A crop disease image retrieval method based on the improvement of inverted index

According to the characteristics of crop leaf disease images, we proposed a new image retrieval method based on the improvement of inverted index to diagnose crop leaf diseases. First of all, the input crop disease images were preprocessed, including compression, denoising, enhancement, etc. And then the features of disease in the whole image were extracted. Meanwhile, in order to reduce the storage space of inverted index feature vectors, the Hash method was adopted to map the inverted index feature vectors to binary values. Hamming distance was used in the similarity calculation between the obtained features data and the lesion features from the constructed disease images indexes. According the ranking of similarities, top 5 images were selected as the candidate diagnostic results list of the input crop disease image. And the results were evaluated by some standard criteria, such as precision, recall, etc. The experiments were conducted on cucumber disease images, including: downy mildew, powdery mildew and target spot disease, and rice disease images, including: rice blast, leaf spot and sheath blight. The results showed that the proposed method can achieve the higher retrieval accuracy than traditional SVM method both of cucumber and rice disease images.