JP5048609B2 - Object tracking device and program - Google Patents

Description

  The present invention relates to an object tracking apparatus and program, and more particularly to an object tracking apparatus and program for tracking an area representing a tracking object.

Conventionally, an image processing method for tracking a moving body is known in which a detected object region is divided into small regions, and each small region is collated between images to perform tracking (Patent Document). 1). Further, there is known an object tracking device that performs tracking by generating and collating a plurality of tracking candidate positions using a motion model of an object (Patent Document 2).
JP 2006-178669 A JP 2007-328746 A

  However, in the technique described in Patent Document 1 described above, in order to stably track an object that is partially deformed, it is necessary to reduce the small area to be divided. There is a problem that an error occurs in the collation and the amount of calculation in the collation increases.

  In addition, the technique described in Patent Document 2 has a problem in that when an object is partially deformed, an error occurs in collation, and thus tracking cannot be performed.

  The present invention has been made to solve the above-described problems, and can track an object stably with simple processing even when the tracked object is partially deformed. And to provide a program.

In order to achieve the above object, an object tracking device according to the present invention includes an imaging unit that images a predetermined area, and an area that represents the tracking object obtained for the first image captured by the imaging unit. A feature point extracting unit that extracts a plurality of feature points, and a point corresponding to each of the feature points extracted by the feature point extracting unit is captured by the imaging unit at a timing different from that of the first image. A search means for searching from each of the two images, a relative position of a specific position in the region representing the tracking object with respect to each of the feature points in the first image, and a search by the search means in the second image. based on each of points in the second image, and each point found by said searching means, the relative position with respect to the feature points from the corresponding said point For each intersection formed from a plurality of straight lines passing through a position, a total value or an average value of distances to other intersections is calculated, and the calculated total value or average value of the distances is equal to or greater than a predetermined value. Extracting the searched points on a common straight line forming an intersection, and when the points are extracted, each of the searched points other than the extracted points and the feature points corresponding to the points Based on the relative position with respect to the specific position estimating means for estimating the specific position in the second image , and based on the specific position in the second image estimated by the specific position estimating means, Area estimation means for estimating an area representing the tracking object in the second image.

Further, the program according to the present invention is a feature point extracting unit that extracts a plurality of feature points from an area representing a tracking target obtained for a first image captured by an image capturing unit that captures a predetermined area. Search means for searching points corresponding to each of the feature points extracted by the feature point extraction means from the second images captured by the imaging means at different timings from the first image, Based on the relative position of the specific position in the region representing the tracking object with respect to each of the feature points in one image, and each of the points searched by the search means in the second image, the second In the image, a plurality of straight lines passing through each of the points searched by the search means and the relative position from the point to the corresponding feature point. For each intersection to be formed, calculate a total value or an average value of distances with other intersections, and on a common straight line that forms a plurality of intersection points where the calculated total value or average value of the distances is equal to or greater than a predetermined value Extracting the searched points, and when the points are extracted, based on each of the searched points other than the extracted points and the relative position with respect to the feature points corresponding to the points, Specific position estimating means for estimating the specific position in the second image, and the tracking object in the second image based on the specific position in the second image estimated by the specific position estimating means. Is a program for functioning as region estimation means for estimating a region representing.

  According to the present invention, a predetermined region is imaged by the imaging unit, and a plurality of feature points are extracted from the region representing the tracking target obtained for the first image captured by the imaging unit by the feature point extracting unit. To do. Then, the search unit searches for points corresponding to each of the feature points extracted by the feature point extraction unit from the second image captured by the imaging unit at a timing different from that of the first image.

  Based on the relative position of the specific position in the region representing the tracking object with respect to each of the feature points in the first image by the specific position estimation unit, and each of the points searched by the search unit in the second image, A specific position in the second image is estimated. Then, the region estimation unit estimates a region representing the tracking target in the second image based on the specific position in the second image estimated by the specific position estimation unit.

  As described above, the tracking in the second image is performed based on the relative position of the specific position in the region representing the tracking target with respect to each of the feature points in the first image and each of the corresponding points in the second image. Even if the tracking object is partially deformed by estimating the specific position in the area representing the object and estimating the area representing the tracking object in the second image, the processing can be performed with a simple process. Can track the tracking object stably.

The specific position estimating means according to the present invention is formed from a plurality of straight lines that pass through each of the points searched by the searching means and positions relative to the corresponding feature points from the points in the second image. For each intersection, the total value or average value of the distances to other intersections is calculated, and the searched on the common straight line forming a plurality of intersections where the calculated distance total value or average value is equal to or greater than a predetermined value A point is extracted, and when the point is extracted, a specific position in the second image is estimated based on each of the searched points other than the extracted point and a relative position with respect to the feature point corresponding to the point. you. Even if the tracked object is partially deformed by excluding the points that exist in the part that operates differently from the other parts and estimating the specific position in the area of the tracked object, The region of the tracking object in the second image can be estimated with high accuracy.

  The object tracking device according to the present invention includes a size of a region representing a tracking object in the first image, a distance between each feature point in the first image and the specific position, and a search unit in the second image. Based on the distance between each of the searched points and the estimated specific position, it further includes size estimating means for estimating the size of the area representing the tracking object in the second image, and the area estimating means is specified. An area representing the tracking object in the second image is estimated based on the specific position in the second image estimated by the position estimation means and the size of the area representing the detection object estimated by the size estimation means. can do.

  The object tracking device according to the present invention is estimated by an area detection unit for detecting an area representing a tracking object from the second image, an area representing the tracking object detected by the area detection unit, and an area estimation unit. And a region correction unit that corrects a region representing the tracking target in the second image based on the region representing the tracking target. Thus, by correcting the area representing the estimated tracking object using the detected area representing the tracking object, the area of the tracking object in the second image can be accurately estimated.

  The object tracking device according to the present invention further includes area detecting means for detecting an area representing the tracking object from the first image, and the feature point extracting means is an area representing the tracking object detected by the area detecting means. From this, a plurality of feature points can be extracted.

  The feature point extraction means according to the present invention uses the second image in which the region representing the tracking object is estimated last time as the first image, and extracts a plurality of feature points from the region representing the tracking object estimated last time. Can be extracted.

  The feature point extraction means according to the present invention uses the second image in which the region representing the tracking object is estimated last time as the first image, and extracts a plurality of feature points from the region representing the tracking object corrected last time. Can be extracted.

  The tracking target can be a partially deformable object.

  As described above, according to the object tracking device and the program of the present invention, the relative position of the specific position in the region representing the tracking object with respect to each of the feature points in the first image, and the second image Based on each of the corresponding points, a specific position in the region representing the tracking object in the second image is estimated, and the region representing the tracking object in the second image is estimated, so that the tracking object is partially Even if it is a case where it deform | transforms automatically, the effect that a tracking target object can be tracked stably by simple processing is acquired.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  As shown in FIG. 1, an object tracking device 10 according to the present embodiment is mounted on a vehicle (not shown), and continuously captures an image of the front of the host vehicle to generate a captured image. The captured image obtained from the imaging device 12 is provided with a computer 14 that tracks a target area representing a pedestrian as a tracking target and displays a tracking result on the display device 16.

  The imaging device 12 captures an image of the front of the host vehicle and generates an image signal of the image, and includes an imaging unit (not shown) configured to perform A / D conversion on the image signal generated by the imaging unit. A / D conversion unit (not shown) and an image memory (not shown) for temporarily storing the A / D converted image signal. Further, the display device 16 is configured by a display mounted on the host vehicle.

  The computer 14 is configured to include a CPU, a ROM storing a program for executing a tracking processing routine to be described later, a RAM storing data and the like, and a bus connecting them. When the computer 14 is described in terms of functional blocks divided for each function realization means determined based on hardware and software, as shown in FIG. 1, an image for acquiring captured images continuously output from the imaging device 12 From the input unit 20, a target detection unit 22 that detects a target region representing a pedestrian from the captured image acquired by the image input unit 20, and each of a plurality of captured images acquired by the image input unit 20. A feature point extracting unit 24 that extracts a plurality of feature points that are easy to track on an image, and a feature that searches for corresponding points between the two images from the feature points in each of the two images obtained by the feature point extracting unit 24 The point search unit 26, each of the corresponding feature points obtained by the corresponding point search unit 26, and the object region in one captured image are input, and the target in the other captured image An object area estimating unit 28 for estimating an area; an object area correcting unit 30 for correcting the estimated object area using the object area detected for the other captured image by the object detecting unit 22; And a display control unit 32 that displays the correction result of the object region correction unit 30 on the display device 16.

  The image input unit 20 acquires two captured images captured at different timings from the imaging device 12. In the following description, of the two captured images, the captured image captured first will be referred to as a first captured image, and the captured image captured later will be described as a second captured image.

  The object detection unit 22 detects an object region representing a pedestrian by pattern recognition from each of the first captured image and the second captured image acquired by the image input unit 20. The object detection unit 22 stores a pattern of image feature values of an area representing a pedestrian obtained by learning. Note that other existing detection methods may be used for the detection method of the object region, for example, detection results by other sensors such as millimeter wave radar and laser radar, time series images, or images captured by a stereo camera. A target area representing a pedestrian may be detected using a detection result obtained by detecting a three-dimensional object from.

  The feature point extraction unit 24 extracts feature points from the object area in the first captured image acquired from the imaging device 12 and extracts feature points from the second captured image acquired from the imaging device 12. A feature point refers to a point that can be distinguished from surrounding points and that makes it easy to obtain a correspondence between different images. The feature points are automatically extracted by using a method (for example, a Harris operator) that detects pixels in which the gradient value of the change in shading increases two-dimensionally. In the method using the Harris operator, feature points are extracted as described below. First, the matrix M is calculated by the following equation (1), where the luminance of the point (u, v) of the image is I (u, v).

However, I _u and I _v represent horizontal and vertical differentiation, respectively, and G _σ represents smoothing by a Gaussian distribution with a standard deviation σ.

Then, using the eigenvalues λ ₁ and λ ₂ of the matrix M calculated by the above equation (1), the corner strength is calculated by the following equation (2).

  However, k is a preset constant, and a value of 0.04 to 0.06 is generally used. In the method using the Harris operator, a point at which the corner intensity is equal to or greater than a threshold value is selected, and the selected point is extracted as a feature point.

  The corresponding point search unit 26 searches the feature points extracted in the second captured image for points corresponding to the feature points in the object area of the first captured image extracted by the feature point extraction unit 24. The feature points are associated.

  In associating feature points between two captured images, a pair of points having similar brightness distributions in a small region set around the feature point is selected, and the selected point pair is associated with To do. A value such as SSD (Sum of Squared Differences) may be used to determine whether the two feature points are similar.

For example, with respect to the feature point p = (x, y) in the object region of the _first captured image I ₁ , the feature point p ′ = (x ′, y ′) of the _second captured image I ₂ The following formula (3) is calculated for each combination as a pair.

ただし、ｒは、小領域の大きさを決定する変数であり、予め定められている。

Here, r is a variable that determines the size of the small area, and is predetermined.

  Then, the feature point of the second captured image when the value calculated by the above equation (3) is minimized is searched for and the point corresponding to the feature point in the object region of the first captured image is obtained. .

  The above calculation is performed for each feature point in the object area of the first captured image, and points corresponding to each of the feature points in the object area of the first captured image are determined as the second captured image. Search from feature points.

  Further, the corresponding point search unit 26 searches for at least three sets of corresponding points between the two captured images with respect to the feature points extracted from each of the two captured images by the feature point extraction unit 24.

  Here, the principle of the present embodiment will be described. When tracking an object in a time-series image, it is possible to search for a region that is the same object by matching between images, but it is possible to track an object that is partially deformed, such as a person. In this case, there may be a case where a corresponding area between images cannot be obtained. It is also possible to estimate the position and size of the object region using the object motion model, but if there is a change in the motion of the imaging device or object, tracking of the object is not possible. It becomes difficult.

  On the other hand, by tracking each of a plurality of feature points in the object region between images, it is possible to stably track an object that is partially deformed.

  Therefore, in the present embodiment, the first imaging obtained by the object region estimation unit 28 and the object region in the first captured image obtained by the object detection unit 22 and the corresponding point search unit 26. The object area representing the pedestrian in the second captured image is estimated using the correspondence between the feature points in the object area of the image and the feature points of the second captured image.

  A method for estimating the object region in the second captured image will be described below. First, as shown in FIG. 2A, the width w, the height h, and the center position of the object area are obtained for the object area detected in the first captured image. Next, as shown in FIG. 2B, feature points pi (i = 1,..., N extracted from the object region of the first captured image, where n is the extracted feature point. The distance di to the center position and the direction θi to the center position are calculated.

  If the point corresponding to the feature point pi of the first captured image is the feature point p′i of the second captured image, as shown in FIG. By drawing a straight line in the θi direction from each feature point p′i (i = 1,..., N), the feature point p′i and the feature point p′i are corresponding to the corresponding feature point pi. Draw a straight line that passes through the position relative to the center position. Note that FIG. 3A illustrates an example in which a point corresponding to the feature point p6 has not been searched, and the points for which the corresponding point has not been searched in the second captured image in this way are straight lines. It is assumed that the process of drawing is not performed.

  Next, a point closest to each straight line drawn from each feature point p′i is obtained and used as an estimated value of the center position of the object region in the second captured image. At this time, if the tracking object is partially deformed, the movement of the feature point in the deformed part is different from the movement of the entire tracking object. Exclude it from being used.

  Here, a method for identifying feature points to be excluded will be described. First, for each intersection formed from a plurality of straight lines drawn for each feature point that is a corresponding point in the second captured image, the total value of the distances to other intersections is calculated, and the calculated distance Intersections whose total value is greater than or equal to the threshold are extracted. In addition, about said threshold value, what is necessary is just to obtain | require the value which multiplied the variance value of the total value of the distance with another intersection, and a predetermined coefficient, and to set as a threshold value.

  When a plurality of intersection points whose total distance is equal to or greater than the threshold value are extracted, feature points on a common straight line that form the plurality of extracted intersection points from the feature points that are corresponding points in the second captured image Are specified and excluded as feature points. In the second captured image as shown in FIG. 3A, each of the intersections formed by the two straight lines drawn from each of the pair combinations of the feature points p′1 to p′4 is Although the positions are almost the same, the intersection formed by two straight lines drawn from the feature point p′5 and each of the other feature point pairs exists at a position away from the other intersections. A feature point p′5 on a common straight line that forms a plurality of intersections where the total value of the distances is equal to or greater than the threshold is determined to be a feature point of the portion where the deformation has occurred, and this feature point is used for the center position estimation process Is excluded from being used.

  Then, as shown in FIG. 3B, a point close to each of the straight lines drawn from each of the corresponding points in the second captured image other than the excluded feature points is estimated as the center position.

  As described above, when the center position of the object region in the second captured image is estimated, next, as shown in FIG. 4A, in the second captured image, each feature point p′i and The distance d′ i from the estimated center position is calculated. At this time, the calculation of the distance from the center position is not performed for the feature points excluded in the center position estimation process.

  Then, for each feature point (corresponding point) for which the distance d′ i to the estimated center position is calculated, the ratio d′ i to the distance from the feature point to the center position in the corresponding first captured image. / Di is calculated, and the average value of the calculated ratios is set as an enlargement ratio α of the object area of the second captured image with respect to the object area of the first captured image.

  Next, in the second captured image, as shown in FIG. 4B, the width α × w and the height α × h are estimated and estimated as the size of the object region in the second captured image. A region having the center position as the center and the estimated size as the size of the region is estimated as the object region.

  The object area estimation unit 28 estimates the object area in the second captured image by the method described above.

  The object area correction unit 30 uses the object area detected by applying the object detection unit 22 to the second captured image, and uses the object area estimation unit 28 to estimate the second captured image. Correct the object area.

  When the detected object area and the estimated object area overlap, the correction of the object area is performed by calculating the average or weighted average of those object areas. The object area is set as a correction value of the object area. In addition, although the case where correction was performed when the detected object area and the estimated object area overlap each other was described as an example, the present invention is not limited to this, and the detected object area is not limited thereto. May be corrected when it is determined that the position and size of the target object are close to the estimated position and size of the object region.

  In addition, when the weighted average is used for correction, the longer the tracking time of the object area representing the pedestrian, the greater the weight of the estimated object area or the reliability of the detected object area. The higher the degree, the greater the weight of the detected object area.

  The target area of the second captured image corrected by the target area correction unit 30 is input to the feature point extraction unit 24 in the next target area estimation. In addition, the second captured image is the first captured image in the next estimation, and the target region is estimated using the captured image captured next as the second captured image, so that it is continuous in time series. Then, the object area representing the pedestrian is tracked.

  The display control unit 32 causes the display device 16 to display the corrected object region estimation result together with the second captured image.

  Next, the operation of the object tracking device 10 according to the present embodiment will be described. In addition, the case where the pedestrian who exists ahead is tracked while driving | running | working the vehicle carrying the target object tracking apparatus 10 is demonstrated to an example.

  When continuous imaging in front of the host vehicle is started by the imaging device 12, a tracking processing routine shown in FIG.

  First, in step 100, two captured images are sequentially acquired from the imaging device 12 as a first captured image and a second captured image. In step 102, an object region representing a pedestrian is detected from the first captured image acquired in step 100 by pattern recognition.

  In the next step 104, a plurality of feature points are extracted from the object area in the first captured image, and in step 106, a plurality of feature points are extracted from the second captured image. In step 108, a point corresponding to each of the plurality of feature points extracted in step 104 is searched from the plurality of feature points extracted in step 106.

  In step 110, the center position, width, and height of the object region obtained in step 102 or step 128 described later are measured for the first captured image. Then, in step 112, the relative position of the center position with respect to the feature point for each of the feature points in the object area of the first captured image that are searched for corresponding points in step 108 above. As described above, the distance to the center position of the object region measured in step 110 and the direction to the center position are calculated.

  In step 114, a straight line extending in the direction to the center position calculated for the corresponding feature point in step 112 is drawn from each of the corresponding feature points in the second captured image searched in step 108. A plurality of intersections are formed by the plurality of straight lines drawn in this way.

  In the next step 116, for each intersection formed by the plurality of straight lines generated in step 114, the total value of the distances from other intersections is calculated. In step 118, intersection points where the total value of the distances calculated in step 116 is equal to or greater than a threshold value are extracted, and when a plurality of intersection points are extracted, corresponding points on a straight line common to the plurality of intersection points are specified. Thus, a feature point on a common straight line that forms a plurality of intersections existing at distant positions and corresponding feature points searched in step 108 are specified, and the specified feature points are excluded. set to target.

  In step 120, a position close to each straight line drawn from a feature point corresponding to the feature point searched in step 108 and other than the feature point excluded in step 118 is determined. It is estimated as the center position of the object region in the two captured images. In the next step 122, each of the corresponding feature points searched in step 108 and other than the feature points excluded in step 118, and the center position estimated in step 120, Each distance is calculated.

  In step 124, the ratio between the distance between the feature point calculated in step 112 and the center position and the distance between the feature point calculated in step 122 and the center position is set to the corresponding feature point pair. The average value of the obtained ratios obtained for each is used as the enlargement ratio of the size of the object area, and the width and height of the object area of the first captured image measured in step 110, and the enlargement ratio Based on the above, the width and height of the object area in the second captured image are estimated. Then, based on the estimated width and height of the target area in the second captured image and the center position estimated in step 120, the target area in the second captured image is estimated.

  In step 126, an object region representing a pedestrian is detected by pattern recognition for the second captured image, and in step 128, the object region detected in step 126 is used and estimated in step 124. Correct the target area.

  In step 130, the object region corrected in step 128 is displayed on the display device 16 together with the second captured image. In the next step 132, the second captured image is used as the first captured image in the estimation of the next target area, and the target area corrected in step 128 is used as the target of the first captured image. This is an area. In step 134, the next captured image is acquired from the imaging device 12, and the process returns to step 104 as the second captured image in the next estimation, and the second newly acquired in step 104 to step 128. The object region in the captured image is estimated.

  As described above, by repeating Step 104 to Step 134, the object region representing the pedestrian is tracked in the continuously captured images.

  As described above, according to the object tracking device according to the present embodiment, the distance to the center position in the object area for each feature point in the first captured image, the direction, and the second captured image. Based on each of the corresponding points in, the center position in the target area in the second captured image is estimated, and the target area in the second captured image is estimated, so that the tracking target is partially Even in the case of deformation, the tracking object can be tracked stably with a simple process.

  Further, in the second captured image, the tracking target object is partially deformed by estimating the center position in the target object area by excluding the feature points that exist in the part that has operated differently from the other part. Even in this case, the region of the tracking target in the second captured image can be estimated with high accuracy.

  In addition, the region of the tracking target in the second captured image is corrected by correcting the target region estimated in the second captured image using the target region detected from the second captured image by pattern recognition. Can be estimated more accurately.

  In the above embodiment, an example is described in which an intersection point that is located at a distance from another intersection point is extracted by extracting an intersection point where the total value of the distances to other intersection points is equal to or greater than a threshold value. However, the present invention is not limited to this, and by extracting the intersection where the average value of the distance to the other intersection is equal to or greater than the threshold, the intersection existing at a position away from the other intersection is extracted. May be.

  Moreover, although the case where the tracking target object is a pedestrian has been described as an example, the present invention is not limited to this, and a partially deformable object other than the pedestrian may be used as the tracking target object.

  Moreover, although the case where the center position of the object area is estimated has been described as an example, the present invention is not limited to this, and a specific position other than the center position in the object area may be estimated. In this case, what is necessary is just to obtain | require the distance to the specific position in the target object area | region with respect to each of the feature points in a 1st captured image, and the direction to a specific position.

  Moreover, although the case where a feature point is extracted from the detected or previously corrected object region has been described as an example for the first captured image, the present invention is not limited to this, and the feature is extracted from the entire first captured image. After extracting the points, the feature points in the object area detected or previously corrected may be extracted from the extracted feature points.

It is a block diagram which shows the target tracking apparatus which concerns on embodiment of this invention. (A) The image figure which shows the target object area | region and center position which were detected from the 1st captured image, (B) The image figure which shows the feature point extracted from the 1st captured image. (A) It is an image figure which shows the feature point corresponding in a 2nd captured image, and the straight line pulled from the feature point, (B) It is an image figure which shows the center position estimated in the 2nd captured image. (A) In the 2nd captured image, it is an image figure which shows the distance of a corresponding feature point and the estimated center position, (B) It is an image figure which shows the target object area | region estimated in the 2nd captured image. It is a flowchart which shows the content of the tracking process routine in the computer of the target object tracking apparatus which concerns on embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Object tracking apparatus 12 Imaging apparatus 14 Computer 22 Object detection part 24 Feature point extraction part 26 Corresponding point search part 28 Object area estimation part 30 Object area correction part

Claims (9)

Imaging means for imaging a predetermined area;
Feature point extracting means for extracting a plurality of feature points from an area representing the tracking target obtained for the first image imaged by the imaging means;
Retrieval means for retrieving points corresponding to each of the feature points extracted by the feature point extraction means from the second images captured by the imaging means at different timings from the first image,
Wherein for each of said feature points in the first image, the relative position of a particular location in the region representing the tracked object, and based on each point found by said searching means in the second image, the first In the image of 2, for each intersection formed from a plurality of straight lines passing through each of the points searched by the search means and the position corresponding to the feature point corresponding to the point, Calculating a total value or an average value of the distances, and extracting the searched points on a common straight line forming a plurality of intersections where the calculated total value or average value of the distances is equal to or greater than a predetermined value, If is extracted, and each point which is the search except that the extracted, on the basis of said relative position with respect to the feature point corresponding to the point, estimates the specific position in the second image A specific position estimation means that,
Area estimation means for estimating an area representing the tracking object in the second image based on the specific position in the second image estimated by the specific position estimation means;
Including object tracking device. The specific position estimation means is formed of a plurality of straight lines that pass through each point searched by the search means in the second image and a position that is the relative position with respect to the corresponding feature point from the point. For each intersection, a total value or an average value of distances to other intersections is calculated, and the search on a common straight line forming a plurality of intersections where the calculated total value or average value of the distances is equal to or greater than a predetermined value Extracted points, and when the points are extracted, a plurality of straight lines passing through each of the searched points other than the extracted points and the relative positions with respect to the feature points corresponding to the points. The object tracking device according to claim 1, wherein the specific position in the second image is estimated based on the second image. The size of the area representing the tracking object in the first image, the distance between each of the feature points in the first image and the specific position, and the point searched by the search means in the second image Size estimation means for estimating a size of a region representing the tracking object in the second image based on a distance between each of the estimated position and the specific position;
The area estimation means is based on the specific position in the second image estimated by the specific position estimation means and the size of the area representing the detection target estimated by the size estimation means. The object tracking device according to claim 1, wherein an area representing the tracking object in the second image is estimated. Area detecting means for detecting an area representing the tracking object from the second image;
An area representing the tracking object in the second image based on an area representing the tracking object detected by the area detecting means and an area representing the tracking object estimated by the area estimating means; Area correction means to correct;
The object tracking device according to any one of claims 1 to 3, further comprising: An area detecting means for detecting an area representing the tracking object from the first image;
5. The object tracking device according to claim 1, wherein the feature point extracting unit extracts a plurality of feature points from a region representing the tracking target detected by the region detecting unit.   The feature point extraction means extracts a plurality of feature points from the region representing the tracking object estimated last time, using the second image in which the region representing the tracking object was estimated last time as the first image. The object tracking device according to any one of claims 1 to 4.   The feature point extraction unit extracts a plurality of feature points from a region representing the tracking object corrected last time, using the second image in which the region representing the tracking object was estimated last time as a first image. The object tracking device according to claim 4.   The object tracking device according to claim 1, wherein the tracking object is a partially deformable object. Computer
Feature point extracting means for extracting a plurality of feature points from the area representing the tracking object obtained for the first image imaged by the imaging means for imaging the predetermined area;
Retrieval means for retrieving points corresponding to each of the feature points extracted by the feature point extraction means from the second images captured by the imaging means at different timings from the first image,
Wherein for each of said feature points in the first image, the relative position of a particular location in the region representing the tracked object, and based on each point found by said searching means in the second image, the first In the image of 2, for each intersection formed from a plurality of straight lines passing through each of the points searched by the search means and the position corresponding to the feature point corresponding to the point, Calculating a total value or an average value of the distances, and extracting the searched points on a common straight line forming a plurality of intersections where the calculated total value or average value of the distances is equal to or greater than a predetermined value, If is extracted, and each point which is the search except that the extracted, on the basis of said relative position with respect to the feature point corresponding to the point, estimates the specific position in the second image Specific position estimation means, and a region estimation means for estimating a region representing the tracking object in the second image based on the specific position in the second image estimated by the specific position estimation means Program to let you.

Images