外文翻譯---一個(gè)魯棒的基于機(jī)器視覺(jué)運(yùn)動(dòng)目標(biāo)檢測(cè)與跟蹤系統(tǒng)（英文）

1、A Robust Vision-based Moving Target Detection and Tracking System Alireza Behrad, Ali Shahrokni, Seyed Ahmad Motamedi Electrical Engineering Department AMIRKABIR University of Technology 424 Hafez Ave. 15914 TEHRAN, I

2、RAN Email: b7723958@aku.ac.ir, ashahrokni@yahoo.com, motamedi@aku.ac.ir Kurosh Madani Intelligence in Instrumentation and Systems Lab. (I2S) SENART Institute of Technology - University PARIS XII Avenue Pierre POINT,

3、F-77127 LIEUSAINT – FRANCE Email: madani@univ-paris12.fr Abstract In this paper we present a new algorithm for real-time detection and tracking of moving tar- gets in terrestrial scenes using a mobile camera. Our algo

4、rithm consists of two modes: detection and tracking. In the detection mode, background motion is estimated and compensated using an affine transformation. The resultant motion- rectified image is used for detection of

5、 the target location using split and merge algorithm. We also checked other features for precise detection of the target location. When the target is identi- fied, algorithm switches to the tracking mode. Modified Mo

6、ravec operator is applied to the target to identify feature points. The feature points are matched with points in the region of interest in the current frame. The corresponding points are further refined using dispar

7、ity vec- tors. The tracking system is capable of target shape recovery and therefore it can successfully track targets with varying distance from camera or while the camera is zooming. Local and re- gional computation

8、s have made the algorithm suitable for real-time applications. The refined points define the new position of the target in the current frame. Experimental results have shown that the algorithm is reliable and can suc

9、cessfully detect and track targets in most cases. Key words: real time moving target tracking and detection, feature matching, affine trans- formation, vehicle tracking, mobile camera im- age. 1 Introduction Visual d

10、etection and tracking is one of the most challenging issues in computer vision. Applica- tion of the visual detection and tracking are nu- merous and they span a wide range of applica- tions including surveillance syste

11、m, vehicle tracking and aerospace application, to name a few. Detection and tracking of abstract targets (e.g. vehicles in general) is a very complex problem and demands sophisticated solutions using conventional pa

12、ttern recognition and mo-1. Select N random feature point from previ- ous frame, and use the standard normalized cross correlation method to locate the corre- sponding points in the current frame. Normal- ized correlat

13、ion equation is given by: 2 / 12 2,2 2 1,12 2 1,1] ) , ( [ ] ) , ( [] ) , ( ][ ) , ( [? ?? ? ?? ?? ? ? ? ?? ?=∑ ∑∑∈ ∈∈f y x f f y x ff y x f f y x frS y x S y xS y x(2) here 2 1and f fare the average intensities of the

14、pixels in the two regions being compared, and the summations are carried out over all pixels with in small windows centered on the feature points. The value r in the above equation meas- ures the similarity between tw

15、o regions and is between 1 and -1. Since it is assumed that mov- ing objects are less than 50% of the whole im- age, therefore most of the N points will belong to the stationary background. 2. Select M random sets of

16、three feature points: (xi , yi , Xi , Yi ) for i=1,2,3, from the N feature points obtained in step 1. (xi ,yi) are coordinates of the feature points in the previous frame, and (Xi , Yi ) are their corresponds in curr

17、ent frame. 3. For each set calculate the affine transforma- tion parameters. 4. Transform N feature points in step 1 using M affine transformations, obtained in step 3 and calculate the M medians of squared differen

18、ces between corresponding points and transformed points. Then select the affine parameters for which the median of squared difference is the minimum. According to the above procedure, the probabil- ity p that at lea

19、st one data set in the background and their correct corresponding points are ob- tained is derived from the following equation [7]: M q M q p ) ) ) 1 (( 1 ( 1 ) , , ( 3 ε ε ? ? ? =(3) where ε(<0.5) is the ratio of t

20、he moving object regions to whole image and q is the probability that corresponding points are correctly find. In [7] it has been shown that the above method will give an accurate and reliable model. 2.2 Moving targ

21、et detection using background motion compensated frames When affine parameters are estimated, they can be used for cancellation of the apparent back- ground motion, by transformation of previous frame. Now difference

22、 of the current frame and transformed previous frame reveals true moving targets. Then we apply a threshold to produce a binary image. The results of the transformation and segmentation are shown is figure 1-a and 1-

23、 b. Some parts are segmented as moving targets due to noise. Connected component property can be applied to reduce errors due to noise. We use split and merge algorithm to find target bounding-boxes. If no target is

24、 found, then it means either there is no moving target in the scene or, the relative motion of the target is too small to be detected. In the latter case, it is possible to detect the tar- get by adjusting the frame r

25、ate of the camera. The algorithm accomplishes this automatically by analyzing the proceeding frames until a tar- get is detected. Our special interest is detection and tracking of the moving vehicles so we used aspec

26、t ratio and horizontal and vertical line as constraints to verify vehicles. Our experiments show that comparison of the length of horizontal and vertical lines in the target area with the pe- rimeter of the target wil

27、l give a good clue about the nature of the target. 3 Target tracking After a target is verified, the algorithm switches into the tracking mode. Modified Moravec op- erator is applied to the target to identify featur

28、e points. These feature points are matched with points in the region of interest in the current frame. Disparity vectors are computed for the matched pairs of points. We used disparity vec- tors to refine the matched