Study On Insect Automatic Identification System And Web Edition Image-text Retrieval Enquiry System

In this dissertation two subjects have been studied:(1) Programed and developed the insect identification systems based on features of outlines and color spots by pattern recognition technique.(2) Developed the web edition image-text retrieval enquiry system by PHP web programing technique.Development for insect automatic identification system based on wing outline: a sample of seven species owlflies (Neuroptera: Ascalaphidae) was used in the system, whose outlines were digitized and transformed to Elliptic Fourier coefficients as feature data. And then, do the classifier model training by pattern recognition of support vector machines with the training data, and optimized parameter of model by the genetic algorithm (GA). In five repeated experiments, the mean accuracy for identification of each species ranged from90%to98%. The accuracy increased to99%when the samples were first divided into two groups based on features of their compound eyes.Development for insect automatic identification system based on wing color spots: four species owlfies was used in this study. We divide the wing color spots with the specific disproportions after normalizing the size and the orientation. And then, translate the subregion to HSV color place. At last, we get the average value which calculated by the Nonuniform Quantization for the region as the input value. Furthermore, we built the identification model with SVM and Adaboost separately, the accuracy with Adaboost was up to100%for4owlfly species, and the other accuracy was between90%and100%.Development for Insect web-image-text retrieval system: In this work, we develop an image-text retrieval system to integrate the test description and the feature image by PHP and MYSQL. We can see the detail features by zooming in and find the geographical distribution easily, which simplify our classification work.The goal of this dissertation study is developing of the application platform for insect identification. The two automatic identification systems and web edition image-text retrieval enquiry system have been packed all complicated calculate and programming in the software. Each system in this study could be used simply. The users of insect taxonomists can try to develop the automatic identification system of other insect groups based on only their professional knowledge, using this platform with the kind of outline and color characters. They can also try to make the wed edition image-text retrieval enquiry system of other insect groups using this platform. It also can be modify and extended easily with the development of the insect groups research. All systems developed in this study are useful tools for insect identification.