| Popis: |
Food is the basic necessity of every living being, unfortunately, vegetables and fruits became harmful because of the overuse of pesticides. Toxins are unsuccessfully spread in farms for increasing the harvest amounts and quantities, needless to say, that causes serious health problems. Therefore, more reliable information regarding pests should be provided to farmers in a user-friendly way. In this work, a novel method for automatic pest identification using local binary pattern (LBP) feature is proposed, implemented, then tested using real-life images. Firstly, the image is resized and the a* plane of the L*a*b* color space is used to extract the region of interest (ROI). Secondly, the histogram of the LBP of the ROI is calculated. Finally, a correlation matching is performed to measure the similarity between input ROI and predefined trained templates of different pests. The model output includes the pest name and symptoms. Experiments conducted on the variety of pests of rice plant infected leaves showed a 92.4% true identification rate which makes this method reliable comparing with the reported results from other works. The detection time also is 10 ms/ frame which fulfills one of the real-time application requirements. |
