Plant classification and identification is the basic work for the research in plant field,and it has the great significance to protect plants, distinguish plant species and explore thegenetic relationship among plants. The most effective and simple way to identify plants isbased on the character of the leaves of plants.This paper uses the neural network to classify and identify plants with somecharacteristic parameters extracted from plants'leaves. As the input sample for neuralnetwork, the extracted parameters include invariant moments, Fourier descriptors andfractal dimension, which are not sensitive to translation, scaling, mirroring and rotation. Inorder to improve the accuracy of characteristic parameters, some pre-processing isnecessary: the conversion from color image to gray image, the binarization of the grayimage, the erosion operations on binary images for filtration and so on.The BP neural network is used to classify and identify the processed images.Experimental results show that neural network trained with the Fourier descriptors canonly identify the training samples, but almost no test samples. when training samplescontain invariant moments and fractal dimension at the same time. Under this condition therecognition rate can reach 92% while the hidden nodes of BP network are 9. |