Weed identification method based on probabilistic neural network in the corn seedlings field

Discrimination between corn seedlings and weeds is an important and necessary step to implement spatially variable herbicides application. This paper proposed a method of weed identification by using the technique of image processing and probabilistic neural network. Otsu´s method for automatic threshold was applied to segment weeds images based on the modified excess green feature, it could distinguish the plant objects from the background effectively whether the plant objects were covered with wheat straw residue seriously or not. The probabilistic neural network classifier was created for recognition of corn seedlings and weeds according to the shape features. Comparing the probabilistic neural network (PNN) method with the back-propagation neural network one, the former is better than the latter seeing from the experimental results. The former method gave the recognition rate of 92.5% (corn seedlings) and 95% (weeds).