JP3581889B2 - Stereoscopic recognition method and apparatus - Google Patents

Description

[0001]
[Industrial applications]
The present invention relates to a method and apparatus for recognizing a three-dimensional image.
[0002]
[Prior art]
Conventionally, three-dimensional measurement is a method of irradiating a point light source or a line light source to an object, measuring a distance from an image forming position of reflected light from the object to the object, vibrating the lens little by little, and partially arranging the image receiving element The distance to the subject is calculated by measuring the distance by minimizing the degree of modulation in the camera, or by photographing the subject using two cameras and measuring the angle difference between the subject and the two cameras There are ways to do that.
[0003]
[Problems to be solved by the invention]
Although the above method is a simple method for measuring the distance to the subject, measuring the distance of the entire screen including the background has the following problems.
[0004]
FIG. 1 is a diagram illustrating a stereoscopic image recognition method of the present invention and a conventional stereoscopic image. FIG. 1A shows an example of a conventional method, and shows an example of the intensity of a captured rectangular image. Here, the image 1 is an image when the size of the subject is sufficiently larger than the size of each pixel.
(A) shows an image when the focus is achieved, and image 1 (b) shows an image when the focus is deviated. Intensity 1 indicates an intensity distribution of a cross section of image 1 along AB. When the image is in focus, the edge portion is sharp as shown in FIG. 7A, but when the image is out of focus, the edge portion is blurred as shown in FIG. If the size of the subject is sufficiently larger than the size of each pixel, blurring occurs at the edge portion, but the image intensity at the central portion of the image does not change.
[0005]
Image 2 shows a case where the size of each pixel is approximately the same as the size of image 2. As the image becomes smaller, the intensity at the center of the image is attenuated when the focus is deviated as shown by intensity 1 and becomes different from the intensity when the image is in focus. That is, in the image 1, when the focus is shifted, the edge portion is blurred, but the intensity of the central portion does not change. However, when the focus shifts in the image 2, the central intensity also shifts as shown in FIG. As described above, there is a case where the intensity distribution is not proportional to the distance depending on the size of the image only with the intensity distribution.
[0006]
An object of the present invention is to improve the inaccuracy of three-dimensional recognition based on the conventional intensity distribution, and to simply measure the distance of the entire screen including the background accurately.
[0008]
[Means for Solving the Problems]
In order to achieve the above object, the first means of the stereoscopic recognition method according to the present invention records, in an image memory, a distance of a subject position when an image is focused on each image photographed by changing a lens focal position. The number of the differentiated image is assigned a value by applying a nonlinear weight to the number of the differentiated image, and the obtained image is displayed as distance information.
[0010]
The second means performs threshold processing on differential images of a plurality of images obtained by photographing the same subject while changing the focal point of the lens, and continuously performs weighting processing with different weights on the images on which the threshold processing has been performed. Thereafter, an image obtained by adding all the weighted images is obtained and displayed.
[0011]
Further, the three-dimensional recognition device is an actuator that changes the focal position of a photographic lens of a subject, an imaging unit that captures a three-dimensional image of the subject, and an image that is captured by the imaging unit is subjected to differentiation processing by an arithmetic processing unit described later. An image memory for accumulating a differential image, sequentially assigning a number to the differential image and accumulating the differential image in the image memory, sequentially comparing values of specific pixels of the accumulated differential image, and calculating a differential image having a maximum value A three-dimensional object having an arithmetic processing unit for obtaining the distance information by repeatedly setting the number to the value of the specific pixel for all pixels, and an image display unit for displaying the distance information calculated by the arithmetic processing unit as an image In the recognition device, the imaging unit is configured by an incoherent / coherent conversion device, an optical contour extraction device, and a CCD imaging device .
[0012]
[Action]
According to the present invention, when the image 1 and the image 2 are differentiated in the horizontal direction as indicated by the differential 1 in FIG. , The intensity distribution as shown in the figure, but the edge portion has an intensity distribution corresponding to the defocus. Therefore, the distance to the image can be calculated irrespective of the size of the image by examining the lens position when the intensity of the edge portion is maximum or minimum.
[0013]
The differential 2 in FIG. 1 (3) is a value when the negative component of the differential 1 is inverted. In this case, both edges of the image are displayed with the same value.
[0014]
As described above, the image of the subject is photographed a plurality of times while changing the focus, and each image is differentiated, and the distance to the subject is obtained from the focal position when the value of the differentiated edge becomes the largest. Can be.
[0015]
That is, when an image of a subject is formed by a lens, an image in which a focused portion is sharpest can be obtained. When the formed image is differentiated, the shadow of the subject is removed and the image becomes a contour image of the subject. By paying attention to the contour of the subject and examining the lens position when the contour is most emphasized, the distance to the contour of the subject can be obtained.
[0016]
【Example】
FIG. 2 is a block diagram of an embodiment of an apparatus for performing the stereoscopic recognition method of the present invention. Reference numeral 21 denotes a photographing lens of a subject 27, reference numeral 22 denotes an imaging unit that captures a stereoscopic image of the subject 27, and reference numeral 23 denotes a lens 21. The actuator 24 to be driven sequentially assigns numbers to the differential images and accumulates them in the image memory 25, sequentially compares the values of the specific pixels of the accumulated differential images, and assigns the number of the differential image having the maximum value to that number. An arithmetic processing unit for repeatedly obtaining the value of the specific pixel for all the pixels to obtain distance information; 25, an image memory for storing a differential image differentiated by the arithmetic processing unit 24; The image display unit displays an image of the distance information calculated in 24.
[0017]
Next, the operation will be described. A plurality of images of the subject 27 are captured by the imaging unit 22 while the focal position is changed by driving the lens 21 by the actuator 23. The photographed image is read into the image memory 25, is subjected to a differentiation process by the arithmetic processing unit 24, and is stored in the image memory 25. Further, if a differential processing function is added to the imaging unit 22, the differential image can be directly stored in the image memory 25. Further, it is converted into distance information by the arithmetic processing unit 24 and displayed as a distance image by the image display unit 26.
[0018]
In the figure, assuming that the distance between the subject 27 and the lens 21 is a, the distance between the lens 21 and the imaging surface of the imaging unit 22 is b, and the focal length of the lens is f, the distance to the subject 27 is obtained by the following formula.
[0019]
(Equation 1)
1 / a + 1 / b = 1 / f
[0020]
(Equation 2)
a = bf / (b−f)
The method of obtaining the focal position when the value of the differentiated edge becomes the largest is performed as follows. Attention is paid to a specific pixel (a set of one pixel or a plurality of pixels) of an image, an image having the maximum value of the specific pixel of each image is selected, and a distance to the pixel is calculated from the imaging conditions of the image. By calculating the distance for each pixel or group of pixels, an image of the distance of the subject can be obtained.
[0021]
Furthermore, by complementing the portion surrounded by the differentiated edge using the image when the edge value is maximized, the image can be reproduced more faithfully to the subject.
[0022]
Next, a method of calculating distance information performed by the arithmetic processing unit 24 will be described. FIG. 3 is a diagram in which n differential images stored in the image memory 25 are arranged. By configuring the number of the differential image having the largest value among the n specific pixels for the specific pixel (i, j) common to the n differential images as new image data, the edge of the subject is obtained. An image corresponding to the distance can be obtained.
[0023]
Further, the distance of the subject position when the image is focused on each image taken by changing the lens focal position is made to correspond to the number of the differential image recorded in the image memory, and the number of the differential image is weighted nonlinearly. The obtained image is displayed as distance information, and threshold processing is performed on differential images of a plurality of images obtained by photographing the same subject by changing the lens focal position. The weighted images may be successively subjected to different weighting processes, and thereafter, an image obtained by adding all the weighted images may be obtained and displayed.
[0024]
If a value obtained by converting the number of the differential image into the distance to the subject calculated from the change in the focal position of the lens is used, image information representing the distance of the edge of the subject can be obtained. In the object, the uniform portion where there is no edge is zero in the differential image, so in the closed edge, the enclosed portion is complemented by the distance value indicated by the edge portion, so that an image closer to the actual Obtainable.
[0025]
Next, a method of displaying distance information will be described. When displaying the distance information obtained as described above on the image display unit, by changing the brightness in accordance with the distance, it is easy to recognize the perspective. As an example, by displaying the closest distance as the brightest and the farthest distance as the darkest on the image display unit, the subject can be displayed as if it were irradiated with illumination light in the dark. In order to colorize this, the most recent distance is displayed warmer and brighter on the image display unit, and the farthest distance is displayed cooler and darker, so that the subject emerges as if it were illuminated more effectively. be able to.
[0026]
Further, the imaging unit 22 using optical processing will be described. If the image is coherent light, differential processing of the image can be performed using light interference based on coherence. As a method for converting an incoherent light image into a coherent light image like a normal camera, there is a device called an optical neuron device (Reference: K. Aiyama, JJAP, Vol. 30, No. 12B, 3887-3892, 1991). When such a device is used, an incoherent image is first converted into an electric signal by the a-Si: H layer, and a difference in the intensity of the electric signal is given to the ferroelectric liquid crystal layer to change the transmittance. By irradiating coherent light from the opposite side of the liquid crystal layer and modulating the intensity of the reflected light, the image can be converted into an image composed of coherent light. By combining this device with the technology of an optical contour extraction device, it is possible to obtain a differential image of a normal lens-focused image by an optical method. What is necessary is just to form this optical differential image with a CCD image sensor.
[0027]
FIG. 4 shows a block diagram of an optical differentiation processing device according to this method. An image is formed on the optical neuron device 42 by the lens 41. The light is converted into a change in the transmittance of the liquid crystal by the optical neuron device 42. The image is converted into a coherent image by the laser light source 43. The coherent image is subjected to a differentiation process by the optical contour extraction device 44, and is converted into an electric signal by the CCD image pickup device 45. In this way, an optical differentiation process is performed.
[0028]
【The invention's effect】
As described above, the present invention uses a single imaging system to move a lens and an imaging element relatively to capture a plurality of images, differentiate each image, and maximize the value of each pixel. By using the resulting image as distance information, a distance image can be obtained by a single imaging system. As a result, with a simple imaging system, an image at a distance over the entire drawing can be obtained and further displayed.
[Brief description of the drawings]
FIG. 1 is a diagram showing a stereoscopic image stereoscopic recognition method of the present invention and a conventional stereoscopic image.
FIG. 2 is a block diagram of an embodiment of an apparatus for performing the stereoscopic recognition method of the present invention.
FIG. 3 is a diagram in which n differential images stored in an image memory of FIG. 2 are arranged.
FIG. 4 is a block diagram of an optical differentiation processing device.
[Explanation of symbols]
1, 2, ... image, 21, 41 ... lens, 22: imaging unit, 23: actuator, 24: arithmetic processing unit, 25: image memory, 26: image display unit, 27: subject, 42: optical neuron device, 43 ... Laser light source 44 Optical contour extraction device 45 CCD image pickup device

Claims (7)

The distance of the subject position when the image is focused on each image taken by changing the lens focal position is made to correspond to the number of the differential image recorded in the image memory, and the number of the differential image is multiplied by a nonlinear weight. A stereoscopic recognition method , wherein the obtained image is displayed as distance information . Threshold processing is performed on differential images of a plurality of images obtained by photographing the same subject while changing the lens focal position , and weighting processing is performed on the images on which the threshold processing has been performed with different weights, and then weighting is performed. A three-dimensional recognition method characterized by obtaining an image obtained by adding all the images obtained and displaying the obtained image. The method according to claim 1, wherein the distance information is displayed as brightness information . 4. The stereoscopic recognition method according to claim 1 , wherein the closest distance information is displayed as the brightest and the farthest distance information is displayed as the darkest as the distance information . Said distance brightest recent distance information in warm as the information, according to claim 1, 2 or 3 steric recognition method, wherein that the darkest to color display the longest distance information cooler. An actuator for changing a focal position of a photographic lens of a subject, an imaging unit for capturing a three-dimensional image of the subject, and an image for storing a differential image obtained by differentiating an image captured by the imaging unit by an arithmetic processing unit described later A memory, sequentially assigning a number to the differential image and storing it in the image memory, sequentially comparing the values of the specific pixels of the stored differential image, and assigning the number of the differential image having the maximum value to the specific pixel. In a three-dimensional recognition device having an arithmetic processing unit that obtains distance information by repeatedly performing the setting of a value for all pixels, and an image display unit that displays an image of the distance information calculated by the arithmetic processing unit, A three-dimensional recognition device , characterized in that the part comprises an incoherent / coherent conversion device, an optical contour extraction device, and a CCD image pickup device . 7. The stereoscopic device according to claim 6 , wherein the input image is converted into a coherent image by the incoherent / coherent conversion device, the contour is emphasized by a diffractive optical element and a one-dimensional grating, and a differential image is photographed by a CCD image sensor. Recognition device .

Images