UAV Based Remote Sensing for Tassel Detection and Growth Stage Estimation of Maize Crop using F-RCNN

The information about the critical growth stage of maize such as tasseling is known very important because it indicates the transition from vegetative to reproductive growth, which will also help in taking potential actions to improve the crop health as well as optimizing the use of agronomic inputs. Manual observation of different stages of the crop is a difficult and challenging task as the height and density of the crop increases day by day. To reduce human efforts, researchers have used computer vision approach to detect and count the number of tassels in the images captured from the static camera. To reduce human efforts, computer vision based approaches are proposed wherein [1]-[2], color, shape, and texture based features used while in [3], convolutional neural network (CNN) based approach used for tassel detection and counting in images from a static camera. However, these works also do not cover the complete growth of tassel at different stages as it changes its color and texture as it grows. It is a difficult problem to detect and count tassel at every stage. In this paper, a framework is proposed to detect and count tassel at different stages using UAV as it is hard to collect data from a static camera in large fields. UAVs have the potential to provide a feasible solution for the large scale and high throughput phenotyping. The processing techniques of static images will be incompatible for images acquired via UAV due to the difference in spatial resolution, variation in background and moment of UAV. To the best of our knowledge, this is the first study on the detection of tassels and estimation of tasseling stages of maize crop using UAV based remote sensing. 2. Proposed Method