Chlorella and chrysophyceae classification and identification method based on image feature deep learning

Abstract

The invention discloses a chlorella and chrysophyceae classification and identification method based on image feature deep learning, and the method comprises the steps: improving an existing ResNet algorithm during the establishment of a detection data set, increasing the speed of network transmission, achieving the free transformation of the dimension of a feature map under the condition of not affecting the size of the feature map, and introducing the nonlinear operation simultaneously. The expression ability of the network is improved, and the classification accuracy is improved to 98%. Besides, a convolutional layer used for image feature extraction is improved on the basis of Faster RCNN, an FPN structure is added in the process of extracting feature maps through a backbone network, and a plurality of feature maps with different resolutions are extracted from the backbone network for subsequent RPN operation, so that it is ensured that the features of smaller chlorella and chrysophyceae are not lost; the size and the length-width ratio of the anchor points are designed, and the detection efficiency is effectively improved.

Description

technical field [0001] The invention relates to a method for classifying and identifying planktonic algae, in particular to a method for classifying and identifying chlorella and golden algae based on deep learning of image features. Background technique [0002] Chlorella and golden algae are two common and similar-looking phytoplankton algae in the ocean, which can be used as bait for seafood farming. In some occasions, it is necessary to classify and identify the two in a mixed state, such as clarifying the primary nutrient structure and source of the ocean, or effectively identifying the source and main factors of marine disasters such as red tides and shellfish poisoning. [0003] Deep learning is the main method for image classification, but in deep learning, as the number of network layers increases, there will be problems such as increased consumption of computing resources, easy over-fitting of the model, gradient disappearance, and gradient explosion. In the VGG n...

AI Technical Summary

Problems solved by technology

[0004] When classifying and identifying Chlorella and Chlorella based on images, if high-magnification microscopic images are used, the target in the field of view of a single sample image is clear, the target unit area is large and easy to identify, which is convenient for manual labeling, but the classification and identification efficiency is low; and When low-magnification microscopic images are used, Chlorella and Chrysophytes occupy a very small area and are densely distributed, and the target recognition level is low. If manual labeling is used, the time cost will be greatly increased and error labeling will easily occur. Case

Therefore, the existing classification technology based on image features is not suitable for the classification and identification of Chlorella and Chrysophytes in low-magnification microscopic images, and there is a problem of low accuracy.

Images