公开(公告)号：EP4322063A1

公开(公告)日：2024-02-14

申请号：EP23187083.3

申请日：2023-07-21

申请人： Nuctech Company Limited

发明人： ZHANG, Li ,  SUN, Yunda ,  HU, Zheng ,  FU, Gang ,  LI, Qiang

IPC分类号： G06N3/0475 ,  G06N3/084 ,  G06N3/09

摘要： A method of training an image decomposition model, an apparatus of training an image decomposition model, a method of decomposing an image, an electronic device, and a storage medium are provided, which may be applied to a field of an image processing technology. The method of training the image decomposition model includes: acquiring a training set; inputting a first training image into the first adversarial neural network and inputting a second training image into the second adversarial neural network, so as to determine a first loss function value; inputting a third training image into the first adversarial neural network and the second adversarial neural network respectively, so as to determine a second loss function value; determining a third loss function value according to a comparison result between an acquired fusion image and the third training image, where the fusion image is generated by fusing a generated image of the first adversarial neural network with a generated image of the second adversarial neural network; and adjusting a parameter of the image decomposition model according to at least one of the first to third loss function values, until a training end condition is met, where the first loss function, the second loss function and the third loss function meet a predetermined weight relationship.

公开(公告)号：EP3343442A1

公开(公告)日：2018-07-04

申请号：EP17210936.5

申请日：2017-12-28

申请人： Nuctech Company Limited

发明人： XU, Yanwei ,  HU, Zheng ,  WANG, Yongming ,  GAO, Chuan ,  YU, Weifeng ,  LI, Qiang ,  MA, Yuan ,  GU, Jianping ,  HU, Yu ,  SUN, Shangmin

IPC分类号： G06K9/00

CPC分类号： G06T7/74 ,  G06K9/00771 ,  G06K9/00791 ,  G06K9/6262 ,  G06N3/0472 ,  G06N3/08 ,  G06T3/40 ,  G06T2207/20081 ,  G06T2207/20084 ,  G06T2207/30248

摘要： The present disclosure discloses a method and system for identifying a part of a vehicle and a vehicle inspection system. The method includes: acquiring (S110) a vehicle body image sequence of a vehicle to be identified; reconstructing (S120) the vehicle body by using a first vehicle body reconstruction model generated through a deep learning algorithm and on the basis of the vehicle body image sequence, so as to acquire a vehicle body reconstruction image of the vehicle to be identified; and identifying (S130) a boundary identifier of the vehicle to be identified on the basis of the vehicle body reconstruction image of the vehicle to be identified.

公开(公告)号：EP3115772A1

公开(公告)日：2017-01-11

申请号：EP15875152.9

申请日：2015-12-23

申请人： Tsinghua University ,  Nuctech Company Limited

发明人： CHEN, Zhiqiang ,  ZHANG, Li ,  ZHAO, Ziran ,  LIU, Yaohong ,  GU, Jianping ,  HU, Zheng ,  LI, Qiang

IPC分类号： G01N23/04

CPC分类号： G06F17/30256 ,  G01N23/04 ,  G01N2223/401 ,  G01N2223/631 ,  G01V5/0016 ,  G06F17/30274 ,  G06K9/42 ,  G06K9/6215 ,  G06K2209/09 ,  G06K2209/23 ,  G06T7/001 ,  G06T7/337 ,  G06T2207/10116 ,  G06T2207/30268

摘要： Disclosed is a vehicle inspection method. The method includes: acquiring a transmission image of an inspected vehicle; acquiring a transmission image template of a vehicle model the same as that of the inspected vehicle from a database; performing registration on the transmission image of the inspected vehicle and the transmission image template; determining a difference between a transmission image after the registration and a transmission image template after the registration, to obtain a change area of the transmission image of the vehicle relative to the transmission image template; and processing the change area to determine whether the vehicle carries a suspected object or not. This solution can avoid the problems of a detection loophole and a poor effect of manually determining an image in a conventional manner, and is important to assist the security inspection for small vehicles.

摘要翻译： 公开了一种车辆检查方法。 该方法包括：获取被检查车辆的发送图像; 从数据库中获取与被检查车辆相同的车辆模型的传输图像模板; 对被检查车辆的发送图像和发送图像模板进行登记; 确定登记之后的发送图像与登记之后的发送图像模板之间的差异，以获得相对于发送图像模板的车辆的发送图像的变化区域; 并处理变更区域以确定车辆是否携带可疑对象。 该解决方案可以避免检测漏洞的问题和以常规方式手动确定图像的不良影响，并且对于协助小型车辆的安全检查是重要的。

公开(公告)号：EP3128496A1

公开(公告)日：2017-02-08

申请号：EP15875153.7

申请日：2015-12-23

申请人： Tsinghua University ,  Nuctech Company Limited

发明人： CHEN, Zhiqiang ,  ZHANG, Li ,  LI, Jianmin ,  ZHAO, Ziran ,  LIU, Yaohong ,  LI, Qiang ,  HU, Zheng ,  GU, Jianping ,  LI, Ying

IPC分类号： G08G1/017

CPC分类号： G08G1/054 ,  G01V5/0008 ,  G01V5/0016 ,  G06K9/4604 ,  G06K9/4642 ,  G08G1/015 ,  G08G1/017

摘要： Disclosed is a vehicle identification method and system. The method includes: acquiring appearance information of an inspected vehicle; obtaining external features of the vehicle based on the appearance information; acquiring a transmission image of the vehicle and obtaining internal features of the vehicle from the transmission image; forming descriptions on the vehicle at least based on the external features and the internal features; and determining a vehicle model of the vehicle from a vehicle model databased by utilizing the descriptions. This method merges various types of modality information, especially introducing the transmission image, and combines the internal structure information with the appearance information, so that the present disclosure can identify a vehicle model more practically.

摘要翻译： 公开了车辆识别方法和系统。 该方法包括：获取被检查车辆的外观信息; 基于外观信息获取车辆的外部特征; 从所述发送图像获取所述车辆的发送图像并获得所述车辆的内部特征; 至少基于外部特征和内部特征对车辆形成描述; 以及通过利用所述描述从数据库的车辆模型确定车辆的车辆模型。 该方法合并各种形态信息，特别是引入传输图像，并将内部结构信息与外观信息相结合，使得本公开可以更实际地识别车辆模型。