It is an object of the present invention to provide a monitoring system that captures and inputs video images of a crowd with a surveillance camera for detecting normal and/or various abnormal states. The monitoring system (MS) comprises: a video image input means (1) that captures and inputs the vi
It is an object of the present invention to provide a monitoring system that captures and inputs video images of a crowd with a surveillance camera for detecting normal and/or various abnormal states. The monitoring system (MS) comprises: a video image input means (1) that captures and inputs the video images of the crowd; an optical-flow calculating means (2) that calculates optical flows from the video images captured at different times; a judgment-block setting means (3) that sets judgment blocks on a surveillance area of each video image, each judgment block serving as a unit to be judged about an abnormality; an optical-flow attribute counting means (4) that counts an attribute of optical flows generated in each judgment block; a normality-evaluating-index calculating means (5) that calculates, from the counted attribute of the optical flows, an evaluation value of a normality-evaluating index for evaluating a normality of the judgment block; an abnormal-state judging means (6) that judges, from the evaluation value of the normality-evaluating index, the abnormal state of the judgment block; and an output means (7) that outputs a result of the judgment made by the abnormal-state judging means onto the video image.
대표청구항▼
1. A monitoring system that captures and inputs video images of a crowd including a plurality of moving objects for detecting an abnormal state of the crowd from movement data obtained from the video images, the monitoring system comprising: a video image input device configured to capture and input
1. A monitoring system that captures and inputs video images of a crowd including a plurality of moving objects for detecting an abnormal state of the crowd from movement data obtained from the video images, the monitoring system comprising: a video image input device configured to capture and input the video images of the crowd;an optical-flow calculating device configured to calculate optical flows from the video images captured at different times;a judgment-block setting device configured to set judgment blocks on a surveillance area of each video image, each judgment block serving as a unit to be judged about an abnormality;an optical-flow attribute counting device configured to count attribute data of optical flows generated in each judgment block;a normality-evaluating-index calculating device configured to calculate, from the counted attribute data of the optical flows, an evaluation value of a normality-evaluating index for evaluating a normality of the judgment block;an abnormal-state judging device configured to judge, from the evaluation value of the normality-evaluating index, the abnormal state of the judgment block; andan output device configured to output a result of the judgment made by the abnormal-state judging device onto the video image,wherein the optical-flow attribute counting device quantifies, as the attribute data, directions of the optical flows generated in the judgment block, divides the quantized directions into groups and counts quantified directions of each group to create a direction distribution of the optical flows,wherein the normality-evaluating-index calculating device determines top two directions of optical flows from the direction distribution, calculates evaluation values of a respective three normality-evaluating indices which are a direction concentration, an angle, and a reverse movement degree, the direction concentration being a degree of concentration of the number of optical flows having the top two directions, the angle being an angle formed between the top two directions, and the reverse movement degree being a ratio of the number of optical flows having one of the top two directions to the number of optical flows having the other direction, andwherein the abnormal-state judging device judges whether the judgment block has a reverse movement against a main stream as the abnormal state, based on the evaluation values of the three normality-evaluating indices. 2. A monitoring system that captures and inputs video images of a crowd including a plurality of moving objects for detecting an abnormal state of the crowd from movement data obtained from the video images, the monitoring system comprising: a video image input device configured to capture and input the video images of the crowd;an optical-flow calculating device configured to calculate optical flows from the video images captured at different times;a judgment-block setting device configured to set judgment blocks on a surveillance area of each video image, each judgment block serving as a unit to be judged about an abnormality;an optical-flow attribute counting device configured to count attribute data of optical flows generated in each judgment block;a normality-evaluating-index calculating device configured to calculate, from the counted attribute data of the optical flows, an evaluation value of a normality-evaluating index for evaluating a normality of the judgment block;an abnormal-state judging device configured to judge, from the evaluation value of the normality-evaluating index, the abnormal state of the judgment block; andan output device configured to output a result of the judgment made by the abnormal-state judging device onto the video image,wherein the optical-flow attribute counting device quantifies, as the attribute data, directions of the optical flows generated in the judgment block, divides the quantized directions into groups and counts quantified directions of each group to create a direction distribution of the optical flows, andwherein the output device assigned colors respectively to the groups of the direction distribution, color-codes the optical flows according to which group each optical flow belongs to, and overlays and displays the color-coded optical flows on the video image. 3. The monitoring system of claim 1, wherein the output device overlays and displays the direction distribution on the judgment block of the video image. 4. A monitoring system that captures and inputs video images of a crowd including a plurality of moving objects for detecting an abnormal state of the crowd from movement data obtained from the video images, the monitoring system comprising: a video image input device configured to capture and input the video images of the crowd;an optical-flow calculating device configured to calculate optical flows from the video images captured at different times;a judgment-block setting device configured to set judgment blocks on a surveillance area of each video image, each judgment block serving as a unit to be judged about an abnormality;an optical-flow attribute counting device configured to count attribute data of optical flows generated in each judgment block;a normality-evaluating-index calculating device configured to calculate, from the counted attribute data of the optical flows, an evaluation value of a normality-evaluating index for evaluating a normality of the judgment block;an abnormal-state judging device configured to judge, from the evaluation value of the normality-evaluating index, the abnormal state of the judgment block; andan output device configured to output a result of the judgment made by the abnormal-state judging device onto the video image,wherein the optical-flow attribute counting device counts, as the attribute data, lengths of the optical flows generated in each judgment block to create a velocity distribution of the optical flows, andwherein the normality-evaluating-index calculating device calculates, as evaluation values of normality-evaluating indices, an average μ of the velocity distribution, a standard deviation σ of the velocity distribution, and either one of a range between a lower limit μ−2σ of it normal range of the velocity distribution and an upper limit μ+2σ of the normal ranges and a range between a lower limit μ−3σ of the normal range and upper limit μ+3σ of the normal range, andwherein the abnormal-state judging device judges, when there is an optical flow outside the either one of the ranges, that the block is in an abnormal state which is a velocity abnormality in which there is a movement different in velocity from others. 5. A monitoring system that captures and inputs video images of a crowd including a plurality of moving objects for detecting an abnormal state of the crowd from movement data obtained from the video images, the monitoring system comprising: a video image input device configured to capture and input the video images of the crowd;an optical-flow calculating device configured to calculate optical flows from the video image captured at different times;a judgment-block setting device configured to set judgment blocks on a surveillance area of each video image, each judgment block serving as a unit to be judged about an abnormality;an optical-flow attribute counting device configured to count attribute data of optical flows generated in each judgement block;a normality-evaluating-index calculating device configured to calculate, from the counted attribute data of the optical flows, an evaluation value of a normality-evaluating index for evaluating a normality of the judgment block;an abnormal-state judging device configured to judge, from the evaluation value of the normality-evaluating index, the abnormal state of the judgment block; andan output device configured to output a result of the judgment made by the abnormal-state judging device onto the video image,wherein the optical-flow attribute counting device counts, as the attribute data, lengths of the optical flows generated in each judgment block to create a velocity distribution of the optical flows, andwherein the optical-flow attribute counting device converts the respective lengths of the optical flows generated in the judgment block into distances based on world coordinates, before creating the velocity distribution of the optical flows. 6. The monitoring system of claim 4, wherein the output device depicts the optical flow outside the either one of the ranges in a color different from one in which the rest of the optical flows are depicted and overlays the optical flow depicted in the different color on the video image before outputting. 7. A monitoring system that captures and inputs video images of a crowd including a plurality of moving objects for detecting an abnormal state of the crowd from movement data obtained from the video images, the monitoring system comprising: a video image input device configured to capture and input the video images of the crowd;an optical-flow calculating device configured to calculate optical flows from the video images captured at different times;a judgment-block setting device configured to set judgment blocks on a surveillance area of each video image, each judgment block serving as a unit to be judged about an abnormality;an optical-flow attribute counting device configured to count attribute data of optical flows generated in each judgment block;a normality-evaluating-index calculating device configured to calculate, from the counted attribute data of the optical flows, an evaluation value of a normality-evaluating index for evaluating a normality of the judgment block;an abnormal-state judging device configured to judge, from the evaluation value of the normality-evaluating index, the abnormal state of the judgment block; andan output device configured to output a result of the judgment made by the abnormal-state judging device onto the video image,wherein the optical-flow attribute counting device count the number of the optical flows generated in the judgment block, andwherein the normality-evaluating-index calculating devices judges that the block is in the abnormal state when a difference between the number of optical flows at the present time and the number of optical flows at a time earlier by a given time period exceeds a threshold value. 8. A monitoring system that captures and inputs video images of a crowd including a plurality of moving objects for detecting an abnormal state of the crowd from movement data obtained from the video images, the monitoring system comprising: a video image input device configured to capture and input the video images of the crowd;an optical-flow calculating device configured to calculate optical flows from the video images captured at different times;a judgment-block setting device configured to set judgment blocks on a surveillance area of each video image, each judgment block serving as a unit to be judged about an abnormality;an optical-flow attribute counting device configured to count attribute data of optical flows generated in each judgment block;a normality-evaluating-index calculating device configured to calculate, from the counted attribute data of the optical flows, an evaluation value of a normality-evaluating index for evaluating a normality of the judgment block;an abnormal-state judging device configured to judge, from the evaluation value of the normality-evaluating index, the abnormal state of the judgment block; andan output device configured to output a result of the judgment made by the abnormal-state judging device onto the video image,wherein the optical-flow attribute counting device counts the number of the optical flows generated in the judgment block, andwherein the abnormal-state judging device judges that a block is in the abnormal state when a difference between the number of optical flows generated in that judgment block and the number of optical flows generated in a block existing in a surrounding area exceeds a threshold value. 9. The monitoring system of claim 1, wherein when the abnormal-state judging device judges the abnormal state of the judgment block, the output device highlights the judgment block judged as being in the abnormal state with a rectangle depicted in a color assigned according to type of the abnormal state. 10. The monitoring system of claim 1, wherein the output device depicts the numeral of the evaluation value calculated by normality-evaluating-index calculating device on the block of the video image, andwhen it is judged that the evaluation value calculated by the abnormal-state judging device exceeds a threshold value, depicts the numeral in a color different from one assigned to a case where it is judged otherwise. 11. A monitoring system that captures and inputs video images of a crowd including a plurality of moving objects for detecting an abnormal state of the crowd from movement data obtained from the video images, the monitoring system comprising: a video image input device configured to capture and input the video images of the crowd;an optical-flow calculating device configured to calculate optical flows from the video images captured at different times;a judgment-block setting device configured to set judgment blocks on a surveillance area of each video image, each judgment block serving as a unit to be judged about an abnormality;an optical-flow attribute counting device configured to count attribute data of optical flows generated in each judgment block;a normality-evaluating-index calculating device configured to calculate, from the counted attribute data of the optical flows, an evaluation value of a normality-evaluating index for evaluating a normality of the judgment block;an abnormal-state judging device configured to judge, from the evaluation value of the normality-evaluating index, the abnormal state of the judgment block; andan output device configured to output a result of the judgment made by the abnormal-state judging device onto the video image,wherein the optical-flow attribute counting device quantifies, as the attribute data, directions of the optical flows generated in the judgment block, divides the quantized directions into groups and counts quantified directions of each group to create a direction distribution of the optical flows, andwherein the normality-evaluating-index calculating device calculates an entropy from the direction distribution, andwherein the abnormal-state judging device judges an area having a greater entropy as an area having a disturbance in a stream. 12. A monitoring method for monitoring movements of a crowd with a monitoring system configured to capture and input video images of the crowd including a plurality of moving objects for detecting an abnormal state of the crowd from movement data obtained from the video images, the monitoring method comprising the steps of: capturing and inputting the video images of the crowd;calculating optical flows from the video images captured at different times;setting judgment blocks on a surveillance area of each video image, each judgment block serving as a unit to be judged about an abnormality,counting attribute data of optical flows generated in each judgment block;calculating, from the counted attribute data the optical flows, an evaluation value of a normality-evaluating index for evaluating a normality of the judgment block;judging, from the evaluation value of the normality-evaluating index, the abnormal state of the judgment block;outputting a result of the judgment on the abnormal state onto the video image;quantifying, as the attribute data, directions of the optical flows generated in the judgment block, dividing the quantified directions into groups and counting quantified directions of each group to create a direction distribution of the optical flows;determining top two directions of optical flows from the direction distribution, provided according to the direction, calculating evaluation values of respective three normality-evaluating indices which are a direction concentration, an angle, and a reverse movement degree, the direction concentration being a degree of concentration of the number of optical flows having the top two directions, the angle being an angle formed between the top two directions, and the reverse movement degree being a ratio of the number of optical flows having one of the top two directions to the number of optical flows having the other direction;judging a reverse movement against a main stream as the abnormal state, based on the evaluation values of the three normality-evaluating indices;depicting the numerals of the evaluation values of the three normality-evaluating indices on the block of the video image; andwhen the result of the judgment is the abnormal state of the judgment block, highlighting, with a rectangle, the judgment block judged as being in the abnormal state.
연구과제 타임라인
LOADING...
LOADING...
LOADING...
LOADING...
LOADING...
이 특허에 인용된 특허 (2)
Dennison, John Charles; Davis, Clayton Houston; Frazier, Brian Edward, Determining angular rate for line-of-sight to a moving object, with a body-fixed imaging sensor.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.