CN102881025A - Method for detecting multiple moving targets - Google Patents

Method for detecting multiple moving targets Download PDF

Info

Publication number
CN102881025A
CN102881025A CN2012103426717A CN201210342671A CN102881025A CN 102881025 A CN102881025 A CN 102881025A CN 2012103426717 A CN2012103426717 A CN 2012103426717A CN 201210342671 A CN201210342671 A CN 201210342671A CN 102881025 A CN102881025 A CN 102881025A
Authority
CN
China
Prior art keywords
pixel
moving
moving object
target
potential
Prior art date
Application number
CN2012103426717A
Other languages
Chinese (zh)
Inventor
修春波
Original Assignee
天津工业大学
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 天津工业大学 filed Critical 天津工业大学
Priority to CN2012103426717A priority Critical patent/CN102881025A/en
Publication of CN102881025A publication Critical patent/CN102881025A/en

Links

Abstract

The invention belongs to the field of target identification and target tracking, and in particular relates to a method for detecting multiple moving targets. In a fixed scene, when a moving object exists, two adjacent acquired frames of images are different, and a moving target can be extracted by differential calculation for adjacent images; a pixel point of the extracted moving target is used as a central point of a potential function; a geopotential graph of the scene is constructed; and a maximum value point of the geopotential graph is determined as a central position of the moving target. The method can be applied to video monitoring systems.

Description

—种多运动目标的检测方法 - kind of moving target detection and more

技术领域 FIELD

[0001] 本发明属于目标识别与目标跟踪领域,涉及一种多运动目标的检测方法,特别涉及一种采用帧差法与势函数法相结合的多运动目标检测方法。 [0001] The present invention belongs to the field of object recognition and object tracking, to a plurality of detecting moving target, particularly to a multi-frame difference moving object detection method in combination with potential function of wears.

背景技术 Background technique

[0002] 运动目标的检测一直是机器视觉和图像处理领域的重要研究内容。 [0002] moving target detection has been an important research machine vision and image processing fields. 在监控系统、安防系统、军事领域等方面有重要的应用价值。 There is great value in terms of monitoring systems, security systems, military and other fields. 常用的目标检测方法有背景差分法、光流法、帧差法等。 The target detection method used the background difference method, optical flow method, a frame difference method and the like. 背景差分法能完整、快速地分割出运动对象。 Background subtraction can complete quickly segment the moving object. 不足之处是易受光线变化的影响。 The downside is susceptible to changes in light. 光流法能检测独立运动的对象,可用于摄像头运动的情况,但计算复杂耗时,很难实时检测。 Optical flow method can be detected independent motion, can be used if the camera movement, complex and time consuming but the calculation, hard real-time detection. 帧差法受光线变化影响较小,简单快速,但不能分割出完整的运动对象,需进一步运用目标分割算法。 Frame difference less affected by law changes in light, simple and fast, but not a complete division of a moving object, the need to further the use of target segmentation algorithm. 另外,由于一般的目标识别系统对目标的检测有实时性的要求,运算量较大的算法很难满足这一要求。 Further, since the general object recognition system real-time requirements of the detection target, a large amount of computation algorithm is difficult to meet this requirement.

[0003] 因此设计一种运算量较小、可靠性较高的多运动目标检测方法具有重要的应用价值。 [0003] An operation amount is small so designed, highly reliable detection of multiple moving targets has important application value.

发明内容 SUMMARY

[0004] 本发明所要解决的技术问题是,设计一种运动目标检测方法,能够实现对连续图像中多运动目标的检测。 [0004] The present invention solves the technical problem is to design a method for moving object detection, detection can be achieved successive images of multiple moving targets.

[0005] 本发明所采用的技术方案是:一种多运动目标检测方法,采用帧差法进行运动目标的提取,利用势函数法进行目标的定位。 The technical proposal [0005] The present invention is: a multiple moving object detection using frame difference extracted moving target object using the positioning potential function. 在一个固定的场景中,当存在运动物体时,采集到的相邻两帧图像之间会存在一定的差异,通过对相邻图像的差分运算可提取出运动目标。 There will be some differences between in a stationary scenario, when there is a moving object, acquired two adjacent frame images, by calculating the difference of adjacent images can be extracted moving target. 以提取到的运动目标的像素点为势函数的中心点,建立场景的位势图,位势图的极大值点确定为运动目标的中心位置。 To pixels of the extracted moving object as the center point of the potential function, establishing a potential map scene, FIG potential site maxima is determined as the center position of the moving object.

[0006] 本发明的目的在于提出一种运动目标检测方法,采用帧差法与势函数法相结合的方式实现运动目标的检测与定位,提高运动目标检测的准确性和有效性。 [0006] The object of the present invention is to provide a moving object detection, by way of frame difference method and the Combination of the potential function and achieve the detection of moving targets, to improve the accuracy and efficiency of moving object detection.

附图说明 BRIEF DESCRIPTION

[0007] 图I序列图像视频图。 [0007] FIG video sequence of images I of FIG.

[0008] 图2运动目标检测结果。 [0008] FIG. 2 moving object detection result.

具体实施方式 Detailed ways

[0009] 下面结合实施例和附图对本发明作进一步详细说明。 [0009] The following Examples and the accompanying drawings of the present invention is described in further detail.

[0010] 本发明采用帧差法进行运动目标的检测。 [0010] The present invention using a frame difference method for the detection of moving targets. SfV1(LyhfkO^y)和fk+1(x,y)分别为k-Ι时刻、k时刻和k+Ι时刻获得的三帧图像在坐标(x,y)处的像素灰度值,对其中相邻两帧图像进行差分运算,并与阈值T进行比较,插值小于等于阈值的像素点确定为背景像素点,用O表示;插值大于阈值的像素点确定为运动目标像素点,用I表示。 SfV1 (LyhfkO ^ y) and fk + 1 (x, y) are three image k-Ι time, k and time k + Ι time obtained pixel grayscale values ​​in the coordinate (x, y) at, for which two frame images adjacent differential operation, and the threshold value T is compared, the interpolation less pixel threshold value is determined as a background pixel is represented by O; pixel interpolation greater than a threshold determined as a moving object pixel, denoted by I. 其公式如下: The formula is as follows:

Figure CN102881025AD00041

[0013] 为了检测出准确的运动目标,对上述两个公式所求得的相邻帧差值后的二值图像F_(x,y)和F+(x,y)进行“与”操作,提取出运动目标的像素位置,即: [0013] In order to accurately detect the moving object, the binary image F_ (x, y) after a difference between the two adjacent frames obtained formulas and F + (x, y) a "and" operation, extracting pixel position of moving targets, namely:

[0014] F (x, y) = F_ (x, y) &F+ (x, y) (3) [0014] F (x, y) = F_ (x, y) & F + (x, y) (3)

[0015] 由上式获得的二值图像F(x, y)中,像素值为“I”的像素点为运动目标像素点,像素值为“O”的像素点为背景像素点。 [0015] obtained by the above formula binary image F (x, y), the pixel value "I" of the pixel as a moving target pixel, a pixel value "O" of the pixel as a background pixel point.

[0016] 通常情况下,上述获得的运动目标像素点的连通性不好,不容易实现运动目标的准确定位。 [0016] Typically, connectivity moving target pixel is not obtained above, it is not easy to achieve accurate positioning of the moving target. 为了能够实现运动目标的有效定位,本发明将上述所求得的二值图像F(x,y)中的运动目标像素点确定为特征点。 In order to achieve effective positioning of the moving object, the present invention will target pixel motion calculated above binary image F (x, y) is determined as the feature point. 设特征点集为^{^0^,5^)},1=1,2,...,1。 Set the feature point set is {^ 0 ^ ^, 5 ^)}, 1 = 1,2, ..., 1. 其中,M为运动目标像素点的总数。 Wherein, M is the total number of pixels of the moving target. 以特征点为中心,确定M个位势函数如下: Feature point as the center, to determine the potential function of M as follows:

[0017] K(v, Vi) = exp[_ α II V-V11 |2],i = 1,2, · · · , M (4) [0017] K (v, Vi) = exp [_ α II V-V11 | 2], i = 1,2, · · ·, M (4)

[0018] 将上述M个位势函数进行求和,获得整幅图像的位势图: [0018] The potential function of the M summed to obtain whole potential map image:

Figure CN102881025AD00042

[0020] 由于同一运动目标的特征点相对集中,其中心位置处于位势图的极大值点处,因此将位势图中大于阈值H的极大值点确定为运动目标的中心位置。 [0020] Because of the feature point relative concentration of the same moving object, which is at the center position of the maxima of the potential map, therefore maxima bit larger than the threshold H is determined as a potential map center position of the moving object.

[0021] 本发明采用帧差法进行运动目标的提取,利用势函数法进行目标的定位,对于不连通的目标也能够实现有效地定位,并能够抑制距离目标较远的噪声对目标定位的影响,有利于提高运动目标检测的准确性。 [0021] The present invention employs a frame difference extracted moving target object using the positioning potential function, the target can be achieved without communication efficiently locate and suppress the influence of noise from distant targets target location It will help improve the accuracy of the moving target detection.

[0022] 实施例 [0022] Example

[0023] 图I给出序列图像的视频图像,图2中的黑十字线标明了序列图像中运动目标的检测结果。 [0023] FIG I gives a sequence of images of video images, FIG. 2 in black crosshair mark detection result in the image sequences of moving objects. 由实例结果可以看出,本发明方法可对连续图像中的运动目标进行较为准确的检测定位。 Examples can be seen from the results, the method of the present invention may be continuous images of moving objects detected accurately positioned.

Claims (3)

1. 一种多运动目标的检测方法,其特征在于,采用帧差法进行运动目标的提取,利用势函数法进行目标的定位。 A multi moving target detection, characterized in that the extraction of moving objects using a frame difference method for positioning a target using potential function. 在一个固定的场景中,当存在运动物体时,采集到的相邻两帧图像之间会存在一定的差异,通过对相邻图像的差分运算可提取出运动目标。 There will be some differences between in a stationary scenario, when there is a moving object, acquired two adjacent frame images, by calculating the difference of adjacent images can be extracted moving target. 以提取到的运动目标的像素点为势函数的中心点,建立场景的位势图,位势图的极大值点确定为运动目标的中心位置。 To pixels of the extracted moving object as the center point of the potential function, establishing a potential map scene, FIG potential site maxima is determined as the center position of the moving object.
2.根据权利要求I所述的一种多运动目标的检测方法,其特征在于,U,y)、fk(x,y)和fk+1(x,y)分别为k-Ι时刻、k时刻和k+Ι时刻获得的三帧图像在坐标(x,y)处的像素灰度值,对其中相邻两帧图像进行差分运算,并与阈值T进行比较,插值小于等于阈值的像素点确定为背景像素点,用O表示;插值大于阈值的像素点确定为运动目标像素点,用I表示。 A method for detecting moving target multi claimed in claim I, wherein, U, y), fk (x, y) and fk + 1 (x, y) are k-Ι time, k three images in time and k + Ι time obtained in the pixel gray value at the coordinates (x, y) at, for which the differential operation two adjacent images, and the threshold value T is compared, the interpolation less pixel threshold It is determined as a background pixel is represented by O; interpolation pixels greater than a threshold determined as a motion target pixel, denoted by I. 其公式如下: The formula is as follows:
Figure CN102881025AC00021
为了检测出准确的运动目标,对上述两个公式所求得的相邻帧差值后的二值图像f_(x,y)和F+(x,y)进行“与”操作,提取出运动目标的像素位置,即: F(x, y) = F_ (X, y)&F+ (x, y) (3) 由上式获得的二值图像F(x, y)中,像素值为“I”的像素点为运动目标像素点,像素值为“O”的像素点为背景像素点。 In order to accurately detect the moving object, the binary image f_ (x, y) after a difference between the two adjacent frames obtained formulas and F + (x, y) a "and" operation, the extracted moving object pixel positions, namely: F (x, y) = F_ (X, y) & F + (x, y) (3) obtained by the above formula binary image F (x, y), the pixel value "I" the pixel as moving target pixel, a pixel value "O" of the pixel as a background pixel point.
3.根据权利要求1、2所述的一种多运动目标的检测方法,其特征在于,将所求得的二值图像F(x,y)中的运动目标像素点确定为特征点。 3. A method of detecting multiple moving object according to claim 1, characterized in that, to determine the determined moving target pixel binary image F (x, y) of the feature point. 设特征点集为V= Ivi(Xpyi)Ki = I,2,...,M。 Set the feature point set to V = Ivi (Xpyi) Ki = I, 2, ..., M. 其中,M为运动目标像素点的总数。 Wherein, M is the total number of pixels of the moving target. 以特征点为中心,确定M个位势函数如下: K (V,Vi) = exp [- a | Wi |2], i = I, 2, . . . , M (4) 将上述M个位势函数进行求和,获得整幅图像的位势图: Feature point as the center, determining the M potential function is as follows: K (V, Vi) = exp [- a | Wi | 2], i = I, 2,, M (4) to the M bits... summing potential function is obtained potential map of the entire image:
Figure CN102881025AC00022
由于同一运动目标的特征点相对集中,其中心位置处于位势图的极大值点处,因此将位势图中大于阈值H的极大值点确定为运动目标的中心位置。 Since the feature point relative concentration of the same moving object, which is at the center position of the maximum point of the potential map, so the potential is greater than the maximum points in FIG threshold value H is determined as the center position of the moving object.
CN2012103426717A 2012-09-17 2012-09-17 Method for detecting multiple moving targets CN102881025A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN2012103426717A CN102881025A (en) 2012-09-17 2012-09-17 Method for detecting multiple moving targets

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN2012103426717A CN102881025A (en) 2012-09-17 2012-09-17 Method for detecting multiple moving targets

Publications (1)

Publication Number Publication Date
CN102881025A true CN102881025A (en) 2013-01-16

Family

ID=47482338

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2012103426717A CN102881025A (en) 2012-09-17 2012-09-17 Method for detecting multiple moving targets

Country Status (1)

Country Link
CN (1) CN102881025A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103674104A (en) * 2013-10-28 2014-03-26 中国科学院长春光学精密机械与物理研究所 Method for increasing tracking distance of photoelectric device
CN104966375A (en) * 2015-07-01 2015-10-07 安徽创世科技有限公司 Security monitoring system and monitoring method
CN106056623A (en) * 2016-05-18 2016-10-26 深圳众思科技有限公司 Fixed scene moving object extraction method and device
CN106651923A (en) * 2016-12-13 2017-05-10 中山大学 Method and system for video image target detection and segmentation

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101702263A (en) * 2009-11-17 2010-05-05 重庆大学 Pedestrian crosswalk signal lamp green wave self-adaption control system and method
CN101806752A (en) * 2010-03-12 2010-08-18 长沙图创机电科技有限公司 Method and equipment for visual detection of visible foreign matters in bottled liquid medicine
CN102270346A (en) * 2011-07-27 2011-12-07 宁波大学 Extracting a target object in an interactive video

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101702263A (en) * 2009-11-17 2010-05-05 重庆大学 Pedestrian crosswalk signal lamp green wave self-adaption control system and method
CN101806752A (en) * 2010-03-12 2010-08-18 长沙图创机电科技有限公司 Method and equipment for visual detection of visible foreign matters in bottled liquid medicine
CN102270346A (en) * 2011-07-27 2011-12-07 宁波大学 Extracting a target object in an interactive video

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
修春波等: "基于兴趣点检测的目标识别算法", 《武汉大学学报(理学版)》 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103674104A (en) * 2013-10-28 2014-03-26 中国科学院长春光学精密机械与物理研究所 Method for increasing tracking distance of photoelectric device
CN103674104B (en) * 2013-10-28 2017-01-25 中国科学院长春光学精密机械与物理研究所 Method for increasing tracking distance of photoelectric device
CN104966375A (en) * 2015-07-01 2015-10-07 安徽创世科技有限公司 Security monitoring system and monitoring method
CN106056623A (en) * 2016-05-18 2016-10-26 深圳众思科技有限公司 Fixed scene moving object extraction method and device
CN106651923A (en) * 2016-12-13 2017-05-10 中山大学 Method and system for video image target detection and segmentation

Similar Documents

Publication Publication Date Title
TWI448977B (en) Method and apparatus for video analytics based object counting
Singla Motion detection based on frame difference method
Ju et al. Depth saliency based on anisotropic center-surround difference
Pan et al. Robust occlusion handling in object tracking
JPWO2010113389A1 (en) Stereo image processing apparatus and stereo image processing method
CN101960490A (en) Image processing method and image processing apparatus
Su et al. Abrupt motion tracking using a visual saliency embedded particle filter
CN101996406A (en) No-reference structural sharpness image quality evaluation method
US20120328161A1 (en) Method and multi-scale attention system for spatiotemporal change determination and object detection
US20150248590A1 (en) Method and apparatus for processing image of scene of interest
Liu et al. Detecting and tracking people in real time with RGB-D camera
Yin et al. Fast-moving target tracking based on mean shift and frame-difference methods
Sugandi et al. Tracking of moving objects by using a low resolution image
US8705894B2 (en) Image rotation from local motion estimates
CN103077539B (en) Motion target tracking method under a kind of complex background and obstruction conditions
TW200929005A (en) Human face detection and tracking method
Shantaiya et al. Multiple object tracking using Kalman filter and optical flow
JP2014192700A (en) Tracking processing apparatus, tracking processing system with the same, and tracking processing method
Placht et al. Rochade: Robust checkerboard advanced detection for camera calibration
CN101303732A (en) Method for apperceiving and alarming movable target based on vehicle-mounted monocular camera
Liu et al. Temporal profile based small moving target detection algorithm in infrared image sequences
CN102866237A (en) Water quality safety online biological early warning and monitoring system based on video identification
Fradi et al. People counting system in crowded scenes based on feature regression
CN102184552A (en) Moving target detecting method based on differential fusion and image edge information
CN102654902A (en) Contour vector feature-based embedded real-time image matching method

Legal Events

Date Code Title Description
C06 Publication
C10 Entry into substantive examination
C02 Deemed withdrawal of patent application after publication (patent law 2001)