【摘要】：隨著智能手機(jī)的普及,移動(dòng)互聯(lián)網(wǎng)的影響已經(jīng)深入到人們生活和生產(chǎn)的方方面面。目前的農(nóng)業(yè)使用的圖像及相關(guān)信息采集傳輸和管理系統(tǒng)存在獲取的圖片數(shù)量過(guò)大、無(wú)法自動(dòng)識(shí)別系統(tǒng)相關(guān)圖片以及檢索效率低等問(wèn)題。為了滿足現(xiàn)代農(nóng)業(yè)生產(chǎn)的實(shí)際需要,幫助農(nóng)業(yè)專家提高工作和研究效率,本文設(shè)計(jì)并實(shí)現(xiàn)了基于Android作物冠層圖像采集傳輸及其WEB端管理系統(tǒng)。該系統(tǒng)先利用Android智能手機(jī)對(duì)圖像及相關(guān)信息進(jìn)行采集并上傳到云服務(wù)器,然后通過(guò)非作物冠層圖像自動(dòng)識(shí)別分類模塊對(duì)上傳圖像進(jìn)行自動(dòng)識(shí)別分類,最后通過(guò)web端系統(tǒng)對(duì)所有上傳圖像及相關(guān)信息進(jìn)行管理。論文主要研究工作如下:(1)介紹了基于Android作物冠層圖像采集傳輸及其WEB端管理系統(tǒng)的研究意義和國(guó)內(nèi)外具體研究現(xiàn)狀。闡述了各研究中所用到的方法及他們所取得的成果,總結(jié)目前研究存在的一些問(wèn)題,確定本文研究的主要目標(biāo)。(2)介紹了Android軟件結(jié)構(gòu)和框架,并對(duì)圖像及其相關(guān)信息數(shù)據(jù)網(wǎng)絡(luò)傳輸相關(guān)技術(shù)進(jìn)行研究和比較,并從中選擇本系統(tǒng)的傳輸方案。接著闡述了服務(wù)器端SSM框架,分析比較了其相對(duì)其他框架的優(yōu)勢(shì)。然后介紹了數(shù)據(jù)庫(kù)sharding策略原理,最后介紹了淺層學(xué)習(xí)和深度學(xué)習(xí)的區(qū)別,并比較了現(xiàn)有的圖像機(jī)器學(xué)習(xí)服務(wù)。(3)本文的重點(diǎn)和創(chuàng)新點(diǎn)是研究非作物圖像識(shí)別實(shí)現(xiàn)和應(yīng)用,使用計(jì)算圖像綠色特征像素覆蓋率的方法作為第一輪判斷依據(jù)。先對(duì)上傳圖像處理光照影響,通過(guò)改進(jìn)的超綠色法得到目標(biāo)強(qiáng)化后的灰度圖,使用黃氏模糊閾值算法將灰度圖二值化為目標(biāo),然后結(jié)合連通域標(biāo)記算法和面積過(guò)濾法填充背景中的雜物和目標(biāo)物中的微小孔洞,最后計(jì)算出綠色特征像素覆蓋率,比較實(shí)驗(yàn)結(jié)果得出通過(guò)以上步驟完整處理的相對(duì)覆蓋率誤差最小。通過(guò)微軟圖像深度學(xué)習(xí)服務(wù)對(duì)上傳圖片的返回值和置信度作為第二輪判斷依據(jù),根據(jù)對(duì)各類別圖片進(jìn)行實(shí)驗(yàn),選取判定上傳圖片為作物圖像的返回值及其置信度。通過(guò)以上兩輪篩選能夠準(zhǔn)確判斷上傳圖片是否為對(duì)于系統(tǒng)有效的作物冠層圖像。在服務(wù)器接收端使用線程池,能夠?qū)崿F(xiàn)多個(gè)Android客戶端同時(shí)上傳圖像及相關(guān)數(shù)據(jù)到服務(wù)器。針對(duì)防止惡意占用服務(wù)器端口線程的問(wèn)題通過(guò)對(duì)連接服務(wù)器的IP地址進(jìn)行檢測(cè)和計(jì)數(shù)解決。(4)根據(jù)需求對(duì)基于Android作物冠層圖像采集傳輸及其WEB端管理系統(tǒng)各模塊功能和流程進(jìn)行設(shè)計(jì),完成了系統(tǒng)從服務(wù)器的數(shù)據(jù)庫(kù)到業(yè)務(wù)邏輯的實(shí)現(xiàn)、頁(yè)面的展示和系統(tǒng)各部分的交互,最后通過(guò)系統(tǒng)測(cè)試得出本系統(tǒng)各部分能達(dá)到預(yù)期效果,在服務(wù)器接收端能夠自動(dòng)識(shí)別分類非作物冠層圖片。
[Abstract]:With the popularity of smart phones, the influence of mobile Internet has reached every aspect of people's life and production. In the current agricultural image collection, transmission and management system, the number of images obtained is too large to automatically identify the related images and the retrieval efficiency is low. In order to meet the practical needs of modern agricultural production and to help agricultural experts improve their work and research efficiency, this paper designs and implements a crop canopy image acquisition and transmission system based on Android and its WEB terminal management system. The system first uses Android smart phone to collect images and related information and upload them to the cloud server, and then automatically recognizes and classifies the uploaded images through the automatic recognition and classification module of non-crop canopy images. Finally, all the uploaded images and related information are managed through the web system. The main research work of this paper is as follows: (1) the research significance of crop canopy image acquisition and transmission and its WEB terminal management system based on Android are introduced, as well as the specific research status at home and abroad. This paper expounds the methods used in each research and their achievements, summarizes some problems existing in the present research, and determines the main objectives of this paper. (2) the structure and framework of Android software are introduced. The related technology of network transmission of image and related information data is studied and compared, and the transmission scheme of this system is selected. Then the server-side SSM framework is expounded, and its advantages compared with other frameworks are analyzed and compared. Then it introduces the principle of database sharding strategy, finally introduces the difference between shallow learning and deep learning, and compares the existing image machine learning services. (3) the emphasis and innovation of this paper is to study the realization and application of non-crop image recognition. The method of calculating image green feature pixel coverage is used as the basis of the first round of judgment. Firstly, the effect of illumination on the uploaded image is processed, then the enhanced gray image is obtained by the improved super-green method, and the binary value of the gray image is transformed into the target by using Huang's fuzzy threshold algorithm. Then the connected domain labeling algorithm and the area filter method are used to fill the debris in the background and the tiny holes in the object. Finally, the green feature pixel coverage is calculated. By comparing the experimental results, it is concluded that the relative coverage error can be minimized by the complete processing of the above steps. The return value and confidence degree of uploaded images are used as the second judgment basis through the Microsoft Image depth Learning Service. According to the experiments of each kind of images, the returned value and confidence of the uploaded images are selected as crop images. Through the above two rounds of screening, we can accurately judge whether the uploaded image is an effective crop canopy image for the system. The thread pool is used in the receiving end of the server, and multiple Android clients can upload images and related data to the server at the same time. The problem of preventing the malicious occupation of server port thread is solved by detecting and counting the IP address of the connection server. (4) according to the requirement, every module of the crop canopy image collection and transmission based on Android and its WEB end management system is realized. Able to design and process, The realization of the system from the database of the server to the business logic, the display of the page and the interaction of the various parts of the system are completed. Finally, through the system test, it is concluded that each part of the system can achieve the desired results. Can automatically identify and classify non-crop canopy images at the server receiving end.
【學(xué)位授予單位】：浙江理工大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類號(hào)】：TP391.41;TP315

【相似文獻(xiàn)】

相關(guān)期刊論文 前10條

1 林耕宇;;觀摩50名Google Android程序開發(fā)競(jìng)賽作品[J];電子與電腦;2008年08期

2 樹子;;Android中文版不完全體驗(yàn)[J];互聯(lián)網(wǎng)天地;2009年04期

3 Jason Whitmire;;產(chǎn)業(yè)軟件專家如何協(xié)助解決Android的分裂困境[J];電子與電腦;2010年02期

4 蔣彬;;10款A(yù)ndroid手機(jī)必備應(yīng)用——Android操作系下的軟件評(píng)測(cè)[J];微電腦世界;2010年04期

5 ;PCWorld Windows Phone 7挑戰(zhàn)Android 毅然崛起的AndroidⅠ洗心革面的Windows Phone 7[J];微電腦世界;2010年08期

6 韓青;;Android平臺(tái)發(fā)展的動(dòng)力與挑戰(zhàn)[J];中國(guó)電子商情(基礎(chǔ)電子);2010年09期

7 方智勇;;Android手機(jī)這樣用[J];電腦迷;2010年15期

8 缺少浪漫;;Android的另一面[J];電腦迷;2010年13期

9 ;ZTE and Three Release Android ，

本文編號(hào)：2270722