Efficient Modern Description Methods by Using SURF Algorithm for Recognition of Plant Species

Plants are one of one of most valuable natural resources. There would be no life on earth without plants. They unlike humans and animals manufacture their own food by photosynthesis. In every food chain, the plants occupy the first position and lead the chain as source of food. Environment and climate are largely interlinked with plants. Rainfall, humidity and temperature are influenced by presence of plants. Cutting down plants also imbalance the environment which will indirectly affect human life. Even the economic importance of plants is also quite large to mankind. Plants are great contributors of economy. Many countries rely on agriculture as one of the main source of revenue. Other benefits of plants are significant applications in different fields. Medical and agricultural applications are just some instances of plants application. Plant recognition can be done by using unique characteristic parts of plants. The used part is leaf. Shapes of leaves are useful to do plant recognition and find the species. Bag of words (BoW) and support vector machine (SVM) methods are applied to recognize and identify plants species. Visual contents of images are used and four steps are performed: (i) image preprocessing, (ii) BoW, (iii) train, (v) test. Three combined methods are used on Flavia dataset. The proposed approach is done by Speed-up robust features (SURF) method and two combined method, HARRIS-SURF and features from accelerated segment test-SURF (FAST-SURF). The accuracy of SURF method is higher than other applied methods. It is 92.28395 %. In addition to visional comparison, some quantitative results are measured and compared