Suzhou Electric Appliance Research Institute
期刊號(hào): CN32-1800/TM| ISSN1007-3175

Article retrieval

文章檢索

首頁(yè) >> 文章檢索 >> 往年索引

一種改進(jìn)SSD算法的輸電線路目標(biāo)檢測(cè)方法

來(lái)源:電工電氣發(fā)布時(shí)間:2021-06-28 09:28 瀏覽次數(shù):602
一種改進(jìn)SSD算法的輸電線路目標(biāo)檢測(cè)方法
 
黃芹芹,董潔,陳玥,朱圓圓
(沈陽(yáng)建筑大學(xué) 信息與控制工程學(xué)院,遼寧 沈陽(yáng) 110168)
 
    摘 要 :電力巡檢在輸電線路部件故障的排除中起著至關(guān)重要的作用。為了實(shí)現(xiàn)復(fù)雜背景下的輸電線路電力小部件的目標(biāo)檢測(cè),提出了一種改進(jìn) SSD 算法的小目標(biāo)檢測(cè)算法——PA-SSD。將反卷積融合單元融合到 PANet 算法中,以改進(jìn) PANet 結(jié)構(gòu),并以此為基礎(chǔ)產(chǎn)生新的特征融合方式,融合不同尺度的特征圖 ;將傳統(tǒng) SSD 算法中的特征圖用新的特征圖替換,形成新的特征金字塔模型。針對(duì)實(shí)際輸電線路中的 4 種目標(biāo)進(jìn)行了測(cè)試,結(jié)果表明,PA-SSD 算法與原始的 SSD 算法相比,其檢測(cè)精度有了明顯提高,檢測(cè)速度也可以滿足檢測(cè)性能的要求。
    關(guān)鍵詞 :目標(biāo)檢測(cè) ;輸電線路 ;SSD 算法 ;PANet 算法
    中圖分類號(hào) :TM726     文獻(xiàn)標(biāo)識(shí)碼 :A     文章編號(hào) :1007-3175(2021)06-0051-05
 
A Transmission Line Target Detection Method with
Improved SSD Algorithm
 
HUANG Qin-qin, DONG Jie, CHEN Yue, ZHU Yuan-yuan
(School of Information and Control Engineering, Shenyang Jianzhu University, Shenyang 110168, China)
 
    Abstract: It is of great importance for the power inspection to troubleshoot the transmission line components. In order to achieve target detection of small power components of transmission lines in complex contexts, this paper proposes a small target detection algorithm based on the improved SSD algorithm-PA-SSD. The deconvolution fusion unit is integrated into the PANet algorithm to improve the PANet structure, and on this basis, a new feature fusion method is generated to fuse feature images of different scales. The feature graph in the traditional SSD algorithm is replaced with a new feature graph to form a new feature pyramid model. The results show that compared with the original SSD algorithm, the detection accuracy of PA-SSD algorithm is significantly improved, and the detection speed can also meet the require ments of detection performance.
    Key words: target detection; transmission line; SSD algorithm; PANet algorithm
 
參考文獻(xiàn)
[1] 鄒棟 . 輸電線路中關(guān)鍵部件圖像識(shí)別及異常檢測(cè)方法研究 [D]. 保定 :華北電力大學(xué),2017.
[2] 徐俊良,曹建,楊凱斌 . 基于局部均值的絕緣子圖像分割算法設(shè)計(jì) [J] . 計(jì)算機(jī)工程,2016,42(9) :262-267.
[3] 趙君 . 基于航拍圖像的桿塔類型判斷及狀態(tài)識(shí)別[D]. 保定 :華北電力大學(xué),2010.
[4] 王麗娟 . 基于深度卷積神經(jīng)網(wǎng)絡(luò)的絕緣子故障檢測(cè)算法研究 [D]. 天津 :天津工業(yè)大學(xué),2018.
[5] LIU W, ANGUELOV D, ERHAN D, et al.SSD :Single Shot MultiBox Detector[C]//Lecture Notes in Computer Science,2016 :21-37.
[6] 于德明,沈建,汪駿,等 . 直升機(jī)在電網(wǎng)運(yùn)行維護(hù)中的研究與應(yīng)用 [J] . 電網(wǎng)技術(shù),2009,33(6) :107-112.
[7] 史高峰,李小勇 .SSD 數(shù)據(jù)結(jié)構(gòu)與算法綜述 [J]. 微型電腦應(yīng)用,2012,28(4) :56-61.
[8] LIN T Y, DOLLAR P, GIRSHICK R, et al.Feature pyramid networks for object detection [C] / /2017 IEEE Conference on Computer Vision and Pattern Recognition(CVPR),2017 :936-944.
[9] LIU S, QI L, QIN H, et al.Path aggregation network for instance segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,2018 :8759-8768.
[10]SHELHAMER E , LONG J , DARRELL T . Fully convolytional networks for semantic segmentation[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2017,39(4) :640-651.