Image analysis – Pattern recognition – Feature extraction
Reexamination Certificate
2000-07-10
2004-04-13
Ahmed, Samir (Department: 2623)
Image analysis
Pattern recognition
Feature extraction
C701S023000, C701S028000, C701S040000, C706S020000, C706S052000
Reexamination Certificate
active
06721453
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Cross Reference to Related Applications
The present application is related to “A Method And Apparatus For Determining The Quality Of An Image Of An Agricultural Field Using A Plurality Of Fuzzy Logic Input Membership Functions”, by Benson, et al., filed on an even date herewith.
2. Field of the Invention
This invention is in the field of automated agricultural guidance systems, specifically relating to processing an image of a field to determine a boundary between the cut crops and the uncut crops.
3. Description of the Related Art
Harvest is perhaps the most important stage of production agriculture. Within the United States, the corn and grain harvest is highly mechanized, with combines and forage harvesters performing the majority of the harvest. Tractors, trucks and wagons are used to transport the harvested crop from the field to the silos and market. Harvest brings both expectation and a tinge of nervousness. Harvest is a time when the farm and community come together, working long hours to bring in the crop.
Unfortunately, these characteristic long hours during harvest season lead to operator fatigue, thereby posing safety and quality control problems. Operators have to manipulate yield monitors, vehicle speed, header height, reel speeds and a host of other controls while they operate the combine. Vehicle automation has brought automatic header height control and improved ergonomics, but the task is both repetitive and complicated.
Automation is ideally suited to repetitive tasks. Agricultural machinery operation combines both repetitive operations (row following or surface coverage) and unique operations (road travel and a myriad of farm tasks). The open and changing environment combined with the safety and robustness requirements add to the difficulty for agricultural automation. Automation can mean simply assisting the driver or it can mean complete autonomous operation of the agricultural vehicle.
Researchers at several institutions have developed methodologies for autonomous agricultural vehicles (Reid and Searcy, 1991 O'Connor et al., 1995, Hoffman et al., 1996, Callaghan, et al., 1997). Different approaches have included mechanical guidance and sensor-based systems. Mechanical systems utilize feelers to detect furrows or rows of plants and the feeler position is converted to a guidance signal. Sensor-based systems rely on electronic sensors to determine the location of the vehicle either locally or with respect to an established coordinate frame.
Several manufacturers have developed and marketed mechanical systems. Sato et al. (1996) demonstrated a feeler based guidance system for Japanese style rice combines. Within the United States, factory and after-market row guidance systems are used in cotton harvesting.
Sensor-based technology has become attractive as sensor capacity has increased while prices have decreased. External sensors, including GPS and field based systems, and internal sensors, including machine vision and inertial systems, have been used for agricultural guidance. Researchers often combine the sensors to provide increased functionality.
For example, tractor guidance must take place regardless of the field conditions. During early season or preseason operations, the plants have not yet reached sufficient maturity to provide a visual reference for guidance. In this case, non-visual sensors such as GPS and inertial systems can provide the guidance signal. However, while many combines are fitted with GPS receivers for field monitors, the receivers do not have sufficient accuracy for guidance.
One of the primary keys to a successful automated agricultural guidance system is to extract certain features of interest from an image of the field. In particular, researchers have developed methodologies to extract the features of crop rows from an agricultural scene (Reid and Searcy, 1991). Generally, the algorithms previously developed assume that the camera is located above and roughly parallel to the camera orientation. For row crop guidance (for example, cultivation), the vehicle is aligned with the rows and the crop is shorter than tractor mounted camera.
Unfortunately, the situation dramatically changes when the camera is used to guide a combine. In particular, the feature of interest is no longer the parameterization of the crop rows, but rather the edge between the cut and uncut crop. In this situation, the harvester heads are up to 9 m wide, thereby necessitating the use of a wide-angle lens to see the entire head swath.
Hoffman, et al., (1996) developed an automated harvester (Demeter) for alfalfa and other field crops. In the Demeter project, cameras were installed on both sides of the cab. However, as the height of the camera is increased, perspective shift becomes a problem. The perspective shift for a given elevated camera location increases with the width of the head. The head width and the crop height make it difficult to accurately detect the cut/uncut boundary in the field using a single top mounted camera.
Due to the difficulties associated with a single top mounted camera, there is a need for an automated agricultural guidance system that determines the boundary between the cut/uncut crops. In particular, there is a need for a reliable method of processing an image of the field to determine the boundary between the cut/uncut crops. In multiple camera systems, the cameras are installed on each end of the head. These head-mounted cameras allow each of the cameras to directly see the cut/uncut boundary without the perspective shift issues of a cab-mounted camera. The head-mounted camera, however, sees a drastically different image than the cab-mounted cameras. Therefore, a new image processing method is needed to accommodate the change in the scene parameters.
SUMMARY OF THE INVENTION
Accordingly, one object of this invention is to provide a method of detecting an edge between a cut crop and an uncut crop in a field. The method includes processing at least two scanlines of pixel data based on an image of the field. A field boundary is generated and one or more characteristics of the image are calculated after each scanline is processed. Finally, the decision whether to continue processing is determined based on the first characteristic, the second characteristic and the vertical location of a particular scanline being processed in relation to the image.
Another object of this invention is to provide an agricultural vehicle configured to be guided through a field of crops by an automatic guidance system. The agricultural vehicle includes at least one camera mounted on the agricultural vehicle, an image processor and a central processing unit. The image processor is configured to process at least two scanlines of pixel data based on an image of the field. The central processing unit is configured to generate a field boundary, calculate a first characteristic of the image after processing each scanline of pixel data, calculate a second characteristic of the image after processing each scanline of pixel data, and determine whether to continue processing the pixel data based on the first characteristic, the second characteristic and the location of a particular scanline being processed in relation to the image.
Yet another object of this invention is to provide a method of detecting an edge between a cut crop and an uncut crop in a field by processing at least two scanlines of pixel data based on an image of the field. A field boundary is generated that divides the cut crop from the uncut crop. A first characteristic and a second characteristic of the image are calculated. The decision whether to continue processing the pixel data is based on a convergence of the first and second characteristics to a first value and a second value, respectively, wherein the first and second values are compared to a predetermined covariance threshold.
REFERENCES:
patent: 5442552 (1995-08-01), Slaughter et al.
patent: 5704200 (1998-01-01), Chmielewski et al.
patent: 5828971 (1998-10-01), Diekhans et al.
patent: 5857539 (1999-0
Benson Eric R.
Reid John F.
Zhang Qin
Ahmed Samir
Le Brian
Stader John William
The Board of Trustees of the University of Illinois
Webb Collin A.
LandOfFree
Method and apparatus for processing an image of an... 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 processing an image of an..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for processing an image of an... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3201149