Abstract:A new rapeseed oil seedling/weed detection method based on visual-attention model was put forward.The RGB images for the rapeseed oil seedling and weed in the seedling rapeseed oil field were obtained.Series of feature saliency using an improved Itti model in terms of the distribution characteristics of the original images were mapped.The ROI using region growing method were extracted.We calculated the shape and texture feature parameters of the regions segmented before and put them as the input of SVM used to identify seedling rapeseed oil regions.The weed regions were obtained by combining the original images with the seedling rape regions using a logical operation.The results showed that the correct segmentation rate,false segmentation rate and error segmentation rate of the proposed method was 92.46%,3.26% and 7.54%,respectively.It is indicated that the proposed method is better than the other two image segmentation methods.Using shape,texture,comprehensive and specifically selected feature parameters as the input,the classification rate of RBF-SVM was 96.00%,94.29%,100.00% and 96.00%,respectively.