Backpropagation neural network design and evaluation for classifying weed species using color image texture

Color co-occurrence method (CCM) texture statistics were used as input vari ables for a backpropagation (BP) neural network weed classification model. Thirty-three unique CCM texture statistic inputs were generated for 40 imag es per class, within a six class data set. The following six classes were s tudied: giant foxtail, large crabgrass, common lambsquarter, velvetleaf, iv yleaf morningglory, and clear soil surface. The texture data was used to bu ild six different input variable models for the BP network, consisting of v arious combinations of hue, saturation, and intensity (HSI) color texture s tatistics. The study evaluated classification accuracy as a function of net work topology, and training parameter selection. In addition, training cycl e requirements and training repeatability were studied The BP topology eval uation consisted of a series of tests on symmetrical two hidden-layer netwo rk, a test of constant complexity topologies, and tapered topology networks . The best symmetrical BP network achieved a 94.7% classification accuracy for a model consisting of II inputs, five nodes at each of the two hidden l ayers and six output nodes (11 x 5 x 5 x 6 BP network). A tapered topology (11 x 12 x 6 x 6 BP network) out performed all other BP topologies with an overall accuracy of 96.7% and individual class accuracies of 90.0% or highe r.