A novel weed recognition scheme based on fuzzy BP overall neural network is proposed. First, the classification features are blurred to deal with the uncertainty of weed features. Second, the genetic algorithm is used to optimize the network structure so as to improve the network’s convergence and stability. Finally, a feature-level data fusion scheme is used. In weed species identification experiments, neural network consists of the 4 BP sub-networks on color feature, main texture feature, secondary texture feature and spectral feature. The results indicate that the overall recognition rate reaches to a good recognition accuracy of 94.2% for 7 weed species. Besides, experiments were put into effect on the corn and its accompanying weeds. The neural network consists of the 4 BP sub-networks on color feature, main texture feature, height feature and spectral feature. The recognition rate reaches to 96.7% with a better recognition accuracy.