JPH0935040A - Image processing method for extracting body movement trace - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide the image processing method for body movement trace extraction which can eliminate the need to decide the identity of a body whose image is picked up at different time and can improve the tracing precision of the body and is improved in toughness against noise. SOLUTION: A time-space image I (x,y;t) is constituted by arranging an image along a time base crossing an X axis and a Y axis at right angles, this image is cut along a plane parallel to the time base, and images appearing on cut surfaces are denoted as cut surface images C (d,t:θ); and the gradients of body traces on the cut surface images are found and only parts including an angle θd are extracted from the respective cut surface images C (d,t;θ) to constitute a new trace sectional image L (s,θ;t).

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing method for extracting an object movement locus, and more particularly to an image processing method for extracting a locus of a moving object in view-based image recognition.

[0002]

2. Description of the Related Art As a method of moving image processing for detecting an object moving in a three-dimensional space, a method of using a spatiotemporal image formed by arranging images continuously obtained from a camera in a time axis direction has been proposed. .

As an example thereof, Taniguchi et al., IEICE Transactions, D-II, VOL. J77-D-II, N
o10, pp. There is "Proposal of a moving image processing method using spatiotemporal images-DTT method-" announced in October, 2019 to 2026, 1994. In this method, on the condition that the moving direction of the object is almost constant, first, significant data about the object is extracted for each frame, and the obtained data is an axis (direction axis) parallel to the moving direction of the object. One-dimensional data is created by projecting onto. Next, by arranging the one-dimensional data of each frame in time series, the spatiotemporal image is converted into a plane image composed of a direction axis and a time axis. In this two-dimensional image, the moving state of the object is represented as one area, and the moving object can be detected by performing simple two-dimensional image processing.

[0004]

However, in this method, it is essential to know the outline of the moving direction of the object in advance and set the cutting plane in that direction. Therefore, there is no inconvenience when handling an object that moves linearly on an image, but there is a problem that a general object that moves in two dimensions cannot be handled.

Therefore, a main object of the present invention is to eliminate the need to determine the identity of objects imaged at different times, to improve the tracking accuracy of the object, and to extract the object movement trajectory that can improve the robustness against noise. Is to provide an image processing method.

[0006]

The invention according to claim 1 is
In an image processing method for determining a moving direction of an object within an imaging range by processing a continuous image and tracking a movement trajectory of an object existing in the image, all continuous images within a predetermined time are arranged in chronological order. A three-dimensional spatiotemporal image is constructed, the spatiotemporal image is cut along a plurality of planes perpendicular to the original continuous image, and the image appearing on the cut surface is extracted as a cut surface image. The result obtained by calculating the inclination and selecting the section image having the maximum inclination at a predetermined time, extracting the portion of the section image at the time point, and applying this to all section images within a predetermined time A part of the obtained many cut surface images is arranged in time order to form a trajectory cross-sectional image.

According to a second aspect of the present invention, in the image processing method similar to the first aspect, a short time portion of a continuous image is arranged in time order to form a spatiotemporal image for a short time portion, and the image is used as the original image. Cut at multiple planes perpendicular to the continuous image, extract the image for a short time appearing on the cut surface as a cut surface image, find the inclination of the object trajectory appearing in the extracted cut surface image, and set a predetermined time A cutting plane image having a maximum inclination in a short time is selected, and the selected cutting plane images are arranged in time order to form a trajectory cutting image.

The invention according to claim 3 is the invention according to claim 1 or 2
In, after performing image processing for feature extraction and filter processing such as edge extraction processing, threshold value processing, and noise removal processing in advance for each of the original continuous images,
A plurality of planes perpendicular to the original continuous image are considered, and an image obtained by projecting the result of the processing on these planes is acquired as a cut plane image.

According to the fourth aspect of the present invention, when a plurality of moving objects are present in one image, the locus cross-section images for each of them are simultaneously extracted.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a flow chart showing the flow of the entire processing of an embodiment of the present invention, FIG. 2 is a view showing a state of a continuous image and a moving object, and FIG. 3 is a spatiotemporal image. FIG. 4 is a diagram showing a cross section of the spatiotemporal image by a plane parallel to the time axis, and FIG. 5 is a diagram for explaining a method of creating a trajectory cross sectional image from a plurality of cut plane images. FIG.

In FIG. 1, continuous images are taken by, for example, a television camera and input to an image processing apparatus. Then, the steps shown in FIG. 1 (abbreviated as SP in the drawing)
In SP1, as shown in FIG. 2, an image (1
1, 12 and 13) are represented by I (x,
y), all the images are arranged in order of the time when this image was obtained. That is, by considering a time axis (T axis) orthogonal to both the X axis and the Y axis, and arranging the images along this axis, a three-dimensional image as shown in FIG. 3, that is, a spatiotemporal image I ( x, y; t) can be constructed. In step SP2, this spatiotemporal image I (x, y;
t) is cut along a plane parallel to the time axis, and an image appearing on the cut surface is referred to as a cut surface image C (d, t; θ).
Here, the angle θ is the X-axis of the original image when the plane parallel to the time axis is
Θ = arctan (y / x), 0 ≦
Let θ <π. Now, when the angle θ is changed, a plurality of cut surface images C (d, t; θ) can be obtained.

In the cut surface image C (d, t; θ), a part of the image of the moving object 21 included in the original continuous image should appear as a locus. Therefore, in step SP3, when the inclination ∂d / ∂t of the object trajectory on the cut surface image C (d, t; θ) at a certain time is calculated, this is a function of the angle θ, and the certain angle θ _d (T)
Is the maximum position. In step SP4, this is applied to all the times, and as shown in FIG. 5, only the part including the angle θ _d is taken out from each cut surface image C (d, t; θ) to form a new image. . When this is referred to as the trajectory cross-sectional image L (s, θ: t), it can be seen that this image is a surface including the velocity vector of the moving object. That is, the trajectory cross-sectional image L (s, θ; t) is the spatiotemporal image I.
It is an image obtained by cutting (x, y; t) with a twisted surface along the moving direction of the object, and this surface completely includes information on the moving of the object, that is, moving speed and moving direction.

In the above embodiment, it was explained that all the continuous images within a fixed time were obtained in advance.
It is not always necessary to perform a series of procedures on consecutive images for an extremely short time, and finally arrange the obtained results in time order to obtain an image corresponding to the trajectory cross-sectional image. Good.

Further, in the above embodiment, the method of simply cutting the spatiotemporal image with a plane in order to obtain a plurality of cut plane images has been described. However, in order to better represent the movement trajectory of the object on the cut plane images. First, the original continuous image is subjected to filtering such as edge detection, noise removal, threshold value processing, and similar feature amount extraction processing in advance, and then a spatiotemporal image is constructed and cut into planes. Alternatively, an image equivalent to the cutting plane image may be acquired by projecting the result of the above-described processing onto a plane, and the trajectory cross-sectional image may be created based on this.

Further, in the above description, the case where there is only one moving object has been described, but even if there are a plurality of moving objects in a continuous image, a plurality of sets of cut plane images are separately created for each of the objects. However, if the trajectory cross-sectional image for each individual object is acquired, the above procedure can be applied as it is even when there are a plurality of moving objects.

[0016]

As described above, according to the present invention, it is possible to realize an adaptive cutting plane selection method so that the moving direction of a moving object of interest does not have to be known in advance. That is, the cutting of the first spatiotemporal image is performed in all directions, and by knowing the optimum cutting direction from this cutting result and leaving only that cutting surface, the cutting surface is a surface twisted along the movement trajectory of the object. This surface is a surface that completely contains the movement vector of the object, and since it contains all the information about the movement trajectory, the general properties common to the methods for cutting the spatiotemporal image are also applicable, so Robustness is improved.

[Brief description of drawings]

FIG. 1 is a flowchart showing an overall processing flow of an embodiment of the present invention.

FIG. 2 is a diagram showing a state of a continuous image and a moving object.

FIG. 3 is a diagram showing a spatiotemporal image.

FIG. 4 is a diagram showing a cross section of a spatiotemporal image by a plane parallel to a time axis.

FIG. 5 is a diagram showing a method of creating a trajectory cross-sectional image from a plurality of cut-plane images.

[Explanation of symbols]

 11, 12, 13 Continuous images 21 Moving object I Spatiotemporal image C Cut plane image L Trajectory cross section image

Claims (4)

[Claims] 1. An image processing method for determining a moving direction of an object within an imaging range by processing a continuous image and tracing a locus of movement of an object existing in the image. All continuous images within a predetermined time. Arrange images in time order 3
A first step of constructing a dimensional spatiotemporal image; the spatiotemporal image constructed in the first step is cut along a plurality of planes perpendicular to the original continuous image, and the image appearing on the cut surface is cut. A second step of extracting the image as a section image, obtaining the inclination of the object trajectory appearing in the second step,
The third step of selecting a cross-sectional image having the maximum inclination at a predetermined time, and the time portion of the cut image selected in the third step are extracted, and all of them are extracted within a predetermined time. An image processing method for extracting an object movement locus, comprising a fourth step of arranging parts of a large number of cutting plane images obtained as a result of application to a cutting plane image in time order to form a locus cross-sectional image. 2. An image processing method for obtaining a moving direction of an object within an imaging range by processing a continuous image to trace a locus of movement of an object existing in the image. The fifth step of arranging in order of time to form a spatiotemporal image for a short time, the spatiotemporal image for a short time is cut along a plurality of planes perpendicular to the original continuous image, and appears on the cut surface. A sixth step of extracting an image for a short time as a cut surface image, obtaining the inclination of the object trajectory appearing in the cut surface image, and selecting the cut surface image for the short time having the maximum inclination at this time. An image processing method for extracting an object movement locus, comprising: a seventh step; and an eighth step of arranging the cutting plane images selected in the seventh step in time order to form a locus cutting image. 3. The second and sixth steps perform filter processing such as edge extraction processing, threshold processing, noise removal processing and image processing for feature extraction in advance for each of the original continuous images. After performing, consider a plurality of planes perpendicular to the original continuous image, and extract the image obtained by projecting the result of the processing onto these planes as a section image. An image processing method for extracting an object movement locus according to item 1 or 2. 4. In the second to fourth steps or the sixth to eighth steps, when a plurality of moving objects exist in one image, trajectory cross-section images for each of them are simultaneously extracted. An image processing method for extracting an object movement locus according to any one of claims 1 to 3, comprising a step.