A plant image segmentation and leaf skeleton extraction method and system

Abstract

The invention relates to a plant image fine segmentation and leaf skeleton extraction method and system. The method comprises the following steps of: constructing a hybrid neural network model comprising a plurality of sub-neural networks; collecting several plant images as training image samples, labelling the corresponding plant class and leaf skeleton according to the training image samples, and training the hybrid neural network model according to the labeled training image samples; The target plant image is collected and input into the trained hybrid neural network model. The trained hybrid neural network model outputs plant image segmentation results and plant leaf skeleton extraction results. The invention relates to a plant image segmentation and a leaf skeleton extraction method,By constructing the hybrid neural network model and training the hybrid neural network model by collecting plant images, the plant image fine segmentation and leaf skeleton extraction can be realizedsimultaneously, and the interference of negative complex background can be restrained effectively, and the precision of plant image segmentation and leaf skeleton extraction can be improved.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a plant image segmentation and leaf skeleton extraction method and system. Background technique [0002] In the field of botanical research, scientists can conduct more in-depth research on genetic inheritance by studying the growth process of plants. With the rapid development of image technology, plant image analysis can quickly and accurately provide quantitative plant growth status indicators to scientists, assisting scientists in further in-depth research. Today's hot deep learning methods have been widely used in plant image segmentation and recognition, such as convolutional neural networks. Commonly used convolutional neural networks include VGG, FCN, and U-Net. These networks have deep structures and can extract deep image features to greatly improve the accuracy of plant image segmentation and recognition. [0003] For complex plant image scenes, although VGG...

AI Technical Summary

Problems solved by technology

[0003] For complex plant image scenes, although VGG can achieve fine segmentation of plants, it is very easy to produce false detections on complex backgrounds (such as the reflection of strong light in the plant incubator, the shadow of plants in the observation mirror, etc.), that is, it is very difficult to interfere with complex backgrounds. Sensitive; FCN has a strong ability to suppress complex background interference, but it is difficult to achieve fine plant segmentation, missing small plants and their leaves; U-Net integrates shallow image detail information and deep image semantic features, in fine plant segmentation and complex background A performance compromise is achieved between interference suppression, but U-Net cannot output plant segmentation results and leaf skeleton extraction results at the same time

Images