JP2005176077A - Camera monitoring system and its monitoring control method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a camera monitoring system for monitoring traffic situation and applying image processing whose fault detection result having a higher correct answer rate to match with the judgment of a watchman. <P>SOLUTION: The camera monitoring system stores an image signal processing algorithm in an image processing portion 1 for processing video image information by a camera 4 to judge fault events, data such as operating parameters, and fault event judgment results into a data server 25, along with a time information (time tag). Consequently, three kinds of estimation results, a watchman's correct detection, wrong detection as a fault event, and failure in detection, can be compared and collated. The stored data and the evaluation results are collated and analyzed. As a result, a control unit 24 controls and adjusts the image processing portion 1 so as to be able to detect fault events correctly, in order to improve the correct answer rate. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a camera monitoring system that performs video monitoring and a monitoring control method thereof.

  There is a camera monitoring system that generates traffic information by monitoring the traffic situation of a road using video information of a monitoring camera and manages traffic on a highway.

  In this camera monitoring system, a surveillance person reports traffic information by determining the driving state of a road by a camera image, a traffic information system other than the camera image, and various reports.

  On highways and the like, a plurality of surveillance cameras are installed, but it is difficult for the surveillance staff to constantly monitor all the surveillance cameras and immediately detect the occurrence of an abnormality. Therefore, in these camera monitoring systems, as an aid to video monitoring, the video information of the camera is subjected to image processing, an alarm for abnormal events such as traffic jams is issued, and the monitoring staff is notified.

  Image processing methods that detect abnormal events from video information from surveillance cameras include various target detection algorithms and abnormal event detection algorithms. For example, the movement of a photographed vehicle is observed and the subject stays for a long time. There are those that detect appearance and disappearance (for example, see Patent Document 1).

  However, each algorithm is suitable for accurately detecting and classifying abnormal events in various traffic situations according to the traffic conditions, and furthermore, the operation parameters that determine the abnormal event detection correctly are also set uniformly. Not. Conventionally, therefore, there is a problem that the determination of an abnormal event has to depend on the determination of a supervisor.

Judgment conditions and operation parameters are set by acquiring video information and image processing operation data in the actual environment, and comparing the operation and abnormal event detection results with the judgment of the monitor to design and adjust the image processing method. Although it was necessary to do so, the camera monitoring system operation results and the monitoring personnel's results were conventionally reported only in sporadic reports in the diary etc. It was not possible to sufficiently verify with the judgment.
JP 2002-262282 A (12th page, FIG. 2)

  There are various algorithms for image processing methods that automatically detect abnormal events from video information of traffic monitoring cameras in camera monitoring systems, but each algorithm also accurately extracts abnormal events corresponding to various traffic conditions. It is suitable for classification, and the setting of operation parameters for correctly determining the abnormal event detection is not uniform. As a result, conventionally, there has been a problem that the determination of an abnormal event has to depend on the determination of a supervisor. In particular, the setting of operation parameters can be performed by measuring video information and image processing operations in a real environment, and comparing the operations and abnormality detection results with the judgment of the supervisor to design and adjust the image processing method. In spite of necessity, conventionally, there has been a problem that the verification data between the operation of the camera monitoring system and the judgment of the monitoring person is not accumulated, which is insufficient for verifying the algorithm and confirming the setting of the operation parameter.

  The present invention has been made to solve the above-described problem, and realizes a camera monitoring system with a high correct answer rate in which an abnormality detection result by image processing of a camera monitoring system that monitors traffic conditions matches a judgment of a monitor. Objective.

  In order to achieve the above object, a camera monitoring system according to the present invention is a camera monitoring system that monitors traffic conditions based on video information of a camera, and outputs a time information for referring to the operation of the camera monitoring system. Means for extracting vehicle image information from the video information of the camera, detection means for detecting abnormal event information relating to traffic conditions from the extracted image information, and an abnormality from the detected abnormal event information. A determination unit that discriminates and determines an event; a notification unit that notifies a determination result of the determined abnormal event to a monitor of the camera monitoring system; and a correct correct result of the determination of the abnormal event notified by the monitor It was evaluated as a detection, or it was evaluated as a false detection where an abnormal event occurred when no abnormal event occurred, or it was different even though an abnormal event occurred. An evaluation input means for inputting any one of the evaluation results of whether the event has not been detected as undetected, the video information, the image information, the abnormal event information, and the determination means include the valve It is characterized by comprising a determination condition to be determined separately, a determination result of the abnormal event, the evaluation result, and the time information, and a control means for monitoring and controlling the respective means.

  The monitoring control method of the camera monitoring system of the present invention includes a camera, a clock means, an extraction means for extracting vehicle image information from video information of the camera, and detecting abnormal event information from the extracted image information. Detecting means for determining, determining means for discriminating and determining an abnormal event from the detected abnormal event information, notification means for notifying a monitoring result of the determined abnormal event to a monitor of the camera monitoring system, and the monitor Evaluates the judgment result of the notified abnormal event as a correct positive detection, or evaluates it as a false detection that an abnormal event has occurred even though no abnormal event has occurred, or an abnormal event has occurred. It comprises evaluation input means for inputting any one of the evaluation results of whether the abnormal event has not been detected and evaluated, storage means, and control means for monitoring and controlling each of the means. A monitoring control method for a camera monitoring system for monitoring a traffic situation, wherein the extraction unit extracts the image information from video information of the camera according to an extraction operation parameter stored in the storage unit, and the detection unit includes: The abnormal event information of the traffic situation is detected from the extracted image information by the detection operation parameter stored and set in the storage means, and is transmitted to the determination means. When the determination means receives the abnormal event information, The determination unit discriminates the abnormal event by collating with the determination condition stored in the storage unit, and the storage unit extracts the extraction operation parameter of the extraction unit, the detection operation parameter of the detection unit, and each of the parameters. A plurality of determination condition candidates for the abnormal event consisting of a combination and at least one determination condition among the stored candidates A determination result determined by the determination means; a determination result of the determination result notified by the notification means to a monitor of the surveillance camera system; an evaluation result input from the evaluation input means; and a reception from the clock means The time information is stored, and the control means selects one of the candidates and performs the control for notifying the determination result based on the determination condition to the notification means, and the monitor who receives the notification performs the control. When the evaluation result input from the evaluation input means is read and the evaluation result is not the positive detection, the data stored in the storage means is read with reference to the time information, the image information, and the abnormal event information The extraction operation parameter and the detection operation parameter are searched, and the extraction operation parameter and the detection operation parameter whose evaluation result is the positive detection are determined. When a new determination condition for the combination is found, the notification for the next and subsequent times is controlled to be corrected and changed so as to notify the result determined by discriminating the abnormal event according to the new determination condition.

  The camera monitoring system according to the present invention includes a video signal processing algorithm for processing abnormal image information by processing video information of the camera, a determination condition including a combination of operation parameters, and data of the determination result, and video information of the camera. Is stored in the data server together with the time information so that it can be compared with the three evaluation result data of whether the monitor has correctly detected for the determination, erroneously determined as an abnormal event, or whether the abnormal event has not been detected, By comparing the stored data with the camera video information and the evaluation result data and analyzing the determination contents, the image signal processing algorithm and operating parameters are adjusted to increase the correct answer rate so that abnormal events can be detected correctly. be able to.

  Embodiments of the present invention will be described below with reference to the drawings.

  FIG. 1 is a block diagram showing components of an example of a camera monitoring system according to an embodiment of the present invention.

  The camera monitoring system in FIG. 1 is a configuration and arrangement example of a camera monitoring system that performs road monitoring using camera images, and includes a camera 4, an image processing unit 1, a network 3, and a monitoring center 2. The camera 4 and the image processing unit 1 are a set of (# 1) to (#n), and the image processing unit 1 is installed in a site such as an electric room of an expressway facility or an office to check traffic conditions on the road. Take a picture.

  The image processing unit 1 measures the speed of the vehicle, which is abnormal event information for determining abnormal events such as traffic conditions and traffic jams, a feature extracting unit 11 that is a means for extracting vehicle images from video information captured by the camera 4 The event detection unit 12 to be calculated, the determination unit 13 that discriminates and determines an abnormal event based on the calculated speed, and the control / communication unit 14 that controls each component of the image processing unit.

  The determination unit 13 includes a table T that lists candidates for determination conditions for abnormal events. The table T is created in the internal memory of the image processing unit 1, and stores judgment condition candidates consisting of combinations of extraction operation parameters of the feature extraction unit 11 and detection operation parameters of the event detection unit 12 in a table.

  The extraction operation parameters are an image extraction algorithm, an image binarization threshold level, a tracking point of the vehicle image, and the like, and the detection operation parameters are a vehicle speed measurement section, a measurement frequency, and the like.

  Processing data and operation parameters in the feature extraction unit 11, event detection unit 12, and determination unit 13 are connected to the control / communication unit 14 via an internal bus or the like, and are further monitored from the control / communication unit 14 via the network 3. 2 is transmitted to the second data server 25.

  The monitoring center 2 selects the monitoring monitor 21 that displays the video information of the cameras 4 (# 1 to #n) received from the network 3, the cameras 4 (# 1 to #n) to display, and the image processing unit 1 The control console 22, the operation of each component of the monitoring center, the control console 22 for performing the control input for reading the data stored in the data server 25, the abnormal event attached to the control console 22 and notified by the image processing unit 1 The evaluation switch 23, the control unit 24, and the data server 25 input the evaluation of “correct detection”, “false detection”, or “not detected” in response to the notification (alarm). The data server 25 stores the video information transmitted from the image processing unit 1, the setting data of the feature extraction unit 11, the setting and operation data of the event detection unit 12, the determination result of the abnormal event of the determination unit 13, and the like. The control unit 24 performs control and data monitoring of each component of the data server, the image processing unit 1 and the like by input setting from the console 22 or a built-in program.

  The data server 25 records the video information of the camera 4 received via the network 3 and various data relating to the processing of the abnormal event acquired corresponding to each determination condition candidate stored in the table T of the determination unit 13. Is stored in the log table TL.

  In addition, when the image processing unit 1 (#c) that performs image processing on the video signal from the camera 4 is installed in the monitoring center 2, the data server 25 is installed at the site where the image processing unit 1 is installed. Sometimes it is done.

  From the video information photographed by the camera 4, the feature extraction unit 11 extracts image information for specifying the position of the vehicle, and the feature extraction unit 11 calculates the traveling speed of the vehicle from the image information to determine whether it is “normal” traveling. , Vehicle traveling speed data for determining an abnormal event such as “traffic jam” is generated. The determination unit 13 interprets the traveling speed of the vehicle from the traveling speed data received from the event detection unit 12 and discriminates whether the abnormal event of the traffic situation is an abnormal event such as “normal” or “congestion”.

  The control / communication unit 14 sets, in the table T, determination conditions for setting each determination condition for the determination unit 13 to perform discrimination determination, and the operation parameters of the feature extraction unit 11 and the feature extraction unit 11. The control / communication unit 14 transmits / receives information and data set in the table T and data of the determination result of the determination unit 13 to / from the data server 25 of the monitoring center 2 via the network 3.

  In FIG. 1, the video information captured by the camera 4 (# 1) is transmitted in two systems. One is transmitted to the feature extraction unit 11 of the image processing unit, and the other is transmitted from the control / communication unit 14 to the monitoring center 2 via the network 3 and displayed on the monitoring monitor 21 of the monitoring center 2, and at the same time the data server 25 is also recorded.

  The feature extraction unit 11 is set with an algorithm for detecting a vehicle image by the control / communication unit 14 and extracts vehicle image information for measuring the traveling speed of the vehicle from the video information. There are many algorithms such as a “background difference” method, an “edge detection” method, and a frame comparison method. Here, the “background difference” method and the “edge detection” method are taken as an example, and the operation when the “background difference” method for extracting the tracking point of the vehicle image from the video information of the camera 4 is set first. explain.

  In the “background difference” method, a camera image of a road that monitors a vehicle is captured and stored in the internal memory of the feature extraction unit 11 as background screen information. For example, video information of a camera that captures a road on which a vehicle is not traveling is background screen information. Since this background screen information changes depending on the sunshine and the surrounding environment of the road, processing and updating such as averaging are performed according to predetermined setting conditions.

  The video information photographed by the camera 4 is obtained for each frame by the feature extraction unit 11 with reference to the background screen information, and the difference video information is further binarized vehicle image information. And the tracking point of a vehicle is further extracted from the information of this vehicle image.

FIG. 2 is a diagram for explaining the “background difference” method of the vehicle image.
FIG. 2A shows an example of vehicle display on the screen of the monitoring monitor 21. The vehicle shape of the difference video information is displayed in the same manner as the original image. FIG. 2B shows the video signal amplitude of the difference signal.

  In FIG. 2B, the 1-1 part is the maximum amplitude value of the video signal, which is the brightest part of the vehicle, and the 2-2 part is the video signal level corresponding to the background. The 3-3 portion is a threshold level Ga for extracting the vehicle portion, and is an intermediate value (or average value) between 1-1 and 2-2.

  The threshold level is set by several parameters such as a value Gb of 1-1 near the maximum video signal level, a value Gc of 2-2 near the lowest signal level, and a fixed signal level. Is done.

  In FIG. 2 (b), the video signal level of the white painted vehicle shows a convex signal level change in the vehicle portion brighter than the background. However, in the case of a black painted vehicle, the signal level of the vehicle is higher than the background. Falls and becomes concave. The signal level and threshold of the vehicle portion vary depending on the paint color, lighting, background color, etc. of the vehicle.

FIG. 3 is a binarized image diagram in which the black portion is the vehicle portion.
FIGS. 3A, 3B, and 3C are binary images whose threshold levels correspond to Ga, Gb, and Gc, respectively. 3 (a) to 3 (c), the shape of the binarized image differs depending on the threshold level. In FIG. 3 (b), the dark part of the vehicle body is deleted, and one vehicle image is formed by a plurality of blocks. doing.

  In each of FIGS. 3A to 3C, a mark frame generated by taking out the outermost tangent in the vertical and horizontal directions of the outer portion of the vehicle portion is a tracking gate TGb of the vehicle image. In FIG. 3B, since there are a plurality of blocks, there is a risk that one vehicle is erroneously determined as a plurality of vehicles. On the other hand, in FIG. 3C, when the video signal level difference (contrast) with the background is small, the binarized vehicle image becomes large, and further, the vehicle cannot be distinguished from the background and the tracking gate TGb cannot be generated. There is danger.

  When the feature extraction unit 11 receives the video information from the camera 4 and generates the tracking gate TGb, the feature extraction unit 11 generates a lock-on signal and transmits it to the internal bus. When the event detection unit 12 receives the lock-on signal, calculation of the traveling speed of the vehicle is started.

  As for the traveling speed of the vehicle, the moving speed on the screen is measured using the corner portion of the tracking gate TGb as a tracking point. One of the four corners of the tracking gate TGb is selected as the tracking point.

  In FIG. 2, since the vehicle is traveling from left to right, the lower left corner close to the rear ground contact surface of the vehicle road is selected as the tracking point LL.

  The combination of the above-described algorithm “background difference” for detecting the vehicle, the threshold level of binarization (here, Ga), and the tracking point LL is used as a parameter of the extraction condition of the image information as an abnormal event described later. It is set and stored in advance in the table T of the determination unit 13 and the table TL of the data server 25 together with the determination condition parameters.

  These settings are performed in a predetermined procedure from an input device such as a keyboard of the console 22 of the monitoring center 2. For example, initial setting is performed in the log table TL of the data server 25, and the set data is transferred to the network 3. Then, it may be downloaded and set in the table T of each image processing unit 1. Alternatively, an adjustment personal computer may be connected via the control / communication unit 14 of the image processing unit 1 and input setting may be performed from the keyboard.

  FIG. 4 is an example of the table T of the determination unit 13 in which the determination conditions are listed.

  In FIG. 4, the determination conditions are “feature extraction method” (“algorithm”, “threshold”, “tracking point”) of the extraction operation parameter, and “event determination condition” (“measurement interval / distance”) of the detection operation parameter. “Measurement frequency”, “use of judgment”) and the like. In the “determination use” column, “O” is set for a determination result to be notified to the monitoring staff among the determination conditions of the abnormal event used for alarm notification.

  The event detection unit 12 calculates the traveling speed of the vehicle by calculating the speed at which the tracking point LL extracted by the feature extraction unit 11 moves on the screen. The vehicle speed calculation condition is set in the log table TL and the table T as a parameter of the abnormal event determination condition.

  Here, the traffic situation is classified into, for example, two abnormal events of “normal”, “low speed”, and “traffic jam”, and examples of conditions for determining abnormal events are as follows.

(Low speed judgment condition)
One of the following conditions is satisfied.
1) When the average vehicle speed is 40 km / h or less.
2) When the rate of change of speed calculated in the section of 20m or more is 50% or more.

(Congestion criteria)
Those that satisfy the following conditions.
1) When the average speed is 40 km / h or less and the traffic volume for 5 minutes is 25 or more.
2) When the traffic density is above a certain value of 40 cars / km / h lane.
3) When the vehicle group speed is a constant value of 40 km / h or less.
Defined by

  The conditions for judging abnormal events in the traffic situation include a combination of the above, but in the following embodiments of the present invention, the description of the operation of the camera monitoring system is limited to only the item 3) of the criteria for judging “congestion”. The procedure for determining abnormal events will be described.

  FIG. 5 is a diagram for explaining a display screen on the monitoring monitor 21 and a method for measuring the traveling speed of the vehicle.

  On the screen of the monitoring monitor 21, the vehicle travels from left to right on a two-lane highway. Of the two lanes, the first lane is displayed in the lower half of the screen and the second lane is displayed in the upper half of the screen. When the traveling vehicle is not shown on the screen, the frame and the tracking gate TGb are not displayed on the screen.

  When the vehicle travels from the left side of the screen in the first lane and enters the screen, the tracking gate TGb is generated and displayed on the screen of the monitoring monitor 21 according to the above procedure. A lock-on signal is transmitted to the unit 12. Then, the event detection unit 12 starts calculating the speed at which the tracking point LL moves on the screen. The moving speed of the tracking point LL in the vertical and horizontal directions on the screen is calibrated so that the position coordinates on the screen correspond to the position on the road and converted to the traveling speed of the vehicle.

  The event detection unit 12 calculates the traveling speed of the vehicle with reference to the event determination conditions in the table T.

  For example, in FIG. 5, when the x1 point and the x2 point on the screen have a distance corresponding to 100 m on the road, if the tracking point LL locked on the section passes in 10 seconds, the feature extraction unit 11 calculates 36 km / h. Further, if the x1 point and the x3 point on the screen correspond to a road distance of 25 m, if the tracking point LL moves between them in 5 seconds, the speed is 18 km / h.

  Further, if the x4 point and the x2 point on the screen correspond to 25 m, if the tracking point LL moves in that section in 2 seconds, the speed is 45 km / h.

  The event detection unit 12 transmits the traveling speed of the vehicle calculated from the moving speed of the tracking point LL to the determination unit 13, and the determination unit 13 has a traffic situation and an abnormal event at any level from the traveling speed. Judgment is made by discrimination.

  By the way, the measurement result of the vehicle traveling speed in each section (hereinafter referred to as vehicle speed data) is used for determination as the vehicle group speed of the traffic situation (abnormal event). Since there is a determination condition consisting of a combination of measurement section distance and measurement frequency, several different measurement results can be obtained correspondingly.

  That is, a vehicle group is called a vehicle group when a plurality of vehicles travel in a lump on the road. However, it is difficult to strictly set the vehicle group as a unit of speed measurement, and in the following embodiments, the average hourly speed of the vehicle sampled with the following operation parameters is used as the vehicle group speed.

For example,
(1) The average speed of 36 km / h obtained as a result of measuring 10 vehicle speed data at the x1 and x2 points at a time at 5 minute intervals is defined as the average speed data (x1-x2).
(2) The average speed 18 km / h obtained as a result of continuously measuring 10 vehicle speed data between the x1 point and the x3 point at a time every 5 minutes is defined as the average hourly speed data (x1-x3).
(3) The average hourly speed of 45 km / h as a result of continuously measuring 10 vehicle speed data between the x2 point and the x4 point at a time at intervals of 5 minutes is defined as the average hourly speed data (x2-x4).
(4) The average speed of 36 km / h obtained as a result of measuring vehicle speed data between the x1 point and the x2 point at five consecutive times at 2-minute intervals is defined as the average hourly speed data (x1-x2).
(5) The average speed 20 km / h as a result of measuring vehicle speed data between the x1 point and the x3 point five consecutive times at intervals of 2 minutes is defined as the average hourly speed data (x1-x3).
(6) The average speed 50 km / h obtained as a result of measuring vehicle speed data between the x2 point and the x4 point five consecutive times at intervals of 2 minutes is defined as the average hourly speed data (x2-x4).

  If the parameters such as the measurement frequency and the measurement section distance are changed as in the case of, etc., the traffic volume will change, and especially if the measurement conditions change such as when there is a curve or junction at the road like an interchange, the vehicle speed data will be calculated differently Is done.

  In the above example, the point x1 is near the point where the interchange merge ends, and x2 is the farthest from the merge end point. Therefore, in order to obtain the average vehicle speed of the video monitoring section from the control / communication unit 14 to the determination unit 13, the vehicle speed data in the 100m section (1) (here, the average speed of 36 km / h) is set to be used. To do.

  Since the vehicle speed is 40 km / h or less according to this setting, the determination unit 13 determines that the “congestion” is an abnormal event and transmits “congestion” to the control / communication unit 14.

  The control / communication unit 14 alerts the abnormal event of “traffic jam”, sets the measurement condition of (1) as the determination condition of the vehicle speed data, and transmits the vehicle speed data of (1) to (6) above to the network 3. To the data server 25 and the control unit 24 of the monitoring center 2. In the data server 25, these determination conditions, vehicle speed data, warnings, and the like are time information (hereinafter referred to as a time tag. Here, XX year OO month 12:10:30) and the log table TL. To remember.

  FIG. 6 is a diagram of an example of the log table TL.

In FIG. 6A, the evaluation column shows the result of evaluating the alarm (abnormal event determined by the determination unit 13) issued by the monitor by “correct detection”, “false detection”, and “not detected”. Remembered. In addition, even when the determination conditions (2) to (6) other than the above (1) are attached, the control unit 24 collates with the evaluation result for (1), and a monitor is assigned to the determination result based on the determination condition. Evaluation contents estimated to be evaluated are stored. (In the log table TL in FIG. 6, it is displayed in parentheses.)
By the way, in the monitoring center 2, when the control unit 24 receives “traffic jam” via the network 3, the monitoring monitor 21 displays the video of the camera 4 and displays a “traffic jam” alarm on the screen of the monitoring monitor 21. Do.

  In the monitoring center, data obtained by directly measuring the vehicle speed by speed measurement radar is acquired from the traffic information system by a method other than video monitoring for the vehicle speed data, or in a service area that is several kilometers away or near an interchange. If traffic information and the like obtained by reading the characters on the license plate of the traveling vehicle from the images of the two installed monitoring cameras and measuring the traveling speed of the vehicle can be obtained, they can be used to determine an abnormal event. In such a case, when an alarm is received, the monitoring person refers to the above information and empirically reads the degree of congestion of the vehicle on the screen of the monitor 21 and the moving speed, and determines “normal”, “congestion”, and the like. Yes.

  However, if there is no means for knowing the speed of the vehicle other than the image processing apparatus, “normal”, “traffic jam”, and the like are determined based on the experience of the observer from only the video of the monitoring monitor 21. Therefore, the judgment includes errors, especially for “congestion”, because various data such as vehicle density necessary for judging “congestion” cannot be obtained from other traffic information systems, etc. I have to rely on the experience of the observer. However, in the camera monitoring system of the embodiment of the present invention, it is possible to improve the consistency between the determination result made by the determination unit 13 and the evaluation based on the determination by the monitoring staff (the correct answer rate is improved) as follows. .

  First, the surveillance staff refers to the traffic information of 80 km / h in the traffic information created from the driving data of the vehicle from which the license plate is read, and the surveillance camera 4 visually determines the traffic condition of the interchange and determines that it is “normal”. To do. In the camera monitoring system, the vehicle speed data of the above (1) is used to determine “congestion”. Therefore, the monitoring staff evaluates that the determination result of the camera monitoring system is incorrect and evaluates the evaluation switch 23 on the console 22. Enter the evaluation result of “false positive”. The evaluation result of “false detection” is stored in the log table TL of the data server 25 together with a time tag (here, XX year OO month 12:10:30).

  On the other hand, when the monitoring person determines that the traffic is “congested” when there is no alarm such as “congested” from the camera monitoring system, the monitoring person inputs “not detected” from the evaluation switch 23 and the evaluation result “ “Not detected” is stored in the log table TL. When the camera monitoring system notifies the “traffic jam” alarm, if the monitor also determines “traffic jam”, the evaluation result of “correct detection” is stored in the log table TL.

  Hereinafter, an abnormal event notification (warning) from the image processing unit 1 or an event in which the monitoring person determines an abnormal event is referred to as an “event”.

  In the conventional camera monitoring system, the result of the “false detection” is only recorded by the monitor using a monitoring note or a takeover memo. Therefore, the data acquisition for setting the criterion for determining the vehicle speed data to the optimum value, the efficiency of setting and adjusting the operation parameters, and the improvement result have been insufficient.

  However, in the camera monitoring system of the present invention, the operation parameters and determination conditions of the camera monitoring system for each event, the determination result, and the evaluation result are stored in the log table TL of the data server 25 together with the time tag. Accordingly, by referring to the time tag for each event going back in the past, various operation parameters of the feature extraction unit 11 and the event detection unit 12, the determination criteria, the determination result, and the evaluation result of the supervisor are read out and collated. The operation of the camera monitoring system can be analyzed, and based on the result, adjustments can be made to improve the operation parameters at the time of “false detection” or “not detected”.

  For example, a monitor or a maintenance engineer of the camera monitoring system can read out and adjust the data from the data server 25 to the camera video information and the log table TL using the keyboard of the console 22 and the monitoring monitor 21, for example. Read and analyze the stored data. Based on these data, data such as each determination condition and operation parameters for controlling each component of the image processing unit 1 are set again in the control unit 24, and each component of the image processing unit 1 is set via the control unit 24. Is transmitted to the control communication unit 14 to adjust various image processing operations.

  Further, for the adjustment and the like, the control unit 24 can set various parameters by incorporating a program in the control unit 24 in advance.

(Change of abnormal event judgment condition parameter)
Hereinafter, an operation in which the control unit 24 acquires data from the log table TL of the data server 25 and controls the processing operation of the image processing unit 1 will be described.

  FIG. 7 is a flowchart for explaining the change of the parameter of the event determination condition.

  The control unit 24 receives image processing data such as determination conditions and determination results of abnormal events, evaluations of monitoring personnel, determination condition candidates listed in the table T, and vehicle speed and the like in the log table TL of the data server 25, and the time Store with tag (step 1).

  In the above description, the determination condition (1) is adopted and it is determined as an abnormal event of “traffic jam”, and an alarm is issued. However, the monitoring person who receives it has determined “false detection”. Therefore, the control unit 24 estimates that the determination condition that the calculation result is 40 km / h or less, which is the determination condition for “congestion”, is evaluated by the monitor as “false detection” (step 2). In addition to the determination conditions (3) and (6), the estimation result of “false detection” is stored in the “evaluation” column of the log table TL (step 3).

  Then, the control unit 24 refers to the log table TL and performs control for selecting determination conditions for the next and subsequent times.

  For example, the table T of the determination unit 13 is set so that the abnormal event determination condition for which “correct detection” is obtained for the latest event including the estimated evaluation is adopted for the next event determination. If the monitor evaluates that the event is not “correct detection” (“false detection” or “not detected”), the control unit 24 refers to the log table TL as “ The determination condition (in this case, (3)) for which “correct detection” has been performed is found, and (3) is set as the determination condition from the next time (step 5).

  If other determination conditions for statistically processing the estimated evaluation result are selected (No in step 4), the control unit 24 determines the evaluation result of the observer for each event and the time corresponding to the event. Are statistically processed (step 6), and a determination condition (here, (6)) corresponding to “correct detection” is searched from the log table TL. In the determination, the determination unit 13 uses the determination condition (6) (step 7) to select a determination condition that enhances the consistency of the experience of the observer and traffic information from a system other than video monitoring. You can also.

  In addition, in order to investigate the operation of the camera monitoring system in detail, the determination condition may be changed and corrected separately by statistical analysis. For example, the evaluation result is a judgment result obtained from a plurality of judgment conditions of the camera monitoring system (here, “congestion”) and a judgment result of the surveillance staff (here, “false judgment” based on the judgment of “normal”). )), The results of the judgment result of “congestion” with many judgment criteria are obtained, and the judgment result of the camera monitoring system is estimated to be more appropriate than that of the observer. In this case, the control unit 24 may display and notify the recommendation information for reviewing the judgment and evaluation of the monitoring person on the monitoring monitor 24.

  In the above description, one determination condition is selected in advance. However, a plurality of determination condition candidates are set, and the determination results of the respective determination condition candidates are totaled to obtain a majority decision (here. Then, “congestion” is selected), and a determination result for notifying the observer of “congestion” may be used.

  As described above, the camera monitoring system according to the embodiment of the present invention can improve the correct answer rate by referring to the stored data and changing and setting the event determination condition assuming an optimum combination. is there.

(Select algorithm)
Furthermore, in order to improve the correct answer rate of the camera monitoring system, it is also effective to select an algorithm for extracting a vehicle image according to the surrounding situation. The operation procedure for selecting an image extraction algorithm will be described below.

  FIG. 8 is a flowchart illustrating an operation for explaining an operation procedure for selecting a feature extraction algorithm.

  The embodiment of the present invention uses the “background difference” method as an algorithm for extracting a vehicle image from a video signal as described above. Since this algorithm extracts the entire vehicle image, stable tracking is possible, so the “background difference” method is adopted as the default algorithm during the daytime.

  However, in the “background difference” method, as shown in FIG. 3C, the level difference of the video signal between the vehicle body portion and the background portion becomes small when the vehicle body is an intermediate color at night or in the daytime or daytime. It is difficult to extract the part and it becomes impossible to lock on the vehicle image, and if the body image has a bright point that is extremely high in brightness compared to the vehicle body like a lamp, the whole vehicle body cannot be extracted. . In addition, when a large vehicle body such as a trailer or a plurality of vehicles are continuously photographed, one vehicle body part does not fit in the tracking gate TGb, so that the vehicle body part cannot be extracted from the video information. .

  In such a case, the vehicle speed is not measured, and the camera monitoring system determines “normal” without detecting an abnormal event, so no alarm is transmitted (step s11). Then, the evaluation based on the result of the monitor determined as an abnormal event by screen monitoring is recorded, and if the monitor determines “congestion”, “not detected” is stored in the “evaluation” of the log table TL together with the time tag. (Step s12).

  In this way, when the shooting situation is such that feature extraction is difficult, “undetected” evaluation inputs increase. Therefore, the control unit 24 changes the threshold level setting when the evaluation input frequency of “undetected” assumed in advance, for example, exceeds 3 times in 10 minutes (Yes in step s13) (step s14).

  FIG. 6B is an example of the log table TL when it is evaluated as “not detected”.

  For example, the vehicle image extraction is improved by changing the parameter setting to match the image characteristics such as setting the threshold level Gc close to the background level, or the threshold level Gb having the largest video signal amplitude in the gate. Plan. A signal for controlling the setting change is transmitted from the control unit 24 to the control / communication unit 14 of the image processing unit 1 through the network 3, and the control / communication unit 14 that receives the signal changes the table T of the determination unit 13. Referring to, a new parameter is set in the feature extraction unit 11.

  The camera monitoring system according to the embodiment of the present invention includes the “edge detection” method for detecting a part of the image of the vehicle body as the vehicle part extraction algorithm in the feature extraction unit 11. If image extraction fails, change to the “edge detection” method to track the vehicle. For example, the control unit 24 receives “not detected” at a predetermined frequency or more, for example, three times in succession (Yes in step s15) and transmits a signal instructing the algorithm change to the control / communication unit 14 to control / The communication unit 14 sets the “edge detection” method in the feature extraction unit 11 (step s16).

  FIG. 9 is a diagram for explaining the “edge detection” method.

  In the “edge detection” method, unlike the “background difference” method, a standby gate g for tracking the vehicle is provided on the screen in advance.

  That is, the background of the vehicle image is not necessarily uniform, and a plurality of vehicles are displayed on the screen. In order to facilitate the extraction of the image signal of a specific vehicle, the vicinity of the image portion of the vehicle portion in advance A standby gate g for extracting and processing only the signal is set and ambushed, and the edge (end) of the vehicle that has entered the standby gate g is tracked (locked on, tracked).

  The standby gate g is placed at a position where the traveling vehicle starts to appear in the first lane and the second lane and waits. Also, the size is set by the control / communication unit 14 as “large”, “medium”, and “small”.

  FIG. 9A is an example of a vehicle shape display on the screen of the monitoring monitor 21.

  FIG. 9B shows the amplitude of the video signal of the vehicle display portion on the horizontal scanning line. FIG. 9C shows the amplitude when the video signal passes through a differentiating circuit.

  In the “edge detection method”, the feature extraction unit 11 differentiates the video signal of the rising or falling portion of the video signal of the vehicle portion to differentiate the end (edge) portion of the vehicle into pulsed edge signals Ep1 and Ep2. To generate a signal indicating an end portion of the vehicle. The “edge detection” method has an advantage that the gate can be made small because either one of the head portion (edge signal Ep1 of the positive pulse) or the rear portion (edge signal Ep2 of the negative pulse) is taken out. Since there is, keep the size as small as possible to track.

  Therefore, when the traveling vehicle enters the standby gate g and detects an edge signal, the gate is contracted to be reduced only in the vicinity of the edge, and the vehicle is tracked by the tracking gate TGe reduced thereafter. Then, tracking of the end portion of the vehicle image is performed so that the pulse of the edge signal comes to the center of the tracking gate TGe (step s17).

  Also, at night, a pulse signal or a signal with a very large contrast difference can be obtained even in a point image such as a headlight or tail lamp. To track using the information of the point image, the “edge detection” method is used. Is suitable.

  The part for tracking the vehicle part is selected by changing the threshold of the edge signal. That is, the head portion of the vehicle that appears brighter than the background on the screen is the edge signal Ep1 of the positive pulse, but the rear portion is the negative edge signal Ep2. On the other hand, if the vehicle appears darker than the background, the relationship is reversed. The signal level of the background and the vehicle portion is automatically detected to reflect the above relationship, and whether the leading portion or the trailing portion is extracted is preset during system adjustment. Further, this setting change may be set by inputting a command for selecting an extraction portion to the control unit 24 by a maintenance engineer or a monitor from an input device such as a keyboard of the control console 22.

  The control unit 24 transmits the command to the control / communication unit 14 of the image processing unit 1 via the network 3. The command is further notified from the control / communication unit 14 to the feature extraction unit 11, and the feature extraction unit 11, for example, when extracting the tail portion, changes the threshold from negative pulse detection to positive pulse detection. Changed to extract the first part from the part.

  When a lock-on notification is received from the feature extraction unit 11, the subsequent processing and operation for acquiring vehicle speed data performed by the event detection unit 12 are performed in the same manner as in the “background difference” method. These various settings are shown in conditions (7) to (12) of the table T in FIG.

  In this manner, the feature extraction unit 11 can extract vehicle image information from the video information from the camera 4 together with the “background difference” method by the “edge detection” method, so that an image extraction algorithm that can more reliably extract the vehicle image is selected. Thus, accurate tracking can be performed and the correct answer rate can be improved. Further, as described above, methods and algorithms other than these two methods may be used as the vehicle image extraction method.

  As described above, the camera monitoring system according to the present invention includes the video information of the camera, the determination condition for determining the abnormal event by processing the video information, the data of the image signal processing algorithm, the operation parameter, and the abnormal event. The determination result is stored in the data server together with the time information so that the monitor can compare and collate the three types of evaluation results for the determination.

  Then, the stored data and the evaluation result are collated and analyzed, and the correct condition rate can be increased by adjusting or automatically changing the determination condition, the image signal processing algorithm and the operation parameter so that the abnormal event is correctly detected.

The block diagram which shows the component of an example of the camera monitoring system concerning the Example of this invention. The figure explaining the background difference method of an image signal processing algorithm. The figure of the vehicle image binarized with the background difference method of the image signal processing algorithm. The figure which shows the table of the list | wrist of the judgment conditions of the camera monitoring system concerning the Example of this invention. The figure which shows an example of the measuring method of the traveling speed of the vehicle of the camera monitoring system concerning the Example of this invention. The figure of the log table of the judgment conditions and operation data of the camera monitoring system concerning the Example of this invention. 6 is a flowchart for explaining a procedure for changing a determination condition parameter for an abnormal event in the camera monitoring system according to the embodiment of the present invention. The flowchart explaining the change procedure of the image extraction algorithm of the camera monitoring system concerning the Example of this invention. The figure explaining the edge detection method.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Image processing part 11 Feature extraction part 12 Event detection part 13 Determination part 14 Control / communication part 2 Monitoring center 21 Monitoring monitor 22 Console 23 Evaluation switch 24 Control part 25 Data server 3 Network 4 Camera T Table TL Log table

Claims (9)

A camera monitoring system for monitoring traffic conditions by video information of a camera,
Clock means for outputting time information for referring to the operation of the camera monitoring system;
Extracting means for extracting vehicle image information from video information of the camera;
Detecting means for detecting abnormal event information related to traffic conditions from the extracted image information;
Determination means for discriminating and determining an abnormal event from the detected abnormal event information;
Notification means for notifying the camera monitoring system of the determination result of the determined abnormal event;
Whether the monitoring personnel evaluated the notified abnormal event determination result as correct positive detection, or evaluated it as a false detection that an abnormal event has occurred even though no abnormal event has occurred, or an abnormal event has occurred An evaluation input means for inputting any one of the evaluation results from the evaluation that the abnormal event did not occur but was not detected,
A storage for storing the video information, the image information, the abnormal event information, a determination condition determined by the determination means, the determination result of the abnormal event, the evaluation result, and the time information Means,
A camera monitoring system comprising control means for monitoring and controlling each of the means. The storage means
A first table for storing a determination condition including an extraction operation parameter of the extraction unit and a detection operation parameter of the detection unit;
Corresponding to the determination condition stored in the first table, the extracted image information, the detected abnormal event information, the determination result of the abnormal event determined by the determination means based on the determination condition, A second table for storing the evaluation result and time information;
The control means includes
Referring to the first table and setting the determination condition in the determination means;
Referring to the time information in the second table, the video information, the set determination condition, the extracted image information stored in the second table, the detected abnormal event information, the abnormality The camera monitoring system according to claim 1, wherein control is performed to collate an event determination result and the evaluation result. The first table is:
A first determination condition based on a combination of the first extraction operation parameter of the extraction unit and the first detection operation parameter of the detection unit, and a combination of the first extraction operation parameter and the second detection operation parameter And the second determination condition by
The control means includes
A first determination result obtained by the determination unit discriminating and determining the detected abnormal event according to the first determination condition is notified to the monitor via the notification unit and stored in the second table. Control,
3. The camera according to claim 2, wherein the determination unit performs control to store a second determination result obtained by discriminating the detected abnormal event based on the second determination condition in the second table. Monitoring system. The means for extracting the image information includes:
It consists of multiple image processing algorithms,
The first table is:
4. The camera monitoring system according to claim 3, wherein one or more extraction operation parameters are stored for each image processing algorithm. The control means includes
When the evaluation result input from the evaluation input means and the first determination result notified to the monitor are compared and collated, and the positive detection is not detected, the second determination result and the input evaluation If the result of comparing the second determination result corresponds to the positive detection,
The control for changing from the first determination condition to the second determination condition is performed as a determination condition for obtaining a determination result to be notified to the monitor of the next abnormal event. Camera surveillance system. The control means includes
If the evaluation result input from the evaluation input means is not the positive detection, the evaluation result input from the evaluation input means stored in the second table and the first notified to the monitor Statistically process the record of the determination result and the second determination result retroactively,
If the second determination result in the previous period has a better match rate with the evaluation result than the first determination result,
4. The control for changing a determination condition for obtaining a determination result of an abnormal event to be notified to the monitoring person from the next time to the second determination condition from the first determination condition is performed. Camera surveillance system. The control means includes
Changing the first or second detection operation parameter of the determination condition to generate a plurality of determination condition candidates;
The evaluation result input from the evaluation means stored in the second table, the determination result based on the first determination condition notified to the monitor, and the plurality of determination results based on the plurality of determination condition candidates The recorded results of
Select the determination condition with the best match rate with the evaluation result from all the determination results,
4. The camera monitoring system according to claim 3, wherein the determination of the abnormal event after the next time is performed by changing the first determination condition to the selected first determination condition. The control means includes
A plurality of determination condition candidates are generated by changing a combination with the first or second detection operation parameter,
A first determination result determined according to the first determination condition;
4. The camera according to claim 3, wherein control is performed such that a determination result by majority vote obtained by aggregating the determination results determined by the plurality of determination condition candidates is selected as the determination result of the abnormal event and transmitted to the notification means. Monitoring system. A camera,
Clock means;
Extracting means for extracting vehicle image information from video information of the camera;
Detecting means for detecting abnormal event information from the extracted image information;
Determination means for discriminating and determining an abnormal event from the detected abnormal event information;
Notifying means for notifying a camera monitoring system monitor of the determination result of the determined abnormal event;
Whether the monitoring personnel evaluated the notified abnormal event determination result as correct positive detection, or evaluated it as a false detection that an abnormal event has occurred even though no abnormal event has occurred, or an abnormal event has occurred An evaluation input means for inputting any one of the evaluation results from the evaluation that the abnormal event did not occur but was not detected,
Storage means;
A monitoring control method for a camera monitoring system for monitoring a traffic situation comprising a control means for monitoring and controlling each means,
The extraction means includes
The image information is extracted from the video information of the camera by the extraction operation parameter stored and set in the storage means,
The detection means includes
Detecting abnormal event information of traffic conditions from the extracted image information by the detection operation parameter stored and set in the storage means, and transmitting to the determination means,
The determination means includes
When the abnormal event information is received, the determination is made by comparing the determination condition stored in the storage means and discriminating the abnormal event,
The storage means
The extraction operation parameter of the extraction means, the detection operation parameter of the detection means, a plurality of determination conditions candidates for the abnormal event comprising a combination of the parameters, and at least one determination condition among the stored candidates The determination result determined by the determination means, the determination result of the determination result notified to the monitoring camera system by the notification means, the evaluation result input from the evaluation input means, and the clock means received Remember time information,
The control means includes
Control the notification means to select one of the candidates and notify the determination result based on the determination condition,
The evaluation result input from the evaluation input unit is read by the monitoring person who has received the notification, and when the evaluation result is not the correct detection, the data stored in the storage unit is read with reference to the time information Searching for the image information, the abnormal event information, the extraction operation parameter, and the detection operation parameter;
When a new determination condition for the combination of the extracted operation parameter and the detected operation parameter whose evaluation result is the positive detection is found, the next notification is determined by discriminating the abnormal event based on the new determination condition. A monitoring control method for a camera monitoring system, characterized in that control is performed to modify and change so as to notify a result.

Images