论文部分内容阅读
In recent years,the UAV(Unmanned aerial vehic le)plant protection machine was widely being popular ized and used for fertilizing and spraying.But the most agricultural UAVs could not collaborat ive decision with pest detection and pesticide technology.It brought some problems of low utilization rate of effective pesticides,excessive pesticide residues in agricultural products,environmental pollution,and etc.The current detection technology of vegetable pests mainly relied on artificial statistics,which had many shortages such as large amount of labor,low efficiency,feedback relay,artificial fault,and also without decision support for pesticide spraying.It is not only reducing manpower and pesticides use by rapid detection technology of image processing for vegetable pests,but also it could provide decision support for the UAV precision spraying and improve the quality of vegetables.Recently,experts and scholars have conducted numerous studies on pests of rice,wheat,corn,rape and other major economic crop.Nevertheless,practical research achievements on the rapid identif ication technology of vegetable pests are still relatively lacking.Based on the above background,this paper presented a classification and recognition scheme based on the model of a bag of features and support vector machine(BOF-SVM)on four southern important vegetable pests including Whitef lies,Phyllotreta Striolata,Plutella Xylostella and Thrips.The scheme of this paper consists of four sub-algorithms.The first sub-algorithm is computing the character description of pest images based on scale invar iant feature transform.The second sub-algorithm is computing the visual vocabulary based on bag of features.The third sub-algorithm is computing the classifier of pests based on support vector machine.The last one is classifying the pest images using the classifier.In this paper,C++ and Python language were used as implementat ion technology with Opencv and Libsvm function library based on BOF-SVM classification algorithm.Experiments showed that the average recognition accuracy was 91.56%for single image category judgment with 80 images from the real environment,and the average time was 0.39 seconds.This algorithm has achieved the ideal operating speed and precision.It could provide decision support for the UAV precision spraying,and also has better application prospect in agriculture.