Statistical approach to identifying and tracking targets within captured image data
원문보기
IPC분류정보
국가/구분
United States(US) Patent
등록
국제특허분류(IPC7판)
G06K-009/00
G06K-009/32
G06T-001/20
G06T-007/40
G06T-007/20
G06K-009/46
H04N-005/33
출원번호
US-0800223
(2013-03-13)
등록번호
US-9373051
(2016-06-21)
발명자
/ 주소
Viviani, Gary Lee
출원인 / 주소
Insitu, Inc.
대리인 / 주소
Perkins Coie LLP
인용정보
피인용 횟수 :
2인용 특허 :
14
초록▼
A facility implementing systems and/or methods for creating statistically significant signatures for targets of interest and using those signatures to identify and locate targets of interest within image data, such as an array of pixels, captured still images, video data, etc., is described. Embodim
A facility implementing systems and/or methods for creating statistically significant signatures for targets of interest and using those signatures to identify and locate targets of interest within image data, such as an array of pixels, captured still images, video data, etc., is described. Embodiments of the facility generate statistically significant signatures based at least in part on series approximations (e.g., Fourier Series, Gram-Charlier Series) of image data. The disclosed techniques allow for a high degree of confidence in identifying and tracking targets of interest within visual data and are highly tolerant of translation and rotation in identifying objects using the statistically significant signatures.
대표청구항▼
1. A method, performed by a computing device having a memory and processor, for identifying targets of interest within captured image data, the method comprising: receiving, from a user, a selection of a first portion of a first frame of image data, the first portion of the first frame of image data
1. A method, performed by a computing device having a memory and processor, for identifying targets of interest within captured image data, the method comprising: receiving, from a user, a selection of a first portion of a first frame of image data, the first portion of the first frame of image data including a target of interest;determining a characteristic value for each of a plurality of pixels within the selected first portion of the first frame of image data;generating a first series approximation for the characteristic values of the plurality of pixels within the selected first portion of the first frame of image data based on pixels only within the selected first portion of the first frame of image data;generating a target signature for the target of interest based at least in part on the generated first series approximation; andfor each of a plurality of second portions of second frames of image data, determining a characteristic value for each of a plurality of pixels within the second portion of the second frame of image data,generating a second series approximation for the characteristic values of the plurality of pixels within the second portion of the second frame of image data,generating a comparison signature for the second portion of the second frame of image data based at least in part on the generated second series approximation, anddetermining a difference between the target signature and the comparison signature generated for the second portion of the second frame of image data. 2. The method of claim 1, further comprising: identifying the second portion of the second frame of image data associated with the comparison signature having the smallest determined difference between the target signature; andindicating that the identified second portion of the second frame of image data includes the target of interest. 3. The method of claim 1, further comprising: determining a first mean value for at least a portion of the determined characteristic values of the plurality of pixels within the selected first portion of the first frame of image data; anddetermining a first variance value for at least a portion of the determined characteristic values for the plurality of pixels within the selected first portion of the first frame of image data,wherein generating the target signature is further based at least in part on the determined first mean value and the determined first variance value. 4. The method of claim 3, further comprising: for each of the plurality of second portions of second frames of image data, determining a second mean value for at least a portion of the determined characteristic values of the plurality of pixels within the second portion of the second frame of image data, anddetermining a second variance value for at least a portion of the determined characteristic values of the plurality of pixels within the second portion of the second frame of image data,wherein generating the comparison signature is further based at least in part on the determined second mean value and the determined second variance value. 5. The method of claim 1, further comprising: normalizing at least a portion of the determined characteristic values of the plurality of pixels within the selected first portion of the first frame of image data, wherein generating the first series approximation for the characteristic values of the plurality of pixels within the selected first portion of the first frame of image data comprises generating the first series approximation based at least in part on the normalized portion of the determined characteristic values of the plurality of pixels within the selected first portion of the first frame of image data. 6. The method of claim 5, further comprising: for each of the plurality of second portions of second frames of image data, normalizing at least a portion of the determined characteristic values of the plurality of pixels within the second portion of the second frame of image data, wherein generating the second series approximation for the characteristic values of the plurality of pixels within the second portion of the second frame of image data comprises generating the second series approximation based at least in part on the normalized portion of the determined characteristic values of the plurality of pixels within the second portion of the second frame of image data. 7. A method, performed by a computing device having a memory and processor, for identifying targets of interest within captured image data, the method comprising: receiving, from a user, a selection of a first portion of a first frame of image data, the first portion of the first frame of image data including a target of interest;determining a characteristic value for each of a plurality of pixels within the selected first portion of the first frame of image data;generating a first series approximation for the characteristic values of the plurality of pixels within the selected first portion of the first frame of image data;generating a target signature based at least in part on the generated first series approximation; andfor each of a plurality of second portions of second frames of image data, determining a characteristic value for each of a plurality of pixels within the second portion of the second frame of image data,generating a second series approximation for the characteristic values of the plurality of pixels within the second portion of the second frame of image data,generating a comparison signature for the second portion of the second frame of image data based at least in part on the generated second series approximation, anddetermining a difference between the target signature and the comparison signature generated for the second portion of the second frame of image data,normalizing at least a portion of the determined characteristic values of the plurality of pixels within the selected first portion of the first frame of image data, wherein generating the first series approximation for the characteristic values of the plurality of pixels within the selected first portion of the first frame of image data comprises generating the first series approximation based at least in part on the normalized portion of the determined characteristic values of the plurality of pixels within the selected first portion of the first frame of image data,wherein the target signature, {right arrow over (S)}, is represented as {right arrow over (S)}=[(L1{f(coeff({right arrow over (X)}norm))}+L2{f(coeff({right arrow over (X)}norm))}),E[{right arrow over (X)}],E[{right arrow over (X)}2]−(E[{right arrow over (X)}])2], wherein {right arrow over (X)}norm represents the normalized characteristic values of the plurality of pixels within the selected first portion of the first frame of image data, wherein coeff(f({right arrow over (X)})) represents coefficients of a series approximation, and wherein {right arrow over (X)} represents the characteristic values of the plurality of pixels within the selected first portion of the first frame of image data. 8. The method of claim 1, further comprising: determining a first vector norm based at least in part on the generated first series approximation; anddetermining a second vector norm based at least in part on the generated first series approximation,wherein generating the target signature is further based at least in part on the determined first vector norm and the determined second vector norm. 9. The method of claim 8, further comprising: for each of the plurality of second portions of second frames of image data, determining a third vector norm based at least in part on the generated second series approximation,determining a fourth vector norm based at least in part on the generated second series approximation,wherein generating the comparison signature is further based at least in part on the determined third vector norm and the determined fourth vector norm. 10. The method of claim 1, further comprising: highlighting the identified second portion of the second frame of image data. 11. The method of claim 1, wherein the characteristic values are intensity values. 12. A method, performed by a computing device having a memory and processor, for identifying targets of interest within captured image data, the method comprising: receiving, from a user, a selection of a first portion of a first frame of image data, the first portion of the first frame of image data including a target of interest;determining a characteristic value for each of a plurality of pixels within the selected first portion of the first frame of image data;generating a first series approximation for the characteristic values of the plurality of pixels within the selected first portion of the first frame of image data;generating a target signature based at least in part on the generated first series approximation; andfor each of a plurality of second portions of second frames of image data, determining a characteristic value for each of a plurality of pixels within the second portion of the second frame of image data,generating a second series approximation for the characteristic values of the plurality of pixels within the second portion of the second frame of image data,generating a comparison signature for the second portion of the second frame of image data based at least in part on the generated second series approximation, anddetermining a difference between the target signature and the comparison signature generated for the second portion of the second frame of image data,wherein the first series approximation is generated based at least in part on f(X→n)=∑s1=0∞…∑sn=0∞(E[∏i=1nHsi(x→i)]∏p=1n[Hsp(x→p)•G(x→p)]sp!), wherein {right arrow over (X)}n=({right arrow over (x)}1, {right arrow over (x)}2, . . . , {right arrow over (x)}n) and represents normalized characteristic values, wherein G(t)=12πⅇ-t22, and wherein Hi represents the ith Hermite polynomial. 13. A method, performed by a computing device having a memory and processor, for identifying targets of interest within captured image data, the method comprising: receiving, from a user, a selection of a first portion of a first frame of image data, the first portion of the first frame of image data including a target of interest;determining a characteristic value for each of a plurality of pixels within the selected first portion of the first frame of image data;generating a first series approximation for the characteristic values of the plurality of pixels within the selected first portion of the first frame of image data;generating a target signature based at least in part on the generated first series approximation; andfor each of a plurality of second portions of second frames of image data, determining a characteristic value for each of a plurality of pixels within the second portion of the second frame of image data,generating a second series approximation for the characteristic values of the plurality of pixels within the second portion of the second frame of image data,generating a comparison signature for the second portion of the second frame of image data based at least in part on the generated second series approximation, anddetermining a difference between the target signature and the comparison signature generated for the second portion of the second frame of image data,wherein the first series approximation is generated based at least in part on f(x→1,x→2)=∑s1=0∞∑s2=0∞(E[Hs1(x→1)•Hs2(x→2)]•Hs1(x→1)•G(x→1)•Hs2(x→2)•G(x→2)s1!s2!) wherein {right arrow over (x)}1 represents normalized characteristic values received from a first sensor, wherein {right arrow over (x)}2 represents normalized characteristic values received from a second sensor, wherein G(t)=12πⅇ-t22, and wherein Hj represents the jth Hermite polynomial. 14. The method of claim 1, further comprising: generating a plurality of target signatures for the target of interest; andin response to determining that the target of interest is to be detected, identifying the plurality of target signatures,prompting a user to select from among the plurality of target signatures, andcomparing the selected target signature to each of a plurality of generated comparison signatures. 15. A computer-readable medium, that is not a transitory propagating signal, storing instructions that, when executed by a computing device having a processor, cause the computing device to perform operations comprising: receiving, from a user, a selection of a first portion of a first frame of image data, the first portion of the first frame of image data including a target of interest;determining a first probability density function based at least in part on characteristic values of a plurality of pixels within the selected first portion of the first frame of image data;generating a first signature for the target of interest based at least in part on the first probability density function using a truncated multivariate Gram-Charlier series; andfor a second frame of image data, for each of a plurality of second portions of the second frame of image data, determining a second probability density function based at least in part on characteristic values of a plurality of pixels within the second portion of the second frame of image data,generating a second signature based at least in part on the second probability density function, anddetermining a difference between the second signature and the first signature. 16. The computer-readable medium of claim 15, the operations further comprising: identifying the second portion of the second frame of image data having the smallest difference between its second signature and the first signature; andindicating that the identified second portion of the second frame of image data includes the target of interest. 17. The computer-readable medium of claim 16, the operations further comprising: determining whether the smallest difference is greater than a signature threshold; andin response to determining that the smallest difference is greater than a signature threshold, updating the first signature. 18. The computer-readable medium of claim 15, the operations further comprising: generating a series approximation to the first probability density function, wherein the first signature for the target of interest is based at least in part on the coefficients of the generated series approximation. 19. The computer-readable medium of claim 15, wherein the first signature is based at least in part on data collected from at least two of a visible light camera, a thermographic camera, or an ultraviolet camera. 20. The computer-readable medium of claim 15, wherein the characteristic values are intensity values. 21. A method, performed by a computing device having a memory and processor, for identifying targets of interest within captured image data, the method comprising: determining a characteristic value for each of a plurality of pixels within a first patch of pixels only within a first image;generating a first series approximation based on values for the first patch of pixels only within the first image;generating a target signature based at least in part on the generated first series approximation;for each of a plurality of patches of pixels within a second image, generating a second series approximation based at least in part on values for pixels within the patch,generating a comparison signature based at least in part on the generated second series approximation, anddetermining a distance between the comparison signature and the target signature;identifying the patch of pixels associated with the comparison signature having the smallest determined distance between the target signature as including the target of interest. 22. A computer-readable medium, that is not a transitory propagating signal, storing instructions that, when executed by a computing device having a processor, cause the computing device to perform operations comprising: determining a characteristic value for each of a plurality of pixels within a first patch of pixels only within a first image;generating a first series approximation based on values for the first patch of pixels only within the first image;generating a target signature based at least in part on the generated first series approximation;for each of a plurality of patches of pixels within a second image, generating a second series approximation based at least in part on values for pixels within the patch,generating a comparison signature based at least in part on the generated second series approximation, anddetermining a distance between the comparison signature and the target signature; andidentifying the patch of pixels associated with the comparison signature having the smallest determined distance between the target signature as including the target of interest.
연구과제 타임라인
LOADING...
LOADING...
LOADING...
LOADING...
LOADING...
이 특허에 인용된 특허 (14)
Neumann, Jonathan, Agile interrogation hyperspectral imager and method of using same.
Cyril C. Marrion, Jr. ; Ivan A. Bachelder ; Edward A. Collins, Jr. ; Masayoki Kawata JP; Sateesh G. Nadabar, Machine vision system for identifying and assessing features of an article.
Ulich Bobby L. (Tucson AZ) Montgomery John W. (Tucson AZ), Process for automatically detecting and locating a target from a plurality of two dimensional images.
Prouty, Jeff; Ullyott, Logan; Canary, Grant, Forestry information management systems and methods streamlined by automatic biometric data prioritization.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.