Identifying  plants is a challenging task considering the large number of existing species in the world. The inter-species similarity and the intra-species variability make the identification task particularly difficult and time consuming.

The difficulty encountered by non-botanists when identifying species using standard flora is centred on three major constraints, namely:

   • The ability to identify the species;
   • The use of dichotomous key;
   • The use of difficult technical terms.
In order to minimise and resolve these constraints, our identification system uses a graphical system which reconstitutes the plants using image processing.

Motivation

• Develop a tool for biodiversity assessment based on image analysis
• Vietnam is one of countries which have the richest biodiversity in the world
• Exploit image content

 

Demos Take an image of the leaf Browse images in the library Identify leaf Database Leaf images of 32 plants species 1907 images, each plant consists about 50 to 70 images, 1600*1200 pixels General images that do not contain leaf Total images: 1952 (training dataset: 1472 images, testing dataset: 480 images)

 

Các  công nghệ và tools sử dụng The algorithm determines the leaf / no leaf: GIST and characterized using the KNN classification  Leaf recognition algorithm: SURF features, BOW and SVM classifier  Android applications on the client-server architecture

• comvis