Research On Microscopic Characteristics For Chinese Herbal Medicine Powder Based On Deep Learning

With the continuous development and progress of computer image processing technology,the microscopic image identification technology of powdered Chinese herbal medicine has been further developed in the field of Chinese herbal medicine recognition due to its unique advantages.However,for large and multi-category microfeature images,using traditional image recognition algorithms for image recognition can cause problems such as low accuracy and poor universality.Therefore,this paper proposes a method for studying microscopic characteristics of powdered Chinese herbal medicine based on deep learning.The main work and innovations of this article are as follows:(1)According to the actual needs of microscopic image detection and classification,a set of reasonable microscopic image classification standards was developed to create Chinese herbal medicine microscopic image data sets.In order to remove the influence of interference factors such as lighting effects,angular offsets,and different scales on the experimental results in the database,a variety of image enhancement methods are used to preprocess the microscopic images of powdered Chinese herbal medicine to obtain more diverse image data and improve the accuracy of detection and identification of the microscopic feature images.(2)An improved detection algorithm based on SSD algorithm is proposed.Aiming at the problem of target fracture and incompleteness in microscopic feature images,the SE module is added after the predictive convolution map of the SSD.The feature channels that have a greater effect on the detection result get greater weight,which makes the network can fully learn the information of important feature channels.In addition,the improved network can fully fuse the information contained in the front and back feature maps,which is conducive to the effective learning of the information of the microscopic feature images by the network and improves the final detection effect.Detection experiments were performed on three types of microscopic characteristic images of thick-walled cells,ducts,and pollen spores.The results show that the Mean Average Precision(m AP)of improved algorithm’s has increased from 79.8% to 81.5%.(3)A binary image transformation method and an image type multi-channel fusion recognition algorithm are proposed.Aiming at the microscopic characteristic images of the catheter,the image type which determines the information content of a image is converted after the detection network in chapter 3.And enrich and expand image information by multi-channel fusion of RGB images,grayscale images and binary images.So that the deep learning network can learn more levels and more effective image information.Thereby,the identification of powdered Chinese herbal medicine can be completed better.The ACNet algorithm was used for Chinese herbal medicine recognition on the fused image,and the results showed that the recognition accuracy of the fused microscopic feature images reached 94.5%.