基于深度信息融合的航拍车辆检测

随着汽车数量的快速增长以及无人机飞控技术的迅速发展,基于无人机航拍的车辆检测技术越来越有用武之地。传统的基于滑动窗口以及手工设计特征的车辆检测不仅计算量巨大,鲁棒性也不够好。卷积神经网络在目标检测方面发挥了显著的优势,但是常见的网络对于航拍遥感图像中的小目标检测效果一般。文中基于faster-RCNN在VGG16网络上使用通道合并融合的方式设计了超特征图,通过结合浅层特征以及深层特征的方式提取小目标的特征以提高检测的召回率。同时修改RPN层的包围框的大小以提高检测的准确性。在慕尼黑车辆数据集以及自己收集的数据上进行了测试,通过对比实验可知,该方法使得车辆检测的效果有了明显提升,在两个数据集上分别达到了72.3%和80.5%的mAP。

基于SLIC和改进区域生长的非结构化道路识别

非结构化道路一般没有车道标识线且道路边界模糊,区分道路区域与背景区域难度较大。针对现有非结构化道路识别方法存在全像素域计算分类处理实时性差、易受噪声数据干扰等问题,提出一种基于SLIC(simple lineariterativeclustering)超像素分割和改进区域生长算法的非结构化道路识别方法。利用均匀化初始聚类中心的SLIC算法生成低分辨率超像素特征图。在此基础上,利用聚类算法与邻域搜索算法自适应选择种子点,并引入CIEDE2000色差理论作为区域生长法生长准则,初步确定道路区域。根据道路连续一致特点,优化超像素级生长图并映射轮廓区域至原图,获得道路最终区域。基于数据集及真实场景的实验结果表明,该方法具有较高的识别率和抗干扰能力。

Sonographic features of duodenal lipomas in eight clinicopathologically diagnosed patients

AIM:To investigate the sonographic features and diagnostic value of endoscopic ultrasonography (EUS) for duodenal lipomas (DLs).METHODS:A total of eight consecutive patients with DL diagnosed pathologically were included in the study.One EUS expert reviewed the ultrasonic images for all lesions,including the original layer of the duodenal wall,the echo intensity and the echo homogeneity.The size of the lesions and the perifocal structures were also investigated.The diagnosis by EUS was compared with the histological results.RESULTS:Using routine endoscopy,only one case was correctly diagnosed as DL.Four cases were classified as submucosal tumors,and three cases were mistaken for stromal tumors.All tumors appeared as round or oval intensive hyperechoic lesions with distinct anterior borders that originated from the submucosal layer on EUS.Tumors ranged from 8 to 36 mm in size,with an average size of 16 mm.Homogeneous echogenicity was seen in all cases except one that had a tubular structure inside the tumor.Echo attenuation was observed only in the area behind the tumors in five cases,and it was observed both inside and behind the tumors in three cases in which the posterior border was obscure or invisible.Seven (87.5%) cases were correctly diagnosed as DL,and one (12.5%) was mistaken as Brunner's gland adenoma by EUS.Pathologically,all tumors originated from the submucosal layer and consisted of mature fat cells without heteromorphism.Among the fat cells,there was a small amount of thick-wall vessels infiltrating the lymphocytes,and abundant fibrous connective tissues.CONCLUSION:On EUS,DL is featured as an intensive homogeneous hyperechoic submucosal lesion with marked echo attenuation and without involvement of the mucosa.

CHARACTERISTICS OF TWO－DIMENSIONAL AND COLOR DOPPLER ULTRASONOGRAPHY IN GRAVES＇DISEASE

Color Doppler flow imaging(CDFI) was performed in 24 patients with Graves' disease and the results were compared with those from 8 patients with Hashimoto's disease, 6 patients with simple diffuse goiter, and 15 normal volunteers. All cases were confirmed by clinical, laboratory tests or pathology. CDFI of Graves' disease showed diffuse or localized hypoechoes within the thyroid glands, rich flow signals in the hypoechoes, accelerated flow velocity, and decreased resistance in the superior thyroid arteries(STA) and arteries within the glands. These appearances were characteristic as compared with the controls. The rate of flow in STA of Graves' disease patients was 8 to 10 times higher than that in the normal volunteers. This study demonstrates that CDFI is convenient and efficient as compared with isotope scan and some other laboratory tests, and can be widely used in the diagnosis of Graves' disase.

EVALUATION OF THE DIAGNOSTIC CRITERIA FOR THE LOCALIZATION OF ACQUIRED ARTERIOVENOUS FISTULAS BY COLOR DOPPLER FLOW IMAGING

Objective. To evaluate the diagnostic criteria for the localization of acquired arteriovenous fistulas (AVFs) by color Doppler flow imaging (CDFI) based on the features of their hemodynamic changes. Methods. The shape and hemodynamic changes of involved vessels which could be helpful to localize the sites of fistulas were studied according to the observation of 10 cases of acquired AVFs. Results. The sites of the fistulas could be shown by two dimensional ultrasonography and color flow imaging in 40%and 80%cases, respectively. In all cases, turbulent high velocity flow was present at the sites of the fistulas, low resistant flow was present in the arteries proximal to the fistulas, and artery like flow was detected in the veins. Conclusion. CDFI was accurate for the localization of acquired AVFs, which were mainly localized by their hemodynamic changes shown by pulse Doppler ultrasound.

Ultrasonographic Characteristics of Intraductal Papillary Mucinous Neoplasm of the Pancreas

Objective To analyze the clinical and ultrasonographic imaging features of intraductal papillary mutinous neoplasm （1PMN） of the pancreas. Methods Twelve patients with IPMN underwent surgery between May 2005 and December 2008, including 4 （33.3%） with adenoma and 8 （66.7%） with adenocarcinoma. IPMN was classified preoperatively into 3 types based on sonographic findings of different sites： main duct, branch duct, and combined type. All clinical presentations and ultrasonographic findings of those paticnts were reviewed and the correlation be- tween ultrasonographic hndings and histopathological results was analyzed. Results There were 9 men and 3 women with a mean age of 60.1±9.6 years （range, 32-73）. OF all the 12 patients with IPMN, 9 （75.0%） had experienced some symptoms of epigastric discomfort and/or pain as cell as backache; 7 cases were with medical history of acute pancreatitis, 5 cases with diabetes, 4 cases with elevated CA19-9, and 2 cases with steatorrhea. All lesions of IPMN have been revealed by transabdominal ul- trasonography. The mean diameters of the lesions were 1.4±0.8 cm （range, 0.5-2.0） and 6.3±6.0 cm （range, 2.0-20.0） in adenomas and adenocarcinomas, respectively. And the mean diameters of the main duct in adelnomas and adenocarcinomas were 1.0±0.8 cm and 1.6±1.0 cm, respectively. Among the 4 adnomas, 3 （75.0%） cases were classified as branch type based on sonographic findings, and 2 were demonstrated as mural nodules in which no color signals was detected. Among the 8 adenocarcinomas, 5 （62.5%） cases were classifled as main duct type, and 3 （37.5%） as combined type. In 7 of the 8 adcnocarcinomas, mural nodules were detected within the dilated ducts or cysts of the lesions in which color flow signals were detected. Conclusions Transabdominal ultrasonography can reveal the pancreatic cystic lesions of IPMN as well as dilated pancreatic： ducts. Some characteristics should be noticed as suggesting the possibility of malignancy： clinical symptoms of pancreatic insufficiency, large tumor size, and mural nodules with color Doppler flow signals. Transabdominal ultrasonography could be a useful tool to help diagnose and make appropriate management of IPMN.

High-resolution Hyper-spectral Image Classification with Parts-based Feature and Morphology Profile in Urban Area

High-resolution hyper-spectral image (HHR) provides both detailed structural and spectral information for urban study. However, due to the inherent correlation between spectral bands and within-class variability in the data, the data processing of HHR is a challenging work. In this paper, based on spectral mixture analysis theory, a new stack of parts description features were extracted, and then incorporated with a stack of morphology based spatial features. Partially supervised constrained energy minimization (CEM) and unsupervised nonnegative matrix factorization (NMF) were used to extract the part-features. The joint features were then integrated by SVM classifier. The advantages of this method are the representation of physical composition of the urban area by the parts-features and the show of multi-scale structure information by morphology profiles. Experiments with an airborne hyper-spectral data flightline over the Washington DC Mall were performed, and the performance of the proposed algorithm was evaluated in comparison with well-known nonparametric weighted feature extraction (NWFE) and feature selection method. The results shown that the proposed features-joint scheme consistently outperforms the traditional methods, and so can provide an effective option for processing HHR data in urban area.