Research On Compression Methods For Image Region Of Interest In Bamboo Germplasm Resources

Bamboo is an important forest resource and the collection of basic data of its germplasm resources is the precondition of comprehensive utilization.However,in the process of collecting basic data of bamboo germplasm resources,the acquisition site is mostly located in remote mountainous areas.Convenient acquisition and real-time transmission become the key,especially the acquisition and transmission of bamboo images.Due to the limitation of communication channel bandwidth in alpine areas,it is difficult to the fast transmission of bamboo image data.This paper takes the data of bamboo germplasm resources as the research object,and focuses on solving three problems,namely,the problem of convenient acquisition of bamboo germplasm resources data,the problem of effective extraction of ROI from bamboo images with complex background in the field,and the problem of region of interest compression of bamboo images based on JPEG2000.The image coding based on the region of interest is proposed,which can preserve the target information to the greatest extent while compressing the image data,and lays a foundation for the research on the classification and recognition of bamboo images.The main research work of this paper is as follows:(1)The Itti visual model ROI extraction method combined with Sobel edge detection is proposed.Aiming at bamboo digital images with complex background in the field,Sobel and Itti visual attention mechanism are combined to extract the color,brightness,direction and edge features of bamboo images,so as to extract ROI from bamboo images.The experimental results show that this method has better extraction effect than the commonly used ROI extraction method,and solves the problem that traditional extraction method is difficult to separate the target from the background in the complex background of bamboo images.(2)The ROI compression method of bamboo images based on Sobel-Itti visual model is studied.Firstly,the image was preprocessed before encoding.Sobel-Itti visual model is used to extract the ROI of bamboo images in the field to calculate the ROI mask.Max-shift image coding method is used to conduct lossless or low-magnification compression for the ROI part and high-magnification compression for the background part.Finally,the evaluation of image quality under ROI-encoded and non-ROI-encoded,the simulation experiment verifies that the algorithm proposed in this paper has better compression performance than the general coding method,and is more consistent with human visual characteristics.(3)The ROI-encoded effect verification model of bamboo images based on SVM is studied.Introducing SVM classification method,comparing the classification results of original image and the same bit rate of ROI-encoded image and non-ROI-encoded image,the results show that ROI-encoded bamboo image has better classification effect;comparing the classification results of ROI-encoded images at different bit rates,the results show that the ROI-encoded can effectively improve the classification effect of bamboo images at a certain bit rates.(4)The data acquisition system of bamboo and rattan germplasm resources based on Android system was developed.A germplasm resource data acquisition system was designed and developed on Android Studio platform with the Java programming and the MVP framework.The image acquisition and image coding modules are integrated on the terminal acquisition equipment.The specific functions of the system include information acquisition(bamboo and rattan geographic information,genus and species information,morphology information and material information),bamboo and rattan retrieval module and user management module,so as to realize convenient and rapid information acquisition,view and management.The research content of this paper provides technical support for collecting bamboo and rattan germplasm data conveniently in the field and the construction of large-scale shared database of bamboo and rattan resources.ROI-encoded has strong practical significance for the classification and recognition of bamboo and rattan images in the later stage.