METHOD AND APPARATUS FOR DETECTING OBJECTS USING STRUCTURED LIGHT PATTERNS
An object detection system is provided that projects one or more patterns onto a monitored area, captures one or more live images of the monitored area, and detects objects that enter the monitored area by detecting changes in the one or more patterns in the live images. Such an object detection system may be less susceptible to dynamic lighting conditions, and more sensitive to object intrusions. One illustrative example projects a pattern of dots onto an area to be monitored, and captures images corresponding to the monitored area, comparing live images to reference images to determine whether an object has intruded a defined area. The area to be monitored does not consist solely of the area captured in the images and may include a volume illuminated by the pattern as well as a volume corresponding to the captured image area. Objects not in the field of view may be detected by the disclosed systems and methods. Several illustrative analytical methods are disclosed as well.
Latest HONEYWELL INTERNATIONAL INC. Patents:
This application is a continuation of U.S. patent application Ser. No. 10/465,267, filed Jun. 19, 2003, entitled METHOD AND APPARATUS FOR DETECTING OBJECTS USING STRUCTURED LIGHT PATTERNS, which is a continuation-in-part of U.S. patent application Ser. No. 10/052,953, filed Jan. 17, 2002, entitled METHOD AND APPARATUS FOR DETECTING OBJECTS, now U.S. Pat. No. 6,841,780, which claims priority under 35 U.S.C. §119(e)(1) to U.S. Provisional Patent Application Ser. No. 60/262,925, filed Jan. 19, 2001, and entitled OBJECT DETECTION USING MOIRÉINTERFERENCE, which are both incorporated by reference herein in their entirety. This application is related to U.S. patent application Ser. No. 09/716,002, filed Nov. 17, 2000, entitled OBJECT DETECTION, U.S. Provisional Application No. 60/275,879, filed Mar. 14, 2001, entitled SAFETY CAMERA, and U.S. patent application Ser. No. 09/981,928, filed Oct. 16, 2001, entitled OBJECT DETECTION, which are all incorporated by reference herein in their entirety.FIELD
The present invention relates to object detection, and more specifically, to object intrusion and/or presence detection within a monitored area or region.BACKGROUND
Motion detection and object detection systems are well known in the art. Frequently, such systems monitor a user-defined area to detect when an object enters or passes through a monitored area. Such systems typically include an image capture device (such as a video camera or still camera) capable of capturing an image of the monitored area and, if required, a device for digitizing the captured images. The digitized images are analyzed in an attempt to detect whether an object has entered the monitored area. There are many different known methods and algorithms for analyzing digitized images for determining when an object has entered a monitored area. One of the most common methods is generally referred to as a change detection method.
Change detection is often accomplished by examining the difference between a current live image and a reference image, where the reference image contains only the static background of the monitored area. A reference image can be thought of as a representation of the monitored area as it would appear if no transitory objects were in view. Change detection algorithms often take two digitized images as input and return the locations in the field of view where differences between the images are identified.
Object detection systems are commonly used in environments that have dynamic lighting conditions. For example, in industrial settings, moving shadows can be cast on a monitored area or region, which can cause significant changes in ambient lighting conditions. Many existing object detection systems, including those that use change detection algorithms to detect objects, can be challenged by such shadows and/or other dynamic lighting conditions.SUMMARY
The present invention overcomes many of the disadvantages of the prior art by providing an object detection system that is less susceptible to dynamic lighting conditions, and/or more sensitive to three-dimensional object motion and/or presence. This is preferably accomplished by projecting one or more patterns onto the monitored area, capturing one or more live images of the monitored area including the one or more patterns, and detecting objects in the monitored area by detecting changes in the one or more patterns in selected captured images.
In a first illustrative embodiment of the present invention, a pattern is projected onto the monitored area. One or more live images of the monitored area are then captured at selected times, and analyzed to detect changes in the projected pattern. The changes in the pattern may indicate a topographical change in the monitored area, and thus the entry or movement of an object in the monitored area. Because the pattern is projected onto the monitored area, changes in the ambient lighting conditions may have less effect on the efficacy of the object detection system. In some embodiments, the projected pattern is provided at a wavelength which will not be affected or at least substantially affected by ambient lighting. For example, a near infrared or infrared wavelength may be chosen since such wavelengths are not as strongly affected by visible light variations in the region of interest if the visible lighting in an area changes. The particular pattern may vary widely, and may be static or dynamic. Additional variations and embodiments are further explained below.
In another illustrative embodiment, an object detection system includes a step of providing an unequal illumination pattern to an area to be monitored, where the unequal illumination pattern contains a predictable pattern. At the same time, an image of the area to be monitored is captured, and data analysis is performed to determine whether the captured image contains the predicted pattern. The predictable pattern may be considered to be a form of data, and the monitoring of the area comprises a determination of the extent to which the data cast in the illumination pattern is altered as measured by the capture of the image. As such, the illustrative embodiment relies not only on the area to be monitored to provide data for analysis, but also uses the illumination pattern to observe additional data. The information in the illumination pattern does not necessarily arise from activity within the field of view of the image capture apparatus, which may be, for example, a camera. Thus events outside the field of view of the image capture apparatus may be monitored by the present invention.BRIEF DESCRIPTION OF DRAWINGS
The following description should be read with reference to the drawings wherein like reference numerals indicate like elements throughout the several drawings. The detailed description and drawings represent select illustrative embodiments and are not intended to be limiting. The figures are not necessarily shown to scale.
The present invention provides an object detection system that may be less susceptible to dynamic lighting conditions, and/or may be more sensitive to object motion and/or presence than prior art systems. The system is also sensitive to three-dimensional extraneous objects that intrude into the monitored area. This may be accomplished by, for example, projecting one or more static or dynamic patterns on the monitored area, capturing one or more images of the monitored area including the one or more patterns, and detecting objects in the monitored area by detecting changes in the one or more patterns in selected captured images.
The illumination source 2 is located above a monitored area 14, such as near a ceiling. The illumination source 2 illuminates the monitored area 14 with a desired pattern. The pattern may be generated by, for example, projecting through a patterned grating, projecting interference patterns where the interference fringes are formed by a phase or wavelength shift, projecting a pattern using a scanning mechanism, or any other suitable method. As shown in
For several embodiments of the present invention, a static or dynamic pattern may be generated having a number of dots or other spatially defined elements created by providing a collimated light source (i.e. a laser source) as illumination source 2, diffusing or splitting the output of the illumination source 2, and directing the diffused or split light output through an aperture array. The apertures in the array may vary in size or shape to improve the resultant pattern on the monitored area 14. For example, given a centrally located illumination source 2, apertures near the edges of an aperture array may be made smaller, since a greater distance may be covered and hence greater dispersion of the output portion passing through the smaller apertures will occur. If, instead, the illumination source 2 were placed on one side of the monitored area 14 (which configuration is not shown in
The illumination of the monitored area 14 may be modulated to prevent any adverse effects of continuous illumination over a long period of time. Such continuous illumination may, for example, cause physical changes in the surface beneath the monitored area 14. Modulation may also allow for the use of a higher intensity illumination source 2 without creating difficulties with heating of the monitored area 14.
The pattern may be static or dynamic. A dynamic pattern is one where the spatial position of the light areas and dark areas changes over time, and in general the movement is periodic in nature. One possible use of a dynamic pattern is to increase the effective detection resolution of the system by preventing an object from “hiding” between illuminated portions of a static pattern.
One way to realize a dynamic moire pattern is to project an interference pattern from a Mach-Zender interferometer, which may include a mirror on a piezo-actuator. The frequency of movement of the mirror may determine the frequency of interference fringe movement. A simpler dynamic pattern may also be realized by using a moving aperture array element. For example, with a centrally located illumination source 2, a pattern of dots may be used as an aperture array element to create a pattern; rotating or otherwise moving the aperture array element can then create a dynamic pattern.
It is contemplated that the monitored area 14 may be rectangular, round, or any other shape, as desired. As shown in
It is contemplated that the illumination source 2 may be any type of illumination source capable of projecting a desired pattern on the monitored area 14. Examples of suitable illumination sources 2 include an incandescent or fluorescent lamp with a patterned mask and illumination optics. Other examples of suitable illumination sources 2 include a lasing source such as a modulated light-stripe illumination module, or a laser diode source uniformly illuminating a patterned grating with the grating re-imaged onto the monitored area with an objective projection lens. A suitable illumination source 2 may also be an infrared source. Using a portion of the spectrum not ordinarily found in or subject to wide variation in the ambient lighting conditions, such as the near infrared, may help keep the projected pattern from being overwhelmed by ambient light, and may also help enhance the differentiation between the projected pattern and other sources of light in the environment. The image capture device may be a sensor (CCD, photodiode, or the like) that is attuned to a desired spectrum, such as the spectrum of the illumination source.
In a further embodiment, multiple wavelengths may be used simultaneously, where the use of the two or more wavelengths may help detect whether a change is a result of a variation in the ambient environment that affects one wavelength more than another. For example, if a change is observed at one wavelength but not the other, the change may be attributed to an ambient condition, rather than an intrusion. Such a variation may also be adapted depending upon the type of intrusions which are anticipated or which are permissible. For example, if an intrusion by an object having a known spectroscopic signature is to be allowed, then the use of two wavelengths may allow objects generally to be detected, while allowing objects having known and anticipated absorbing effects to be ignored. The multiple wavelengths may each be given distinct patterns or dynamic features as well.
The illumination source 2 preferably projects at least one pattern on the monitored area 14. The pattern used may depend on the particular application at hand. For example, in one illustrative embodiment, the pattern may be any pattern that has transitions between areas that have illumination (e.g. light areas) and areas that lack illumination (e.g. dark areas). Generally, the maximum distance between the centroids of illuminated area should be no more than one half the size of the smallest object for which detection is desired, although this is not required. Examples of suitable patterns included striped or checker board patterns where illuminated and non-illuminated areas alternate. Another suitable pattern is one in which a collection of dots are defined against a general background, with either the dots or the background being the illuminated area. It should be noted that the figures are not necessarily drawn to scale and the particular resolutions, sizes and features described below are merely illustrative and not intended to be limiting.
In an illustrative embodiment, once the monitored area 14 is illuminated the image capture device 4 captures a reference image of the monitored area 14. The reference image is preferably an image of the monitored area 14 with no transitory objects in the monitored area 14. The reference image is preferably stored, at least temporarily, in the image storage device 6. A new reference image may be captured periodically, if desired. Once a reference image is captured, the image capture device 4 may capture successive live images of the monitored area 14, preferably at a selected time interval. Like the reference image, the live images are preferably stored, at least temporarily, in the image storage device 6. The image storage device 6 may provide the reference image and the live images to the processing device 8 for processing.
The processing device 8 preferably analyzes the live images to detect changes in the illuminated pattern. In some embodiments, the monitored area is divided into a number of image segments called mask windows. The size of each mask window is preferably chosen so that it is no bigger than the approximate size of the smallest object for which detection is desired. While objects smaller than the mask window may be detected, the probability of detecting such objects decreases with object size. The position of the various mask windows may be chosen so that the entire area to be monitored is covered by overlapping mask windows. The image area that corresponds to each mask window may be analyzed separately for object detection, if desired. In addition, it is contemplated that the analysis method that is used to analyze the various mask windows may differ across the image, and the triggered response may vary depending on which mask window detects an object, if desired.
The comparison between a reference image and a live image can be accomplished in any number of ways. One method is to simply do a pixel-by-pixel comparison of the projected pattern in the images. If there is no entry or movement of objects in the monitored area, the projected pattern in the two images will substantially cancel out. However, if there is entry or movement of an object in the monitored area, part of the projected pattern shown in one image may be shifted or otherwise deformed relative to the pattern shown in the other image. A threshold value may be used to help determine if there is a sufficient difference between the reference image and a live image to indicate a detected object, as further described below.
Another method for comparing one image to another is to calculate a difference between the value of the brightness levels corresponding to the light areas of the pattern (such as in a mask window), and the value of the brightness levels corresponding to the dark areas in the mask window of the reference image. A similar calculation may be made for the mask windows of a live image. Whenever the second calculation is different from the first calculation by a specified amount, a change may be inferred. A further discussion of this and other methods for comparing images is included in co-pending U.S. patent application Ser. No. 09/716,002 entitled OBJECT DETECTION, which is incorporated herein by reference.
Yet another method for comparing one image to another is to measure a correlation between each pixel and some neighboring pixels and/or a correlation between selected features, and then compare the correlation values. Whenever the correlation values are different by a specified amount, a change may be inferred. Alternatively, or in addition, the image analysis may extract the moire spatial frequency and phase using a Fourier transform. Other image analysis techniques may also be used, such as, for example, unsharp masking, thresholding, contrast segmentation, filtering processing, skeletonization processing, multi-resolution analysis, deformable contour modeling, image clustering, morphology, etc. These comparison methods are meant to be only illustrative, and any suitable method may be used to compare selected characteristics of the images, depending on the application.
It is also contemplated that the reference and/or live images may be preprocessed before they are compared. For example, the reference image and/or live images may be provided to a filter that helps removes speckle, provides smoothing, changes overall intensity, or otherwise cleans-up the images. In one illustrative example, the images may be modified to reflect changes in ambient light intensity by, for example, taking into account the average received intensity across an entire image.
Changes that are detected in the illuminated pattern may indicate a topographical change within the monitored area 14, and thus entry or movement of an object in the monitored area 14. When an object is detected, the processing device 8 may sound an alarm, shut down the machine 18, and/or provide some other alarm or action. Images of the monitored area with the detected object present may be retained for subsequent analysis, and/or sent to a monitoring station if desired.
The reference pattern in
It can be seen that the pixels may not exactly correspond in shape or position to the regions 36, 38. In an actual system, the individual pixels would be sensed individually and either subtracted in analog or digital fashion, where each pixel is represented as a value of light received represented in terms of a voltage, for example for analog subtraction, or in terms of a number value generated by analog-to-digital conversion (ATD) for a digital subtraction. If subtraction occurs using analog methods, an ATD step may follow the subtraction to yield results as schematically demonstrated in
The steps for comparing a received image to a reference image are known in the art, and any acceptable methodology may be used. The method illustrated by
The area of the triangle MNO can be written in two ways:
Using FOV as the field of view, one writes:
Treating x as the location of the pixel going to the left on
Then the following results:
Where px represents the number of pixels in the field of view in the x-direction, and wx represents the imager size in the x-direction. It can be seen that as x increases (i.e. the disturbance occurs farther away from the center of the field of view), the pixel shift is reduced.
For example, given H=3.0 meters, h=40 millimeters, Y=1.0 meters, focal length=3.6 millimeters, imager size=4.8 mm, and pixels=480, then at x=0, the pixel shift is 2.88. Thus, an illumination feature (such as a portion of a dot) that would have been sensed at one location without the object having h=40 millimeters would be sensed 2.88 pixels away. Optimal detection will occur when the angle between illumination device and the image capture device is 90 degrees. As can be seen from the above formula for pixel shift, one would seek to maximize Y and minimize H to achieve better sensitivity; however, it should be noted that distortion will increase as Y gets larger relative to H, so there is a tradeoff to take into consideration.
As a further example, if the data processing system associated with a particular safety camera can reliably detect or sense a shift of a single pixel across a border from light to dark in a pattern, a pixel shift of 1.0 may be used as a detection threshold. Then the following formula may be used to determine how far above the surface 64 an object must be (or if the object lies on the surface 64, how thick the object must be) to be detected reliably for the illustrative system:
If H=3.0 meters, Y=1.0 meters, focal length=3.6 millimeters, imager size=4.8 mm, and pixels=480, using pixel shift of 1.0 and assuming that the region of interest which will be monitored has borders that are 1 meter away from the center of the camera field of view such that x is less than or equal to 1 meter, then the minimum value for h which will be reliably detected at any place in the region of interest is 1.8 centimeters. The actual values at any given location in the region of interest will vary, with even thinner objects detected toward the center of the region (as x goes to zero, h becomes smaller) More sensitive electronics or data treatment schemes may allow for use of lesser pixel shift minimum values to achieve higher overall sensitivity to objects which are quite near the ground. Note also that the 1.8 centimeter value is actually the minimum height or level for an object entering sensed in the region of interest. Thus a very thin item such as a piece of paper which enters the region of interest some distance greater than 1.8 centimeters above the surface 64 can be sensed regardless of how thick it is.
The analysis for
Then, using similar triangles again:
Also, the area for triangle utm may be written two ways:
Using focal length=s and image displacement=d, the similar triangles on opposing sides of a lens will yield:
Next, referring to
S=S1+S2=H*cos θ+X*sin θ
Using p as pixel displacement, w as imager size, and n as pixel count,
If the camera is aligned such that θ =0, the X term drops out and leaves:
This result is illustrated in
For an illustration, if height H=3.0 meters, object height h=40 millimeters, separation B=2.0 meters, focal length s=3.6 millimeters, imager size w=4.8 millimeters, and pixel count n=480, with θ =0 then p=3.2 pixels of pixel shift. With this newer formulation, using θ =0 allows a more simple determination of the minimum height at which an object can be detected, because the calculation does not require allowing for different pixel displacements depending on position. For example, regardless of the value of X, with θ =0, using p=1 as a threshold and with H=3.0 meters, B=2.0 meters, s=3.6 millimeters, w=4.8 millimeters, and n=480, the minimum height at which an object can be detected would be calculated by:
Which results in h=12.5 millimeters for a minimum height at which an object will be detected reliably using p=1 as a threshold for reliable detection by electronics.
The image capture apparatus 62 captures images across an area defined by the angle Φ0 from the position of the image capture apparatus 62 to the far edge of the area 64. As such the angle Φ0 may be defined by the following formula (note Xc is treated as negative when measured as illustrated):
Three angles 80, 82, 84 are defined by the imposition of an object with the height h. The first angle 80 is the initial angle formed at a location x with respect to the image capture apparatus 62 and a vertical axis, and can be defined by:
The second angle 82 is the larger angle formed by the new reflection from a height h and can be defined in two ways, the latter being the simpler:
The third angle 84 is the difference between these first two. Using these angles and the number of pixels N, a pixel shift p may be defined as:
Now, let z equal:
Using simple geometry, the following two equations are readily calculated.
To further simplify the expression, one may use the following expression:
This can be simplified to:
Making a substitution in the original equation for z above results in:
Now, substituting x+Δx+Δy for z in the equation for p given above yields:
Using the above equations for p and z one may readily determine a desired resolution for detection of an object having a predetermined height h. For example, using N=400, h=15 mm, H=200 cm, Xc=−5 cm, Xs=105 cm, W=100 cm, and x=50 cm, then z=50.83 cm and p=3.19 is the pixel shift that would be sensed at that location given that height. Thus the data processing equipment and/or programming would need to be set up to sense a pixel shift of 3.19 or greater to detect an object fifteen millimeters high in the middle of the area of interest.
In many applications the borders of the area of interest will be where such calculations are made, i.e. when x=0 or x=W. Using the same values as above, except for x, then p=3.44 for x=0 and p=2.69 for x=W. Making Xc=0 improves performance by increasing the lower of the pixel shifts (pixel shift farther away from the image capture apparatus 62), so that p=3.42 for x=0 and p=2.74 for x=W. To avoid wasting pixels, the field of view should correspond to the size of the area to be monitored. Selecting the longest focal length for the image capture apparatus 62 that covers exactly the area to be monitored allows for maximum resolution with minimum distortion.
Any of a wide variety of actions may be performed in response to detection of an intruding object either individually or in combination. For example, a machine in operation in or near the area being monitored may be turned off in response to a detected intrusion. An audible or visible alarm may be sounded. A message or signal may be sent to a remote location such as a police station or other authorities. Additionally, images may be recorded in a permanent or semi-permanent fashion in response to a detected object, for example, where images are captured and discarded in a FIFO or LIFO manner, once an object is detected subsequent images may be placed into a separate memory and not discarded.
Those skilled in the art will recognize that the present invention may be manifested in a variety of forms other than the specific embodiments described and contemplated herein. Accordingly, departures in form and detail may be made without departing from the scope and spirit of the present invention as described in the appended claims.
1. A method for detecting an object in a monitored area, the method comprising the steps of:
- illuminating the monitored area using two or more wavelengths to create at least one pattern on the monitored area;
- capturing a live image of the monitored area, including at least a portion of the at least one pattern; and
- detecting an object in the monitored area by comparing the live image to a reference image, wherein a change in the at least one pattern between the live and reference images indicates the presence of an object in the monitored area.
2. The method according to claim 1, wherein the pattern comprises a number of dots.
3. The method according to claim 1, wherein each of the two or more wavelengths produces a different pattern.
4. The method according to claim 1, wherein a change in the at least one pattern at both wavelengths indicates the presence of an object in the monitored area.
5. The method according to claim 1, wherein a change in the at least one pattern at either wavelengths indicates the presence of an object in the monitored area.
6. The method according to claim 1, wherein at least one of the wavelengths is in the near infrared or infrared.
7. A system for monitoring a volume of space, the system comprising:
- an illumination apparatus placed to illuminate at least a portion of a monitored area with two or more wavelengths; and
- an image capture apparatus placed to capture images of at least part of the monitored area, wherein the image capture apparatus is sensitive to each of the two or more wavelengths;
- wherein the volume of space monitored includes a volume corresponding to the space defined between the illumination apparatus and the monitored area; and
- wherein the volume of space monitored includes a volume corresponding to the space defined between the monitored area and the image capture apparatus.
8. The system according to claim 7, wherein the illumination apparatus is adapted to create at least one pattern using the two or more wavelengths and project the at least one pattern onto the monitored area.
9. The system according to claim 8, wherein the illumination apparatus is adapted to create and project a different pattern onto the monitored area with each wavelength.
10. The system according to claim 8, wherein the at least one pattern is not present in the monitored area in the absence of the illumination apparatus.
11. The system according to claim 8, wherein the illumination apparatus is adapted to project the at least one pattern onto a surface in the monitored area, such that an object in the monitored area will have the pattern projected onto its surface.
12. The system according to claim 7, wherein the portion of the monitored area illuminated by the illumination apparatus is not rectangular.
13. The system according to claim 7, wherein the portion of the monitored area extends around at least two sides of a machine.
14. The system according to claim 7, wherein the illumination apparatus illuminates at least a portion of a monitored area with a different pattern for each of the two or more wavelengths.
15. The system according to claim 7, wherein the portion of the monitored area does not include the pattern prior to the illuminating step.
16. The system according to claim 7, wherein at least one of the wavelengths is in the near infrared or infrared.
17. A method for detecting an object in a monitored area, the method comprising the steps of:
- creating at least one pattern using two or more wavelengths;
- projecting the at least one pattern onto the monitored area;
- capturing one or more live images of the monitored area at the two or more wavelengths, including at least a portion of the at least one pattern; and
- detecting an object in the monitored area by comparing the one or more live images to one or more reference images, wherein a change in the pattern between the one or more live and reference images indicates the presence of an object in the monitored area.
18. The method according to claim 17, wherein each of the two or more wavelengths produces a different pattern.
19. The method according to claim 17, wherein a change in the at least one pattern at both wavelengths indicates the presence of an object in the monitored area.
20. The method according to claim 17, wherein the monitored area does not include the pattern prior to the projecting step.
21. The method according to claim 17, wherein at least one of the wavelengths is in the near infrared or infrared.
International Classification: G06M 7/00 (20060101); G01J 5/02 (20060101); G08B 13/18 (20060101);