CN106033614B - A kind of mobile camera motion object detection method under strong parallax - Google Patents

A kind of mobile camera motion object detection method under strong parallax Download PDF

Info

Publication number
CN106033614B
CN106033614B CN201510124336.3A CN201510124336A CN106033614B CN 106033614 B CN106033614 B CN 106033614B CN 201510124336 A CN201510124336 A CN 201510124336A CN 106033614 B CN106033614 B CN 106033614B
Authority
CN
China
Prior art keywords
camera
depth
image
constraint equation
point
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201510124336.3A
Other languages
Chinese (zh)
Other versions
CN106033614A (en
Inventor
顾国华
丁祺
孔筱芳
徐富元
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Wuxi Nust New Energy Electric Vehicle Technology Development Co ltd
Original Assignee
Nanjing University of Science and Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nanjing University of Science and Technology filed Critical Nanjing University of Science and Technology
Priority to CN201510124336.3A priority Critical patent/CN106033614B/en
Publication of CN106033614A publication Critical patent/CN106033614A/en
Application granted granted Critical
Publication of CN106033614B publication Critical patent/CN106033614B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Landscapes

  • Image Processing (AREA)
  • Measurement Of Optical Distance (AREA)
  • Image Analysis (AREA)

Abstract

本发明提出一种强视差下的移动相机运动目标检测方法。用旋转矩阵R和平移矩阵t描述两帧之间的相机运动,获得第二帧图像上的点与世界坐标系的关系,从而获得深度约束方程;将第二帧图像上所有匹配点的坐标和对应的深度信息带入深度约束方程,获得深度约束方程的Rt矩阵的最优解;利用深度约束方程计算的图像点坐标与对应的深度信息相乘;将该乘积和利用深度约束方程估计出来的结果分别作为两个三维坐标点,计算该两个三维坐标点之间的距离,并将距离归一化,若归一化距离大于设定的阈值,则判断该距离对应的点为运动目标,否则,该距离对应的点为背景。本发明能够解决在车载手持等移动平台下,三维场景会产生视差的问题,对运动目标进行实时高效的检测。

The present invention proposes a moving camera moving target detection method under strong parallax. Use the rotation matrix R and the translation matrix t to describe the camera motion between the two frames, and obtain the relationship between the points on the second frame image and the world coordinate system, so as to obtain the depth constraint equation; sum the coordinates of all matching points on the second frame image and The corresponding depth information is brought into the depth constraint equation, and the optimal solution of the Rt matrix of the depth constraint equation is obtained; the image point coordinates calculated by the depth constraint equation are multiplied by the corresponding depth information; the product sum is estimated by the depth constraint equation. The results are taken as two three-dimensional coordinate points respectively, the distance between the two three-dimensional coordinate points is calculated, and the distance is normalized. If the normalized distance is greater than the set threshold, the point corresponding to the distance is judged to be a moving target. Otherwise, the point corresponding to this distance is the background. The present invention can solve the problem of parallax generated in three-dimensional scenes on mobile platforms such as vehicle-mounted handhelds, and can perform real-time and efficient detection of moving objects.

Description

A kind of mobile camera motion object detection method under strong parallax
Technical field
The present invention is to belong to Digital Image Processing and area of pattern recognition, especially belongs to the target based on image procossing and knows Not, field is tracked, and in particular to the mobile camera motion object detection method under a kind of strong parallax.
Background technique
Automatic Detection for Moving Target is the key technology in Target Acquisition And Track System, be subsequent track association, The technologies such as target identification, tracking provide preliminary information.In traditional moving object detection system, video camera is usually quiet Only, it has a wide range of applications in the systems such as video monitoring, Vehicle Detection, however under the mobile platforms such as vehicle-mounted, hand-held, by In the movement of camera, three-dimensional scenic can lead to the problem of parallax, just need effectively to estimate scene parallax at this time.
Conventional method is usually the Epipolar geometry relationship utilized under motion platform between image, such as plane homography, basis Matrix and trifocal tensor etc. compensate or estimate the movement of video camera with this.Wherein, homography matrix is using on same plane Mapping relations between matching double points, situations such as unmanned plane can be almost ignored especially suitable for scene parallax;Basic square Battle array detects moving target using the emitter-base bandgap grading geometrical relationship between images match point pair;The multi-view image that trifocal tensor utilizes by Two views increase to three-view diagram, are not the mathematical quantity of the inherent projective geometry relationship shifted with scene structure between representative image Trifocal tensor is become by fundamental matrix, but is calculated increasingly complex.
But the above conventional method needs geological information known referring to calibration object, the scaling method based on active vision needs The azimuth information of video camera is provided at any time using the accurately hardware platforms such as holder, it is high to hardware requirement, it is not suitable for just Camcorder occasion;Method based on self-calibration such as fundamental matrix, three visual angle tensors, although to calibration scene and calibrating instrument It is of less demanding, but often heavy workload, poor robustness, accuracy be not high.
Summary of the invention
The purpose of the present invention is to provide the mobile camera motion object detection method under a kind of strong parallax, this method is combined The grayscale information and depth information of image, establish depth constraints equation, whether meet depth constraints by the point judged on image Equation carries out moving object detection, is able to solve under vehicle-mounted hand-held equal mobile platforms, three-dimensional scenic can generate asking for parallax Topic carries out the detection of real-time high-efficiency to moving target.
In order to solve the above-mentioned technical problem, the present invention proposes the mobile camera motion target detection side under a kind of strong parallax Method, comprising the following steps:
Step 1: two field pictures are matched, and extract the depth information of two field pictures;Camera is demarcated, is obtained The inner parameter of camera;
Step 2: position when using camera shooting first frame image converses on first frame image as world coordinate system The corresponding world coordinate system coordinate of point coordinate;
Step 3: the camera motion between two frames is described with spin matrix R and translation matrix t, is obtained on the second frame image Point and world coordinate system relationship, to obtain depth constraints equation;
Step 4: the coordinate of all match points on the second frame image and corresponding depth information are substituted into depth constraints side Journey obtains the optimal solution of the Rt matrix of depth constraints equation;
Step 5: it is multiplied using the picture point coordinate of depth constraints equation calculation with corresponding depth information;By the product With the result that is estimated using depth constraints equation respectively as two three-dimensional coordinate points, calculate two three-dimensional coordinate points it Between distance if normalized cumulant is greater than the threshold value of setting, judge that this apart from corresponding point is to move and by range normalization Target, otherwise, this is background apart from corresponding point.
Further, in step 3, shown in the depth constraints equation such as formula (1):
In formula (1), (u2,v2) indicate front and back two field pictures match point image coordinate, u2、v2Pixel is respectively indicated to exist Abscissa and ordinate on a later frame image;Zc2The match point for respectively indicating front and back two field pictures is believed relative to the depth of camera Breath;fx、fy, s, m, n be parameter in camera internal parameter K respectively, wherein fx, fyRespectively with x, the pixel dimension in the direction y The camera of expression is along x, the focal length in the direction y;S is warp parameters;(m, n) is center point coordinate in camera imaging plane;Xw、Yw、Zw To put the homogeneous coordinates in corresponding world coordinate system on first frame image;r1,r2...r9And t1,t2,t3It is spin matrix respectively With the parameter of translation matrix, and spin matrixTranslation matrix
Further, in step 4, the optimal solution for meeting the Rt matrix of depth constraints equation is solved using least square method.
Further, between two three-dimensional coordinate points described in step 5 shown in the calculation such as formula (2) of distance,
In formula (2), Distance (i) is distance between two three-dimensional coordinate points,For using deeply The product of angular coordinate and corresponding depth information on the second frame image that degree constraint equation calculates;(Zc2u2,Zc2v2,Zc2) For the product of angular coordinate and corresponding depth information on the second frame image.
Further, normalized threshold is chosen between 0 to 1.
Compared with prior art, the present invention its remarkable advantage is, the present invention utilizes two dimensional image coordinate system to three-dimensional generation Relationship between boundary's coordinate system according to image depth information and photography geometrical principle, and combines in front and back two field pictures and camera Portion's parameter proposes depth constraints equation, and carries out moving object detection using depth constraints equation, compared to traditional movement Camera motion object detection method has the advantage that the moving target detecting method under (1) this strong parallax, due to combining The grayscale information and depth information of image can be eliminated due to the mobile influence generated to moving object detection of camera, reject view Difference reduces false alarm rate, is applicable not only to the take photo by plane Deng negligible scene of parallaxes, and in the vehicle for being frequently used for strong parallax environment It carries, also have good effect in handheld device;(2) realization of object detection method under the new moving camera of one kind is proposed Depth constraints equation is not only applied in the frame of target detection by journey, this method, but also can be in a variety of depth detectors It is realized on platform;(3) algorithm is simple, fast speed, lower to hardware platform requirements, and moving object detection rate is high.
Detailed description of the invention
Fig. 1 is the method for the present invention flow chart.
Fig. 2 be in present invention experiment using apart from the closer small vehicle model of camera as the source of strong parallax, camera to In the case where left and simultaneously rotation counterclockwise, when tank model moves from left to right, the movement of continuous three frame video image Object detection results.
Fig. 3 is the verification and measurement ratio comparison schematic diagram of the method for the present invention Yu epipolar-line constraint and homography constrained procedure.
Fig. 4 is the accurate rate comparison schematic diagram of the method for the present invention Yu epipolar-line constraint and homography constrained procedure.
Specific embodiment
In conjunction with Fig. 1, mobile camera motion object detection method under the strong parallax of the present invention, comprising the following specific steps
Step 1: original image pretreatment: firstly, matched to the pixel of front and back two field pictures, mature KLT with Track algorithm is due to having many advantages, such as that speed is fast, precision is high, anti-noise ability is strong, using KLT track algorithm to two frame figure of front and back As pixel is matched, the image coordinate (u of two field pictures match point is obtained1,v1), (u2,v2), wherein u1、v1It respectively indicates Abscissa and ordinate of the pixel on previous frame image, u2、v2Respectively indicate abscissa of the pixel on a later frame image With ordinate;Secondly, opposite using such as Kinect camera, laser radar even depth detector, the match point for obtaining two field pictures In the depth information Z of camerac1, Zc2, i.e. camera range-to-go;Finally, carrying out camera mark using Zhang Zhengyou camera calibration method It is fixed, obtain the inner parameter of cameraWherein, fx, fyRespectively indicated with x, the pixel dimension in the direction y Camera is along x, the focal length in the direction y;S is warp parameters;(m, n) is center point coordinate in camera imaging plane.
Step 2: the point coordinate under world coordinate system solves: using the corresponding camera coordinates system of first frame image as world coordinates System, according to photography geometrical principle, obtain the point on first frame image to world coordinate system corresponding relationship, as shown in formula (1):
Wherein, (u1,v1, 1) and it is the homogeneous coordinates put on first frame image;(Xw,Yw,Zw, 1) and it is that the corresponding world of the point is sat Homogeneous coordinates in mark system;Zc1For depth information of this under camera coordinates system;I is unit matrix.Is obtained using formula (1) Point (u on one frame image1,v1) corresponding world coordinate system coordinate (Xw,Yw,Zw, 1), as shown in formula (2):
Step 3: the foundation of depth constraints equation: since camera itself is moving, when shooting the second frame image, the position of camera It sets and is changed relative to the position of shooting first frame, described between two field pictures with spin matrix R and translation matrix t Camera motion, therefore, shown in the relationship such as formula (3) of point and world coordinate system on the second frame image:
Wherein, point (u2,v2) it is point (u1,v1) match point coordinate on the second frame image;Zc2It is the point in Current camera Depth information under coordinate system, willSubstitution formula (3) (wherein r1,r2...r9And t1,t2,t3Point It is not the parameter of spin matrix and translation matrix), and it is organized into the form of AX=B, obtain depth constraints equation:
The solution of step 4:Rt matrix: the coordinate of all match points on the second frame image and corresponding depth information are substituted into Depth constraints equation solves the optimal solution for meeting the Rt matrix of depth constraints equation using least square method.
Step 5: the moving object detection based on depth constraints equation: firstly, using step 4 calculate Rt matrix it is optimal Solution calculates the value of depth constraints equation right side of the equal sign, i.e., angle point on the second frame image come out using depth constraints equation calculation The product of coordinate and corresponding depth informationSecondly, by angular coordinate on the second frame image with it is corresponding Product (the Z of depth informationc2u2,Zc2v2,Zc2) and using depth constraints equation calculation come out productIt sees Two three-dimensional coordinate points are done, the two three-dimensional coordinate points are substituted into formula (5), calculate the distance between two o'clock:
Finally, Distance (i) normalization is obtained Distance_normal (i) and is judged, if normalized cumulant Greater than the threshold value of setting, then judge this apart from corresponding point for moving target, conversely, then judge this apart from corresponding point for back Scape, to realize moving object detection.Normalized threshold is chosen between 0 to 1, according to theory deduction and experiment experience, is chosen The size of normalized threshold should be positively correlated with camera motion speed, negative moves speed about target about camera frame frequency, negative Degree.If normalized threshold is chosen excessive, it may appear that the missing inspection of moving target;If normalized threshold is chosen too small, it may appear that false-alarm, General normalized threshold takes 0.7 or so.
Beneficial effects of the present invention can be further illustrated by following experiment:
For the embodiment of the present invention using Matlab2012b as experiment porch, experiment is 640 × 480 using image size, and frame frequency is The Xtion three-dimension sensor of 30Hz carries out video capture and extraction of depth information.Make using apart from the closer small vehicle model of camera For the source of strong parallax, in the case where camera is to left and rotates counterclockwise simultaneously, tank model moves from left to right.Such as Shown in Fig. 1, specific steps are as follows:
(1), according to the present invention described in step 1, the original image obtained to Xtion three-dimension sensor is pre-processed, false If moving target can be by Based on Feature Points, firstly, extracting the angular coordinate of two field pictures with Harris algorithm;Secondly, using KLT track algorithm matches front and back two field pictures angle point, obtains the image coordinate (u of two field pictures match point1,v1), (u2, v2), in Fig. 2 (a), (d), in (g) shown in white point;Again, the matching of two field pictures is obtained using Xtion three-dimension sensor Depth information Z of the point relative to camerac1, Zc2, in Fig. 2 (b), (e), shown in (h), color shows that distance is closer more deeply feeling;Most Afterwards, camera calibration is carried out using Zhang Zhengyou camera calibration method, obtains the inner parameter K of camera.
(2), according to the present invention described in step 2, the match point coordinate under world coordinate system is solved: according to formula (1):
Obtain the corresponding world coordinate system coordinate (X of all matching angle points on first frame imagew,Yw,Zw, 1), such as formula (2) It is shown:
(3), according to the present invention described in step 3, establish depth constraints equation: angle point and the world on the second frame image are sat Shown in the relationship such as formula (3) for marking system:
It willSubstitution formula (3), and it is organized into the form of AX=B, obtain depth constraints side Journey:
(4), according to the present invention described in step 4, the Rt matrix in depth constraints equation is solved: by institute on the second frame image The coordinate and corresponding depth information for having matching angle point substitute into depth constraints equation, are solved using least square method and meet depth about The optimal solution of the Rt matrix of Shu Fangcheng.
(5), according to the present invention described in step 5, moving object detection is carried out using depth constraints equation: firstly, utilizing formula (4) angular coordinate that obtains using depth constraints equation and the product of corresponding depth information are calculatedIts It is secondary, by the product (Z of angular coordinate and corresponding depth information on the second frame imagec2u2,Zc2v2,Zc2) and utilization depth constraints side What journey estimatedRegard two three-dimensional coordinate points as, the two three-dimensional coordinate points is substituted into formula (5), meter Calculate the distance between two o'clock:
Finally, Distance (i) normalization is obtained Distance_normal (i) and is judged, if normalized cumulant Greater than the threshold value of setting, then judge this apart from corresponding point for moving target, conversely, then judge this apart from corresponding point for back Scape, to realize moving object detection, moving target is as shown in black color dots in (c) in Fig. 2, (f), (i).Wherein, threshold is normalized Value is chosen for Th=0.9.(c) in Fig. 2, (f), (i) are respectively the testing result of three frame images, what the point of black indicated to detect Motion corner point., it is evident that the motion corner point in three frame images all is correctly detected out by the present invention, and strong parallactic angle Point is not detected among out.
20 frame images are randomly selected, hand labeled goes out moving target angle point, is denoted asWherein t indicates t frame;To selection 20 frame images out carry out moving object detection, and the angle point for the moving target that every frame calculates is denoted asIt defines D (t) Indicate the verification and measurement ratio of t frame:
Defining P (t) indicates the accurate rate of t frame:
Wherein, N indicates the number at set midpoint.The verification and measurement ratio that algorithm is described with formula (6), that is, what is detected is movement mesh Target angle point accounts for the ratio of real motion Corner;The accurate rate that algorithm is described with formula (7), that is, what is detected is movement mesh Target angle point accounts for the ratio (being inversely proportional to false alarm rate) of all angle points that detected.The higher expression of the value of verification and measurement ratio and accurate rate should The performance of algorithm is better.The method of the present invention and epipolar-line constraint algorithm and homography bounding algorithm are compared, by 20 frame images Algorithm verification and measurement ratio and algorithm accurate rate are depicted as line chart, as shown in Figure 3, Figure 4.From figure 3, it can be seen that the method for the present invention Verification and measurement ratio is apparently higher than epipolar-line constraint algorithm, slightly above homography bounding algorithm, and apparent missing inspection occurs in epipolar-line constraint algorithm; Figure 4, it is seen that the accurate rate of the method for the present invention is apparently higher than homography bounding algorithm, slightly above epipolar-line constraint algorithm, There is apparent false-alarm in homography algorithm.The method of the present invention has good performance on verification and measurement ratio and accurate rate.
In conclusion the grayscale information and depth information of present invention combination image, establish depth constraints equation, pass through judgement Whether the point on image meets depth constraints equation to carry out moving object detection.Experiment show this method not only verification and measurement ratio compared with Height, and the depth constraints equation proposed can effectively remove the influence because of the mobile parallax generated of camera to target detection, It can be not only used for that parallax is negligible to take photo by plane, monitor, and also have for the equipment such as vehicle-mounted, hand-held very strong practical Property.

Claims (3)

1.一种强视差下的移动相机运动目标检测方法,其特征在于,包括以下步骤:1. a moving camera moving target detection method under a strong parallax, is characterized in that, comprises the following steps: 步骤一:将两帧图像进行匹配,提取两帧图像的深度信息;对相机进行标定,获取相机的内部参数;Step 1: Match the two frames of images to extract the depth information of the two frames of images; calibrate the camera to obtain the internal parameters of the camera; 步骤二:以相机拍摄第一帧图像时的位置为世界坐标系,换算出第一帧图像上的点坐标对应的世界坐标系坐标;Step 2: Take the position of the camera when the first frame of image is captured as the world coordinate system, and convert the coordinates of the world coordinate system corresponding to the point coordinates on the first frame of image; 步骤三:用旋转矩阵R和平移矩阵t描述两帧之间的相机运动,获得第二帧图像上的点与世界坐标系的关系,从而获得深度约束方程;Step 3: Use the rotation matrix R and the translation matrix t to describe the camera movement between the two frames, and obtain the relationship between the point on the second frame image and the world coordinate system, thereby obtaining the depth constraint equation; 步骤四:将第二帧图像上所有匹配点的坐标和对应的深度信息代入深度约束方程,获得深度约束方程的Rt矩阵的最优解;Step 4: Substitute the coordinates of all matching points and the corresponding depth information on the second frame image into the depth constraint equation to obtain the optimal solution of the Rt matrix of the depth constraint equation; 步骤五:利用深度约束方程计算的图像点坐标与对应的深度信息相乘;将该相乘乘积和利用深度约束方程估计出来的结果分别作为两个三维坐标点,计算该两个三维坐标点之间的距离,并将距离归一化,若归一化距离大于设定的阈值,则判断该距离对应的点为运动目标,否则,该距离对应的点为背景。Step 5: Multiply the image point coordinates calculated by the depth constraint equation with the corresponding depth information; use the multiplication product and the result estimated by the depth constraint equation as two three-dimensional coordinate points respectively, and calculate the difference between the two three-dimensional coordinate points. If the normalized distance is greater than the set threshold, the point corresponding to the distance is judged to be a moving target, otherwise, the point corresponding to the distance is the background. 2.如权利要求1所述强视差下的移动相机运动目标检测方法,其特征在于,步骤三中,所述深度约束方程如式(1)所示:2. the moving camera moving target detection method under strong parallax as claimed in claim 1, is characterized in that, in step 3, described depth constraint equation is as shown in formula (1): 式(1)中,(u2,v2)表示前后两帧图像匹配点的图像坐标,u2、v2分别表示像素点在后一帧图像上的横坐标与纵坐标;Zc2表示前后两帧图像的匹配点相对于相机的深度信息;fx、fy、s、m、n分别是相机内部参数K中的参数,其中,fx,fy分别为以x,y方向的像素量纲表示的相机沿x,y方向的焦距;s为扭曲参数;(m,n)是相机成像平面上中心点坐标;Xw、Yw、Zw为第一帧图像上点对应的世界坐标系中的齐次坐标;r1,r2...r9和t1,t2,t3分别是旋转矩阵和平移矩阵的参数,且旋转矩阵平移矩阵 In formula (1), (u 2 , v 2 ) represent the image coordinates of the matching points of the two frames before and after the image, u 2 and v 2 respectively represent the abscissa and ordinate of the pixel on the next frame of image; Z c2 represents the front and back The matching points of the two frames of images are relative to the depth information of the camera; f x , f y , s, m, and n are the parameters in the camera's internal parameters K, respectively, where f x and f y are the pixels in the x and y directions, respectively The focal length of the camera along the x, y directions represented by the dimension; s is the distortion parameter; (m, n) is the coordinate of the center point on the camera imaging plane; X w , Y w , Z w are the world corresponding to the point on the first frame image Homogeneous coordinates in the coordinate system; r 1 , r 2 ... r 9 and t 1 , t 2 , t 3 are the parameters of the rotation matrix and the translation matrix, respectively, and the rotation matrix translation matrix 3.如权利要求1所述强视差下的移动相机运动目标检测方法,其特征在于,步骤四中,利用最小二乘法解得符合深度约束方程的Rt矩阵的最优解。3 . The method for detecting a moving camera moving target under strong parallax according to claim 1 , wherein, in step 4, the optimal solution of the Rt matrix conforming to the depth constraint equation is obtained by using the least squares method. 4 .
CN201510124336.3A 2015-03-20 2015-03-20 A kind of mobile camera motion object detection method under strong parallax Active CN106033614B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510124336.3A CN106033614B (en) 2015-03-20 2015-03-20 A kind of mobile camera motion object detection method under strong parallax

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510124336.3A CN106033614B (en) 2015-03-20 2015-03-20 A kind of mobile camera motion object detection method under strong parallax

Publications (2)

Publication Number Publication Date
CN106033614A CN106033614A (en) 2016-10-19
CN106033614B true CN106033614B (en) 2019-01-04

Family

ID=57149206

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510124336.3A Active CN106033614B (en) 2015-03-20 2015-03-20 A kind of mobile camera motion object detection method under strong parallax

Country Status (1)

Country Link
CN (1) CN106033614B (en)

Families Citing this family (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108759665B (en) * 2018-05-25 2021-04-27 哈尔滨工业大学 Spatial target three-dimensional reconstruction precision analysis method based on coordinate transformation
CN109255803B (en) * 2018-08-24 2022-04-12 长安大学 Displacement calculation method of moving target based on displacement heuristic
CN109495641B (en) * 2018-10-24 2021-01-08 维沃移动通信有限公司 Reminding method and mobile terminal
CN109462746B (en) * 2018-10-26 2020-11-06 北京双髻鲨科技有限公司 Image stabilization method and device
WO2020237441A1 (en) * 2019-05-24 2020-12-03 Guangdong Oppo Mobile Telecommunications Corp., Ltd. User equipment and method of oblique view correction
CN110798634B (en) * 2019-11-28 2020-10-09 东北大学 Image adaptive synthesis method, device and computer readable storage medium
GB2593227A (en) * 2020-03-20 2021-09-22 Sony Corp Image processing method and apparatus
CN114596333A (en) * 2020-12-03 2022-06-07 北京华航无线电测量研究所 A Target Tracking Method Applicable to Vehicle Environment
CN112950679A (en) * 2021-01-28 2021-06-11 武汉博雅弘拓科技有限公司 Automatic redundant point cloud eliminating method and system based on photogrammetry technology

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080089556A1 (en) * 2006-06-20 2008-04-17 Garbis Salgian Moving Target Detection in the Presence of Parallax
CN103729857A (en) * 2013-12-09 2014-04-16 南京理工大学 Moving target detection method under movable camera based on secondary compensation

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080089556A1 (en) * 2006-06-20 2008-04-17 Garbis Salgian Moving Target Detection in the Presence of Parallax
CN103729857A (en) * 2013-12-09 2014-04-16 南京理工大学 Moving target detection method under movable camera based on secondary compensation

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Moving object detection and camera motion analysis from moving camera;Andrey Vavilin et al;《38th Annual Conference on IEEE Industrial Electronics Society》;20121224;4171-4176
移动平台下运动目标检测的MOF算法;陈海欣 等;《红外与激光工程》;20140228;第43卷(第2期);620-624

Also Published As

Publication number Publication date
CN106033614A (en) 2016-10-19

Similar Documents

Publication Publication Date Title
CN106033614B (en) A kind of mobile camera motion object detection method under strong parallax
CN105550670B (en) A kind of target object dynamically track and measurement and positioning method
CN106251305B (en) A real-time electronic image stabilization method based on inertial measurement unit IMU
US11521311B1 (en) Collaborative disparity decomposition
WO2019100933A1 (en) Method, device and system for three-dimensional measurement
CN104463880B (en) A kind of RGB D image acquiring methods
CN108230392A (en) A kind of dysopia analyte detection false-alarm elimination method based on IMU
US20090167843A1 (en) Two pass approach to three dimensional Reconstruction
Islam et al. Stereo vision-based 3D positioning and tracking
CN102609941A (en) Three-dimensional registering method based on ToF (Time-of-Flight) depth camera
JP2020112438A (en) Sea level measurement system, sea level measurement method and sea level measurement program
US11212510B1 (en) Multi-camera 3D content creation
CN107560592A (en) A kind of precision ranging method for optronic tracker linkage target
CN113724335B (en) Three-dimensional target positioning method and system based on monocular camera
CN106203429A (en) Based on the shelter target detection method under binocular stereo vision complex background
CN113240749B (en) A long-distance dual target determination and ranging method for UAV recovery on offshore ship platforms
CN104504691B (en) Camera position and posture measuring method on basis of low-rank textures
CN108090930A (en) Barrier vision detection system and method based on binocular solid camera
Zhang et al. High quality depth maps from stereo matching and ToF camera
CN115880643B (en) A method and device for monitoring social distance based on target detection algorithm
Thangarajah et al. Vision-based registration for augmented reality-a short survey
Kollmitzer Object detection and measurement using stereo images
CN116151320A (en) Visual odometer method and device for resisting dynamic target interference
Ye et al. Iterative Closest Point Algorithm Based on Point Cloud Curvature and Density Characteristics
Diskin et al. UAS exploitation by 3D reconstruction using monocular vision

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
TR01 Transfer of patent right

Effective date of registration: 20210917

Address after: Room 329, building A2, Nanjing Science and Technology Park, 129-3 Guanghua Road, Qinhuai District, Nanjing, Jiangsu 210008

Patentee after: TECHNOLOGY TRANSFER CENTER CO LTD NANJING University OF SCIENCE AND TECHNOLOGY

Address before: 210094 No. 200, Xiaolingwei, Jiangsu, Nanjing

Patentee before: NANJING University OF SCIENCE AND TECHNOLOGY

TR01 Transfer of patent right
TR01 Transfer of patent right

Effective date of registration: 20211210

Address after: 214104 room 402, 4th floor, block a, Xidong chuangrong building, No. 78, Danshan Road, anzhen street, Xishan District, Wuxi City, Jiangsu Province

Patentee after: WUXI NUST NEW ENERGY ELECTRIC VEHICLE TECHNOLOGY DEVELOPMENT CO.,LTD.

Address before: Room 329, building A2, Nanjing Science and Technology Park, 129-3 Guanghua Road, Qinhuai District, Nanjing, Jiangsu 210008

Patentee before: TECHNOLOGY TRANSFER CENTER CO.,LTD NANJING University OF SCIENCE AND TECHNOLOGY

TR01 Transfer of patent right