Image analysis – Applications – Vehicle or traffic control
Reexamination Certificate
2000-07-10
2004-03-30
Mariam, Daniel (Department: 2623)
Image analysis
Applications
Vehicle or traffic control
C382S199000, C382S307000, C348S113000, C701S023000, C701S028000
Reexamination Certificate
active
06714662
ABSTRACT:
CROSS-REFERENCES TO RELATED APPLICATIONS
The present application is related to “A Method and Apparatus for Processing an Image of an Agricultural Field” filed by Benson, et al. on an even date herewith.
FIELD OF THE INVENTION
This invention relates generally to processing of sensor signals. More particularly, it relates to the determination of the quality of sensor signals received from a sensor in deciding whether to use them or not to use them in the automatic control of an industrial process. Even more particularly, it relates to the determination of the quality of a camera image of an agricultural field, and whether data extracted from that image should be used to guide the vehicle through the field.
BACKGROUND OF THE INVENTION
All sensors are noisy. By noise, we mean anything that causes the sensor signal to depart from an ideal expected value. This degree of noise or error can be substantial in many sensors.
In most industrial applications, quick and dirty rules of thumb are used to determine whether a sensor is providing signals that are truly representative of the physical phenomenon being monitored. For example, an electronic thermometer being used to monitor the temperature of water boiling at atmospheric pressure and temperature, wherein the temperature is used to control the electricity provided to a heating element, might be configured to reject the signal from the temperature transducer if it exceeds 240° F. The underlying assumption is that the transducer is in water and the water is boiling at atmospheric temperature and pressure. Therefore, a temperature of 240° F. would indicate an impossible state of events. At standard atmospheric pressure, water should never rise above 212° F. It would be appropriate to hard limit a temperature sensor to a temperature slightly above 212° F., since such a temperature reading from the sensor would clearly be an error.
Sensor readings, however, and the systems in which they are employed, are often not so easily handled. For example, using the above situation, if the water was being heated in a pressure vessel, it is quite possible that a temperature of 240° F. could properly indicate the temperature of boiling water. The likelihood that a temperature of 240° F. was incorrect would vary, depending upon the pressure of the pressure vessel in which the water was being boiled. The larger the pressure, the more likely that a temperature of 240° F. was still acceptable for use in the temperature feedback control system.
The determination of the acceptability, or the unacceptability of a signal, i.e. the likelihood that the signal accurately represents a true physical phenomenon, may therefore depend on many different machine or process perameters, most significantly including any change or difference from a preceding measurement of the sensor signal. What do we mean by this? Imagine that in the example above, the heating element, heating the water to a boiling point is only capable of raising the temperature by 5° F. per minute. If the temperature sensor indicated that the temperature of the water increased by 5° in one second, this radical and sudden change (e.g., if changed from 180° F. to 185° F.), would indicate that the sensor signal was incorrect.
Thus, to provide more accurate control of an automated system, especially in systems where there is a significant amount of noise, it would be beneficial to determine whether a particular sensor reading or a particular signal value, acceptably represented the underlying measured physical phenomenon by taking into account a plurality of different factors, and possibly different measurements, that are not limited to fixed limits.
One class of applications are particularly well suited to this technique, in particular, machine vision systems for guidance or control used in agricultural or construction applications. In recent years, construction and agricultural vehicles, either tracked or wheeled, have been equipped with camera systems that attempt to extract information indicative or characteristic of the surrounding environment from a series of images taken by a camera that is typically fixed to the vehicle. We normally think of computers in the current day as being extremely fast. This is not so in the case of image processing, especially image processing in chaotic visual environments with a great deal of random colors, intensity, and shapes. Such environments include, agricultural applications in which the camera is directed towards the ground where the crops adjacent to the vehicle, and construction environments in which the camera is directed to churned-up earth in the areas adjacent to the vehicle. In both these situations, the shear randomness of much of the image requires extensive processing to extract a signal that can be used to control operation of the vehicle. Any control algorithm that extracts characteristics of the environment from the image itself will inevitably have a huge component of noise. Yet, since the information extracted from this sequence of images is typically used for real time control of the vehicle, it is essential that this image processing happen extremely fast, on the order of several times a second. Inevitably, this means that any extraction of characteristics of the image and hence of the surrounding scene must happen extremely fast and must be (at least at the current time) relatively crude and noisy.
One specific vehicular application in which the characteristics extracted from the image may be noisy is that described in the patent application entitled “A Method and Apparatus for Processing an Image of an Agricultural Field”, filed contemporaneously herewith. That application is incorporated by reference in its entirety herein for any purpose.
In that application, a series of images are taken by a camera mounted on the harvesting head of a combine. This camera is pointed directly ahead of the vehicle in the direction that the harvesting head travels as it harvests crops. The camera is aligned such that it travels through the field substantially on top of the boundary between the previously cut and uncut regions of the field. In other words, one side of the image shows the portion of the field with the uncut crop and the other side of the image shows the portion of the field with the harvested crop.
FIG. 3
is a typical image taken by a camera in accordance with this invention showing the boundary between the cut and uncut regions extending as a line from the lower left corner of the image frame to the upper central portion of the image frame. This line is approximated in
FIG. 4
as described in the above-identified specification.
As each image from this camera is processed, the central processing unit extracts two characteristics of the scene from each image: a numeric value indicative of the boundary line and a numeric value indicative of the intercept of the line. Thus, image processing extracts two numeric values that collectively define a line extracted from the camera image and representative of the physical boundary between the cut and uncut portions of the field. For example, the preferred harvest crop is corn which is planted in rows. The header travels down these rows in which these lines also correspond to the row of planted crop.
Given the rather precise boundary shown in
FIG. 3
, one might expect that it would be relatively easy to extract the boundary line, the row line from images of like quality. That is true. Unfortunately, in real life, there are a variety of factors, such as variations in leaf growth, insects, dirt, fog, and dust, as well as the curvature of the boundary line in the field that cause the slope and intercept to vary, sometimes considerably, from image frame to successive image frame. For example,
FIG. 5
shows an image in which a corn leaf was flashing in front of the camera at the instant the camera took its picture. Superimposed on top of the image is the characteristic boundary line that would have been extracted from this image. As can be seen, the slope and intercept of this line varies considerably
Benson Eric R.
Reid John F.
Zhang Qin
Case Corporation
Henkel Rebecca I.
Mariam Daniel
Maurer Brant T.
LandOfFree
Method and apparatus for determining the quality of an image... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for determining the quality of an image..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for determining the quality of an image... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3238246