Feature Extraction of Plant Leaf Based on Visual Consistency

Feature extraction of plant leaf based on image processing technology has an great application prospect in plant taxonomy and intelligent agriculture and forestry production. In order to achieve feature data which can not only meet the automatic processing demands of computer but also be consistent with human understanding and determination on a leaf, a new idea on feature extraction named as feature extraction based on visual consistency (FEBVC) is presented. The main idea of FEBVC is conducting feature extraction in the same way as people describe an object. The key point of FEBVC is how to determine the direction in which to describe an object. FEBVC has been tried on shape feature extraction of plant leaf. Firstly, the plant leaf is rotated to a certain orientation with an improved inertia axis method according to human habit of observing an object. Then six shape feature parameters are designed to describe the shape of plant leaves according to human habit of describing an object. Many plant leaves with different shapes have been tested and the results show a good feasibility. FEBVC is very applicable to the establishment of intelligent expert systems.