Plant identification through images: Using feature extraction of key points on leaf contours(1.)

Area of Science:

• Botany
• Computer Vision
• Image Recognition

Background:

• Plant identification requires extensive botanical knowledge and field time.
• Leaf characteristics are crucial for species identification.
• Automated leaf recognition from images is a significant research challenge.

Purpose of the Study:

• To develop an automated leaf recognition method using image analysis.
• To propose a novel feature extraction technique for leaf contours.
• To evaluate the performance and invariance properties of the proposed method.

Main Methods:

• Feature extraction from leaf contours using lines between centroid and contour points.
• Creation of a length histogram to represent contour distance distribution.
• Application of a statistical classifier for matching leaf templates and queries.

Main Results:

• Achieved top accuracy of 92.7% and 97.3% for the first two values.
• Demonstrated high scale invariance with correlation coefficients between 0.98611 and 0.99992.
• Showcased strong rotation invariance with correlation coefficients between 0.98071 and 0.99988.

Conclusions:

• The proposed leaf feature extraction method is invariant to scale and rotation.
• The method demonstrates superior performance compared to Zernike moments and curvature scale space.
• This approach offers a robust solution for automated plant identification via leaf images.