Study on spike detection of cereal plants

The spike of a cereal plant is the grain-bearing organ whose physical properties are therefore critical components for plant yield. The ability to detect spikes from 2D images of cereals, such as wheat, provides vital information on tiller number and plants yield potential. We propose a novel spike detection method, which uses both RGB and fluorescence images. Firstly, an improved colour index method is used in the segmentation of plant feature from visible light RGB images, while threshold colour segmentation is applied to fluorescence images. Secondly, morphology algorithms are used to refine the initial segmentation results. Thirdly, a neural network-based method using Laws texture energy is performed for spike detection; connected and overlapping spikes are identified based on the average area and average perimeter of spikes and separated by a straight line through the center of the overlapping region. Finally, the spikes are extracted using area and height thresholds. Experimental results with two data sets have shown that the proposed method is simple, practical and can identify spikes with an accuracy of over 80%.