本文選題：目標識別 + 圖像處理�。� 參考：《農(nóng)業(yè)工程學報》2017年15期

【摘要】：為實現(xiàn)番茄不同器官的快速、準確檢測,提出一種基于深度卷積神經(jīng)網(wǎng)絡(luò)的番茄主要器官分類識別方法。在VGGNet基礎(chǔ)上,通過結(jié)構(gòu)優(yōu)化調(diào)整,構(gòu)建了10種番茄器官分類網(wǎng)絡(luò)模型,在番茄器官圖像數(shù)據(jù)集上,應(yīng)用多種數(shù)據(jù)增廣技術(shù)對網(wǎng)絡(luò)進行訓練,測試結(jié)果表明各網(wǎng)絡(luò)的分類錯誤率均低于6.392%。綜合考慮分類性能和速度,優(yōu)選出一種8層網(wǎng)絡(luò)用于番茄主要器官特征提取與表達。用篩選出的8層網(wǎng)絡(luò)作為基本結(jié)構(gòu),設(shè)計了一種番茄主要器官檢測器,結(jié)合Selective Search算法生成番茄器官候選檢測區(qū)域。通過對番茄植株圖像進行檢測識別,試驗結(jié)果表明,該檢測器對果、花、莖的檢測平均精度分別為81.64%、84.48%和53.94%,能夠同時對不同成熟度的果和不同花齡的花進行有效識別,且在檢測速度和精度上優(yōu)于R-CNN和Fast R-CNN。
[Abstract]:In order to detect tomato organs quickly and accurately, a classification and recognition method of tomato main organs based on deep convolution neural network was proposed. On the basis of VGGNet, 10 kinds of tomato organ classification network models were constructed by optimizing and adjusting the structure. The network was trained by using a variety of data augmentation techniques on the tomato organ image data set. The test results show that the classification error rate of each network is lower than that of 6.392. Considering the classification performance and speed, an 8-layer network was selected for feature extraction and expression of major organs of tomato. Using the selected 8-layer network as the basic structure, a tomato main organ detector is designed, and the candidate detection region of tomato organ is generated with Selective Search algorithm. The results showed that the detection accuracy of the detector was 81.64%, 84.48% and 53.94%, respectively, which could be used to identify the fruit of different maturity and the flower of different flower age at the same time, the results showed that the detection accuracy of the detector was 81.64% and 53.94g% respectively, and the results showed that the detection accuracy of the detector was 81.64% and 53.94%, respectively. It is superior to R-CNN and Fast R-CNN in detecting speed and accuracy.
【作者單位】： 沈陽農(nóng)業(yè)大學信息與電氣工程學院;
【基金】：遼寧省科學事業(yè)公益研究基金(2016004001) 國家自然科學基金(31601218)
【分類號】：S641.2;TP391.41

【相似文獻】

相關(guān)期刊論文 前2條

1 劉繼龍;張振華;謝恒星;;果園土壤貯水量神經(jīng)網(wǎng)絡(luò)估算模型研究[J];農(nóng)業(yè)系統(tǒng)科學與綜合研究;2007年01期

2 薄永軍;;溫室溫度控制系統(tǒng)神經(jīng)網(wǎng)絡(luò)PID控制算法研究[J];安徽農(nóng)業(yè)科學;2014年13期

相關(guān)碩士學位論文 前4條

1 陳銘暉;基于神經(jīng)網(wǎng)絡(luò)的蔬菜育苗潮汐灌溉策略研究[D];上海交通大學;2014年

2 米雅婷;基于GA-BP神經(jīng)網(wǎng)絡(luò)的溫室番茄病害診斷研究[D];東北林業(yè)大學;2016年

3 王曉娟;基于模糊控制與RBF神經(jīng)網(wǎng)絡(luò)的桃病蟲害發(fā)生預測研究[D];河北農(nóng)業(yè)大學;2011年

4 郝琪;智能溫室遠程控制系統(tǒng)研究與設(shè)計[D];燕山大學;2011年

，

本文編號：1848819