23%, and the automatic measurements had a good agreement with manual measurements, regardless of panicle type. To evaluate the efficiency of this prototype, 3108 panicle samples were tested under continuous-measurement conditions, and the measuring efficiency was approximately 900 panicles per hour, 6 times over manual method. In conclusion, the system automatically extracts panicle length while providing three advantages over the manual method: objectiveness, high efficiency and high consistency.