Patent Grant
10932445
The invention relates to the field of natural resource management and more specifically to the automated identification of individual subjects in a target population to perform population and/or biological monitoring.
It is a problem in the field of natural resource management to perform population and/or biological monitoring to uniquely identify an individual subject in a target population. In particular, while work has been done to differentiate among various species in a study area, in order to produce a gross count of the members of each species in the study area, there is a lack of any system that can automatically and accurately differentiate among the individual members of a particular species in a target population.
It is well known that the species in various taxa have external characteristics that are useful in uniquely identifying an individual member of the species. This process of identifying an individual member of the species is commonly done on a manual basis, more commonly in the management of mammals, where the herd managers can visually distinguish among members of a particular herd. This process is experientially based in that the herd manager over time remembers various characteristics that distinguish members of the managed herd. However, this knowledge is typically not recorded in any formal manner and is therefore not simply transferable to others. In addition, the knowledge base of a particular herd is not extensible to any other herd and the data collection and member recognition process must be laboriously reinitiated when a herd manager encounters a new herd or even additions to an existing herd. Finally, the identification process is not rigorous, in that it is subject to the personal determinations of the herd manager, may not be based on scientifically accurate distinctions, and may be limited to only gross identifying characteristics that are easily recognized.
In the case of very large populations of subjects, where the data collection and subject distinguishing processes are beyond a simple visual and human-based memory process, there is presently no method to enable the recognition of an individual member of the population. In these populations, the process in identifying the individual must be automatic, rapid and accurate and extensible to any size population or any population among a larger set of populations. In addition, any recognition process must be able to account for temporal and growth changes in the members of the population, such as seasonal variations and changes associated with aging. In addition, in order to distinguish among members of a very large population, a scientifically rigorous process must be used and is typically based on subtle and/or complex differentiable characteristics.
One aspect of the invention relates to a system for uniquely identifying subjects from a target population which operates to acquire, process and analyze images of subjects to create data which contains indicia sufficient to uniquely identify an individual in a population of interest. The process used by the system in identifying the individual is automatic, rapid and accurate.
There are various species of interest in the field of natural resource management which have external characteristics that are useful in uniquely identifying an individual member of the species. These characteristics can be any recognizable feature or pattern of features, which in total have sufficient specificity to create a “signature” that is unique to the selected individual. These selected characteristics are typically members of the class of identifying characteristics that include: overall size, color, color/spot/stripe features, overall patterning, size and shape of physical features, features of horns/tusks/antlers, maturity-juvenile/adolescent/adult, scars, deformities, and the like.
The present system for uniquely identifying subjects from a target population implements an automated, image-based process that captures data indicative of a selected set of external characteristics for subjects that are members of a target population of a predetermined species. The captured data is then processed to create a database record which is associated with the selected subject and which can be used to uniquely identify the selected subject. The subject identification process is repeatable, so any subject can be accurately and repeatedly mapped to their associated database record, which database record can then be updated to reflect changes in the subject and/or presence of the subject in a target population.
The system for uniquely identifying subjects from a target population operates to acquire, process and analyze images to create data which contains indicia sufficient to uniquely identify an individual in a population of interest. This system implements an automated, image-based process that captures data indicative of a selected set of external characteristics for subjects that are members of a target population of a predetermined species.
The subject identification apparatus 100 is comprised of a number of elements. In particular, a subject positioning apparatus 111 can be used to present the selected subject in a predetermined orientation and/or position to enable the imaging apparatus 112 to capture image data which enables the subject identification apparatus 100 to uniquely identify the selected subject. Alternatively, the present system can be used to process images that would be taken by field biologists of marginally-restrained (hand-held) or even free-ranging animals such as free-ranging giraffes or hand-held fish. Alternatively, a “lay” labor force of anglers can be used to take pictures of the fish they catch and release or “bird-watchers” can photographically verify the birds they view and add to their “life-lists”. The captured image data is transmitted from the imaging apparatus 112 to a processor 113, which executes a database routine 115 that serves to create a database record for the selected subject and store the database record in a memory 114, where the database record includes the captured image data and/or processed image data that comprises a set of indicia that serves to uniquely identify the selected subject. The processor 113 can also store additional data in memory 114, such as: date, time, location, subject measurement data (e.g. weight, age, physical measurements, automatically received GPS data, etc), and the like as part of the database record for the selected subject. The processing of the captured image data can be in real time as the selected subject is positioned in the imaging apparatus 112 or off-line, after the selected subject is released from the imaging apparatus 112.
Furthermore, the imaging apparatus 112 can consist of a number of elements that are cooperatively operative to generate the captured image data. These elements are a function of the species of subject that is to be imaged as well as the selected set of characteristics that are to be monitored to create the signature that uniquely identifies the selected subject. Within these general considerations, it is expected that there are numerous implementations possible for the imaging apparatus 112 and the description of any such apparatus herein represents an illustration of such apparatus and should not be construed as any limitation of the inventive concepts described and claimed herein. Therefore, at an architectural level, it is expected that the imaging apparatus 112 would typically include a subject illumination apparatus 121 to provide the desired illumination and/or contrast, if it is determined that the ambient light is insufficient, to enable the subject identification apparatus 100 to accurately determine the presence of and perform measurements of the selected characteristics for this species of subjects. In addition, an image capture device 122, such as a digital camera or other such image rendering apparatus, is used to determine the presence of and perform measurements of the selected characteristics for this species of subjects. Optionally, a display device 123 is provided to enable the operator to view and monitor the operation of the image capture device and controls 124 can be provided to further enable the operator to regulate the operation of the image capture device 122 and/or the positioning of the subject within the subject positioning apparatus 111.
System Operation—Image Capture
The typical imaging process is initiated at step 201, where a subject is selected from the population and, at step 202, the subject is positioned or oriented to enable a clear view of the subject. At least one image of the subject is captured by image capture device 122 at step 203, of which at least one image is typically in the form of a digital image. The image can optionally be reviewed by an operator on display device 123 at step 204 to ensure that there is sufficient quality of the image and number of features of the desired characteristic contained in the image to enable the subject identification process to accurately render an identification of the subject. If the quality and/or content of the captured image are inadequate, processing returns to step 202 (or 203) and the image capture process is reactivated. If the image quality and content are adequate, a determination is made at step 205 whether the subject is simply to be added to the database as a newly acquired image or whether it must be determined whether the subject has previously been entered into the database. If this is a newly acquired image, the subject is then further processed at step 206, additional data, if any, associated with the further processing entered into the database in a database record associated with this subject and the subject is then released at step 207. The data generated from the image generation process is stored in memory 114 at step 208 for processing at step 209 either contemporaneously with the image capture process or off line as a background process. In fact, if the image processing is contemporaneously done with image capture, the image processing step can optionally signal the operator when a successful image processing has been effected and the subject can be released.
If it is determined at step 205 that this subject imaging process may include subjects already recorded in the database, then it must be determined whether this subject has previously been entered into the database, and processing advances to step 210 where the data generated from the image generation process is stored in memory at step 211 for processing at step 212, as described below. The image processing at step 213 either returns an indication that the subject is a newly acquired subject, at which point processing advances to step 206 as described above, or an indication that the subject matches an existing database record for this subject, at which point the existing database record associated with this subject is identified and all additional data generated in this process is appended to the existing database record. Processing now advances to step 206 as described above.
Example of Selected External Characteristics of a Selected Species (24) As noted above, it is well known that the species in various taxa have external characteristics that are useful in uniquely identifying an individual member of the species. These characteristics can be any recognizable feature or pattern of features, which in total have sufficient specificity to create a “signature” that is unique to the selected individual. These selected characteristics are typically members of the class of identifying characteristics that include: overall size, color, color/spot/stripe, overall patterning, size and shape of physical features, particularities of horns/tusks/antlers, maturity-juvenile/adole-scent/adult, scars, deformities, and the like. The features within a characteristic are the specific unique implementations of the characteristic.
As an example, the rainbow trout (Oncorhynchus Mykiss) has the characteristic color/spot/stripe which has as its set of features the details of the colors/spots/stripes, for example, the location and/or size and/or color of the spots. In the characteristic of overall patterning, the features are not individual datum points, such as location of a spot, but instead are the pattern details. For example, the brook trout (Salvelinus fontinalis) has as one of its visual characteristics vermiculations on its back. The pattern of vermiculations does not have ‘spots’ per se, and unless considered at the discrete pixel level, it does not have ‘multiple datum points.’ In this case, the totality of the differences in intensity is the ‘pattern’ that distinguishes subjects. The plurality of features in this instance is therefore the plurality of differences in intensity.
Thus, an external characteristic, that contains a sufficient number of features to provide a unique signature for the subject, can be used to identify the subject. The definition of “characteristic” and “features” are a function of the species and these terms are used herein to provide a consistent structure in describing the operation of the present system for uniquely identifying subjects from a target population.
In order to illustrate the operation of the present system for uniquely identifying subjects from a target population, a species of fish—rainbow trout (Oncorhynchus Mykiss)—is used as an example. In this case, one characteristic of interest is the spot pattern on the fish, and the features of this characteristic are the plurality of spots that are imaged. Subjects can be located or confined in a monitoring area or manually positioned on an image platen to thereby capture an image of the subject or pertinent section of the subject that contains the desired characteristics. For example, fish can be captured using fyke traps, electro fishing, seines, trawls, angling and the like.
1.) Fish as separated from background
(function of background/background color/lighting)
2.) Components of the head:
Given this ordered priority of components, a determination must be made regarding the characteristics captured by the image that are required to enable the subject identification apparatus 100 to define a signature that uniquely identifies the subject. This determination includes the use of color in the image as an alternative to gray scale intensity. With color, the image data can be filtered to thereby enhance the distinction between datum points of the features of the selected characteristic(s) and non-essential image data. Other components of the features can include: size, location of the spot on the image, roundness—shape of the spot, other features. All of these features must be oriented in a three dimensional space using the imaging features: roll, pitch, yaw. Therefore, in addition to determining image characteristics, the subject must be oriented using either automatic positioning or manual positioning—which typically entails projecting a superimposed grid on the subject to facilitate manual positioning of the subject with relation to the image capture apparatus.
System Implementations for a Selected Species
In
In
System Operation—Image Processing
The database is populated with a plurality of data records, each of which comprises the image data and/or the processed image data. The matching technique is often referred to as “template matching” which uses autocorrelation as a method for extracting weak signals from a noisy channel. Many image processing techniques are devoted to the same principle—extracting the “signal” from “noisy” image. The autocorrelation process used in signal processing is exactly (mathematically) the same process applied to images. In the case of images, it is applied to each line of the image both horizontally and vertically. For example, in a 1K×1K image, the resulting autocorrelation is a (2K−1)×(2K−1) matrix/image, made up of the single-dimensional (signal processing-type) of each horizontal and vertical row/column of pixels.
)
At step 305, the operator can resize the Area-Of-Interest to a lower resolution to facilitate increased speed of processing the captured image data. The typical autocorrelation process requires 2n.sup.2 floating point operations and therefore a reduction in the number of features (spots) comprising the selected characteristic (spot pattern) that must be identified significantly improves the processing efficiency. In the specific case of the Rainbow trout, there is a lower limit to the uniqueness of the spot pattern and discriminating thresholds become difficult to determine with fewer than three spots if only a correlation process is used. However, combining correlation with additional image data processing features can improve the signature determination and processing speed. Therefore, at step 306, an additional process can be activated, such as determining specific optical centers of each spot (exact x, y location of the optical center of the spot in the Area-Of-Interest), which elevates the discriminatory ability of the system as a whole. Augmenting the correlation process with additional image data processing routines represents species-specific “tweaking”. Thus, components of the features of the selected characteristic(s) can be used as well as more than one characteristic to obtain the unique signature for the subject.
At step 307, the processor 113 of the subject identification apparatus 100 accesses the database of Area-Of-Interest data for this species, using database process 115 and selects appropriate data records from the database of data records for comparison with the presently determined pattern for the selected subject. At step 308, the processor performs the correlation operation 116 to determine the correspondence between each selected database record representing a known pattern and the presently determined characteristic for the selected subject. When comparing two subjects' characteristics using this method, the Area-Of-Interest for the known characteristics and the presently determined characteristic do not have to be exact, only “close”, since the correlation technique allows a certain amount of “flexibility” in selecting the exact Area-Of-Interest and still provides excellent matching ability. At step 309, the processor 113 determines whether the resulting correlation coefficient is below the species-specific pre-selected values (threshold), resulting in a ‘match’ with an existing known pattern for a previously determined subject as indicated at step 310, or a failure to match, resulting in the present subject being added to the database at step 311.
The characteristic(s) being used for identification purposes need only be discernable from the background. The above outlined process, using a spot pattern on the fish as the characteristic, requires that a ‘mask’ be made that allows “extraction” of the individual features (spots) and the characteristic (spot pattern) from the background. This process is generally referred to as “feature extraction”. Identifying “real” spots is no small matter, and a number of techniques can be used to enhance this process. In particular, using the electromagnetic frequency domain (color) to assist the feature extraction process is a possibility in some species where the frequency response (the “transfer function” in the electromagnetic domain) of the photographic sensor is selected or a spectral filter is used to isolate “real” features and reject the background or anomalies that could be mistaken for features.
Instead of pixelization of the pattern from the image data, a Euclidian distance computation can be done in a polar coordinate system which uses a point of reference and computes the angle from a reference line and the Euclidian distance from the point of reference to the optical center of each selected spot. However, such an alternative system is less amenable to automation and simplified processing of the image data when the subject has a large number of spots. However, this method may have merit in dealing with other species of subjects where the characteristics have traits that require Euclidian distance computations, such as determination of the antlers on ungulate species. Furthermore, there are numerous pattern matching techniques that can be used and these are well published and within the knowledge of one of ordinary skill in the art. Therefore, an extensive discourse regarding these techniques is unnecessary for the purposes of describing the present subject identification apparatus 100 and is omitted in the interest of simplicity of description.
This application is a continuation of U.S. patent application Ser. No. 15/077,788, filed Mar. 22, 2016, entitled “SYSTEM FOR UNIQUELY IDENTIFYING SUBJECTS FROM A TARGET POPULATION,” now U.S. Pat. No. 9,693,535, which is a continuation of U.S. patent application Ser. No. 14/480,526, filed Sep. 8, 2014, entitled “SYSTEM FOR UNIQUELY IDENTIFYING SUBJECTS FROM A TARGET POPULATION,” now U.S. Pat. No. 9,295,227, which is a continuation of U.S. patent application Ser. No. 13/850,139, filed Mar. 25, 2013, entitled “SYSTEM FOR UNIQUELY IDENTIFYING SUBJECTS FROM A TARGET POPULATION,” now U.S. Pat. No. 8,826,863, which is a continuation of U.S. patent application Ser. No. 13/352,934, filed Jan. 18, 2012, entitled “SYSTEM FOR UNIQUELY IDENTIFYING SUBJECTS FROM A TARGET POPULATION,” now U.S. Pat. No. 8,438,997, which is a continuation of U.S. patent application Ser. No. 12/716,083, filed Mar. 2, 2010, entitled “SYSTEM FOR UNIQUELY IDENTIFYING SUBJECTS FROM A TARGET POPULATION,” now U.S. Pat. No. 8,113,151, which is a continuation of U.S. patent application Ser. No. 10/498,188, filed Jun. 8, 2004, entitled “SYSTEM FOR UNIQUELY IDENTIFYING SUBJECTS FROM A TARGET POPULATION,” now U.S. Pat. No. 7,841,300, which claims priority from PCT application PCT/US2003/035581, filed Nov. 6, 2003, which claims priority from U.S. Provisional Application Ser. No. 60/424,917, filed Nov. 8, 2002. All of these applications are hereby incorporated by reference into this disclosure in their entirety.
| Number | Name | Date | Kind |
|---|---|---|---|
| 3040980 | Mann | Jun 1962 | A |
| 3096600 | Gwyther | Jul 1963 | A |
| 3564132 | Baker et al. | Feb 1971 | A |
| 4084148 | Koshikawa | Apr 1978 | A |
| 4095092 | Neff | Jun 1978 | A |
| 4226540 | Barten et al. | Oct 1980 | A |
| 4351438 | Morton | Sep 1982 | A |
| 4392236 | Sandstrom et al. | Jul 1983 | A |
| 4626992 | Greaves et al. | Dec 1986 | A |
| 4743742 | Espedalen | May 1988 | A |
| 4854455 | Faivre | Aug 1989 | A |
| 4934537 | DeBourke | Jun 1990 | A |
| 5063603 | Burt | Nov 1991 | A |
| 5222458 | Pippy | Jun 1993 | A |
| 5280436 | Kubota et al. | Jan 1994 | A |
| 5816196 | Webster et al. | Oct 1998 | A |
| 5823140 | Pittet et al. | Oct 1998 | A |
| 5930804 | Yu | Jul 1999 | A |
| 5991429 | Coffin et al. | Nov 1999 | A |
| 6006126 | Cosman | Dec 1999 | A |
| 6234109 | Andersson et al. | May 2001 | B1 |
| 6256737 | Bianco | Jul 2001 | B1 |
| 6310966 | Dulude | Oct 2001 | B1 |
| 6318289 | Pratt | Nov 2001 | B1 |
| 6418235 | Morimoto et al. | Jul 2002 | B1 |
| 6546368 | Weninger | Apr 2003 | B1 |
| 6687386 | Ito et al. | Feb 2004 | B1 |
| 6697509 | De La Torre-Bueno | Feb 2004 | B2 |
| 6700669 | Geng | Mar 2004 | B1 |
| 6880485 | Massey | Apr 2005 | B2 |
| 6922145 | Piesinger | Jul 2005 | B2 |
| 6974373 | Kriesel | Dec 2005 | B2 |
| 7190814 | Miichi et al. | Mar 2007 | B2 |
| 7841300 | Skvorc, II | Nov 2010 | B2 |
| 8113151 | Skvorc, II | Feb 2012 | B2 |
| 8438997 | Skvorc, II | May 2013 | B2 |
| 8826863 | Skvorc, II | Sep 2014 | B2 |
| 9295227 | Skvorc, II | Mar 2016 | B2 |
| 9693535 | Skvorc, II | Jul 2017 | B2 |
| 20020079249 | Lawandy et al. | Jun 2002 | A1 |
| 20030043042 | Moores, Jr. | Mar 2003 | A1 |
| 20050011470 | Skvorc, III | Jan 2005 | A1 |
| 20100215278 | Skvorc, III | Aug 2010 | A1 |
| 20120114186 | Skvorc, II | May 2012 | A1 |
| 20130216104 | Skvorc, II | Aug 2013 | A1 |
| 20140376783 | Skvorc, II | Dec 2014 | A1 |
| Number | Date | Country |
|---|---|---|
| 44 20 254 | Jan 1995 | DE |
| 1 018 297 | Jul 2000 | EP |
| 2634573 | Jan 1990 | FR |
| 2201772 | Sep 1988 | GB |
| 1-126546 | May 1989 | JP |
| 1-195364 | Aug 1989 | JP |
| 5-304851 | Nov 1993 | JP |
| Entry |
|---|
| Dong, C., et al., “Particle Retention Measurement Using Image Processing for a Spray Marking System”, Communications, Computers and Signal Processing, IEEE Pacific Rim Conference, May 19-21, 1993, pp. 790-793, XP-010141679. |
| Pike Angler's Club of Great Britain , “Recognition of Individual Pike”, Internet Address http://web.archive.org/web/20020707205535/http://freespace.virgin.net/vicky.mark/aboutpike/recognition.html, Mar. 24, 2004, p. 2, XP-002274805. |
| English translation of Japanese Patent No. 5-304851, “Method for Sorting Males and Females of Ayu Fishes and its Device”, Katsuhiko Kawaguchi, Nov. 1993. |
| English translation of Japanese Patent No. 1-126546, “Method for Identifying Rate of Movement of Multiple Fish”, by Takashi Iida et al., May 1989. |
| Jacob et al., Translation of French Patent No. FR 2634573, Jan. 26, 1990. |
| English translation of French patent 2,634,573, “Fish Counting Device” by M. Jacob et al., Jan. 1990. |
| Number | Date | Country | |
|---|---|---|---|
| 20180070558 A1 | Mar 2018 | US |
| Number | Date | Country | |
|---|---|---|---|
| 60424917 | Nov 2002 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 15077788 | Mar 2016 | US |
| Child | 15634594 | US | |
| Parent | 14480526 | Sep 2014 | US |
| Child | 15077788 | US | |
| Parent | 13850139 | Mar 2013 | US |
| Child | 14480526 | US | |
| Parent | 13352934 | Jan 2012 | US |
| Child | 13850139 | US | |
| Parent | 12716083 | Mar 2010 | US |
| Child | 13352934 | US | |
| Parent | 10498188 | US | |
| Child | 12716083 | US |
