Herbarium classification method and system integrated with multi-scale multi-directional texture characteristics

Abstract

The invention discloses a herbarium classification method and system integrated with multi-scale multi-directional texture characteristics. The herbarium classification method integrated with multi-scale multi-directional texture characteristics includes the steps: acquiring an image of herbarium; preprocessing the image; extracting the interesting area; segmenting the interesting area through a connected domain, and obtaining a segmented plant lamina set; calculating the geometrical characteristic of single lamina; through the amplitude-gradient histogram of the single lamina, determining the main direction of the texture of the lamina; utilizing different scales of Gaussian kernel to perform Gaussian filtering on the single lamina, and calculating the texture characteristic in the main direction; utilizing a characteristic matrix, and using an SVM (Support Vector Machine) algorithm to classify the single lamina, and obtaining a result set E of all the laminas of a specimen of a plant lamina; and taking the result with the greatest appearing time in the set E as the final classification result of the herbarium to be tested. The herbarium classification method integrated with multi-scale multi-directional texture characteristics has the advantages of being high in detection efficiency and classification accuracy, fully considering the multi-scale and difference factors, and greatly reducing the labor intensity of detection staff.

Description

technical field [0001] The invention belongs to the technical field of computer pattern recognition, and in particular relates to a method and system for classifying plant specimens by integrating multi-scale direction texture features. Background technique [0002] Plants are the most diverse and widely distributed form of life in the world, and have a very close relationship with human life and the environment. At the same time, plants are a very important part of the earth's biosphere and play an irreplaceable role in many fields of human production and life. With the development of modern civilization and the destruction of the ecological environment, plant species continue to decrease. Therefore, plant classification and identification are very important to protect plant diversity and maintain ecological balance. [0003] Compared with plant roots, stems, branches and other organs, plant leaves can extract features more easily and survive longer. At the same time, pl...

AI Technical Summary

Problems solved by technology

People with professional background knowledge to observe and determine different types of plants, although to a certain extent meet the needs of classification, but there are several problems in manual detection: one is that people tend to appear in the process of classifying plant leaves for a long time Wrong; second, it takes a lot of time for people to observe and identify similar plant leaves, which is inefficient; third, there are many kinds of plants, which require a lot of manpower and material resources, which wastes the use of professional talents to a certain extent.

[0005] It can be seen that simple manual detection and classification is time-consuming and labor-intensive and inefficient

The Chinese invention patent "Automatic Plant Recognition Method and System" with the publication number CN103617417B discloses a method and system that can finally determine which species a plant belongs to through point cloud data construction, feature extraction, and category discrimination calculation. This method and system can realize the classification of plants. However, it only uses shape geometric features without considering factors such as direction, scale, and texture, so it cannot be applied to the identification environment of leaf specimens.

Images