Rice heading stage automatic observation by multi-classifier cascade based rice spike detection method

The rice heading stage is an essential phase of rice production as it directly affects the rice yield. This paper transforms the issue of rice heading stage automatic observation into the problem of rice spike detection and proposes a new method for automatic observation of the rice heading stage. Rice spike detection is achieved using a new multi-classifier cascade method comprised of the following steps: First, SVM with color feature as input is utilized to distinguish the rice spike image patches from the background patches (leaf, soil, water, etc.); Second, a gradient histogram method is applied to remove the yellow leaf patches from consideration; Third, a convolutional neural network (CNN) is utilized to further reduce the false positive rate. The arrival of the rice heading stage is determined by the number of the detected spike patches. To evaluate the proposed method, it was applied to the automatic rice heading stage observation of six image sequences collected by the designed observation device between 2011 and 2013. In the experiment, the proposed method produced similar results to the conventional manual observation method in determining the arrival of the rice heading stage. The differences between the proposed method and manual way were within two days. Experiments demonstrated that the proposed method is an effective approach of automatic observation of the rice heading stage in paddy fields and can be utilized to replace the manual observation.