Maize Pest Identification Based On SVM And DS Image Data Fusion

For the problem that corn planting person lack of systematic training and feel difficult to identify the corn pests accurately,the article took the corn pests as research objects,researching corn pests recognition algorithm based on the image data fusion of SVM and DS.the article studied on the identification method of 3 kinds of pests--cutworm larva,corn borer larva and prodenia litura larva which was common in corn planting process.Image shooted under the natural condition in image acquisition nodes would have the problem of the uneven illumination image,first,needed to modify the illumination so as to reduce the impacts for the images.After that,used median filter for imagine denoising,then conducted graying,binarization and open operation for easy extracting of image features.After image preprocessing,extracted the image eigenvalue.the image eigenvaluemay measure the images in numeralization and be the basis of image recognition.Different pests had some difference Thereinto,skewness as it's color features,calculated image's gray level co-occurrence matrix in gray space and calculated the it's contrst,correlation and energy as the texture feature,taking Hu matrix as the shape feature.Corn pests recognition algorithm adopted image data fusion algorithm based on SVM ans DS,this algorithm fully combined the advantages of SVM in dealing with the problem of small sample classification and the information fusion ability of multi-feature of DS in evidence theory.First adopting multi-classification SVM to calculate 3 kinds of characteristic of image and basic reliability distribution in the same identification frame respectively,then using the DS evidence theory to fuse the result,getting the final recognition result according to the decision rules.For the convenience of the use of corn planting personnel,designed and realized corn pests identification platform based on JavaWeb.The platform had many functions including login,configuration,network,pests identification,environmental monitoring and the historical data query.Image information adopted 3 g network transmission,environmental information transmitted through ZigBee,and set the corresponding data transfer protocol.At last,conducted the experiment on the recognition effect of corn pests through platform.Experiment was divided into twogroups: experiment images would be the corn pests picture under the natural background and which under the pure color background respectively.Each pest image had 200 pieces in two group respectively,calculated the classification results for SVMA on three kinds of image features respectively and fusion result for DS evidence theory.The final result showed that the SVM-DS algorithm could recognize corn pests better and met expectations,could be applied to corn field,better pest of corn,identified the image in real-time through the image acquisition device,thereby improved the corn yield and quality,saved manpower and material resources.