JPS6261178A - Method and apparatus for monitoring moving body - Google Patents

Abstract

PURPOSE:To remove influence due to the change of weather or illumination by finding out a difference between a current image signal and a time average value, and when the difference exceeds a regulated value and a special change more than a regulated value exists, deciding the invasion of a moving body. CONSTITUTION:Since a ring pointer system for storing difference data found out at present in the address of the oldest contents of a difference memory is adopted in a monitor mode and a value obtained before TNsec from the current time is always used, data are not influenced by peripheral continuous change. When the device is decided as normal after generating an alarm and checking the state by the operator, a registration button 16a is depressed to adopt the value obtained at the generation of the alarm newly as an allowable value. Such kind of study makes it possible to provide the machine with a deciding mechanism through the uncertain deciding function is ordinary applied to the operator.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Application of the Invention] The present invention relates to a method and apparatus for monitoring a moving object (eg, one person, two cars, a train, etc.).

[Background of the invention]

In a building or facility, if a moving object approaches a certain area, it may be dangerous or inconvenient. That is, for example, if a person falls onto the railroad tracks inside a station, the person's life is at risk. Therefore, it is necessary to monitor and prevent it. This kind of thing is not limited to station premises, but also the boarding and alighting areas of amusement park rides, the escalator boarding and alighting areas of department stores, etc.
It is also necessary in places where there is a possibility of contact between moving objects (for example, people and cars).

Now, the ``System for Detecting Moving Persons in Outdoor Environments'' presented by Kotani, Nakamae, and Nishida at the 1981 China Chapter Federation of Electrical Engineers of Japan Conference was known. This system detects when a person enters a dangerous premises and issues an alarm. Images are captured by an ITV and processed by a microcomputer via an image input device for a personal combinator. In this system, in order to distinguish between a still image and a person, the presence of a person is detected by comparing the measured brightness value with the brightness value of the background when no person is present as a reference value.

However, because this system uses a fixed background brightness value when no people are present, the input brightness value changes significantly due to changes in the weather, and it has been reported that this change can cause malfunctions. ing.

In addition, Japanese Patent Application Laid-open No. 59-99580 discloses that a moving object moving in a stationary background is moved by increasing the difference between video signals at two adjacent times to erase complex background information. Extract (cut out) information about only the object,
Cut out the target original image based on the extracted information,
A technique for performing a pattern LH job using the cut out original image has been disclosed. However, this technology is related to the recognition of moving objects, and is not related to a moving object monitoring system that detects that a moving object has entered a dangerous area and issues an alarm.

[Purpose of the invention]

SUMMARY OF THE INVENTION An object of the present invention is to provide a highly reliable method and device for monitoring moving objects that is not affected by weather or lighting variations.

[Summary of the invention]

The present invention captures an image of an area where a moving object may exist with a camera, stores the digitalized image signal in an image memory, calculates the temporal average value of the brightness of the stored image signal, Calculate the difference between the input current image signal and the temporal average value, and if this difference exceeds a specified value and there is a spatial change greater than the specified value, it is determined that a moving object has invaded. It is characterized by

[Embodiments of the invention]

Hereinafter, the present invention will be explained in detail using specific examples.

Although the following embodiments will be explained using a system for monitoring fallen persons inside a station as an example, the present invention is not limited to such a system. It can be applied to any other system with

Now, Figure 2 shows cameras inside the station (for example, TV
An example of the arrangement of cameras) is shown below. ! In Figures 1 and 2, 1 shows the train, 2 the platform, and 3 the tracks. The camera is
Cameras 4a, 4b, and 4c are arranged along the moving direction of the train so that they can image the platform and track areas. Cameras 5a, 5b, and 5c are similarly arranged on the opposite lane. Cameras have a limited field of view, so one camera cannot cover the entire area of the platform. For this purpose, multiple units have been installed. In addition, in detecting fallen persons, in order to ensure the detection of the smallest detected object such as a child, it is necessary to increase the resolution, but in order to increase this resolution, it is necessary to image a narrow area as large as possible. However, the imaging area of the camera is limited.

In other words, as shown in Figure 3, if the image monitored by one television camera is decomposed into M x N by digitization, the area monitored by the camera is actually Mo (m) x
When NO(m) is assumed, the resolution per pixel is generally said to be 3 pixels for image distortion including the lens system, so if the minimum detection object for an object with a margin is 5 pixels, then Its resolution is 5×△Px, 5X△P
, becomes.

Next, the hardware configuration of this system will be explained.

! FIG. 81 is a control block diagram of this system.

A video signal is input from the television camera 4 to an image input section 6, where the analog video signal is digitized and replaced with binary coded data according to brightness. The timing of digitization here is synchronized with the clock pulse output from the oscillator 7.

The digitized video signal is input to a switching circuit 8 that switches image memory access, and is input via the switching circuit 8 to an image memory 9m that stores the current image. The image memories 9a and 9b store the TV screen in MxN format as shown in FIG.
The minimum unit of the division, that is, each pixel, is divided into four parts as binary data according to the brightness level described above.
As shown in the figure, data is stored in order according to each pixel starting from the edge of the screen.

The switching circuit is connected to an image memory 9b that stores image data for calculation, and an arithmetic processor 10 that reads data between these image memories and performs comparison calculations.

The arithmetic processor 10 includes a memory for storing programs (
ROM) ua, memory as work memory (R
AM) llb is connected, the output circuit is connected for interface with peripheral devices, and the display control unit port is connected for outputting the video signal of the monitor television 14. Operation unit 1
6 is composed of a start push button 16b and a registration push button 16a in this embodiment. The operation unit 16 is
The present invention is not limited to such a configuration, and any device such as a keyboard that can interface between a person and a machine may be used. In this embodiment, the alarm unit is composed of a visual alarm device 15b such as a lamp or a display device, and a listener@device i 15m that issues an alarm by a buzzer, voice, or the like. This alarm unit may also be of a type other than those described above.

In platform monitoring, in order to detect a foreign object within the track, it is sufficient to compare the image of the standard state and the image of the current state. However, the condition of the station platform +4
It changes from moment to moment depending on the movement (number) of passengers, the approach of vehicles, the clothes of passengers, the weather, etc. Therefore, rather than a fixed standard state, it is necessary to create a standard state that responds to changes.

In this system, a standard state image is created using the following procedure. That is, it is created based on the operation flow shown in FIG. 1O.

First, if there is no reference image, the standard speed registration mode is entered in step F1, and image sampling is performed for TN seconds. As for the processing during these 18 seconds, in the first sampling, the current image is transferred to the calculation image memo 'J9b. This is step F3. F5. This is the process of F9. In the next sampling, the current image is transferred to the image memory 9m, the calculation processor 10 calculates the difference between the image memory 9a and the calculation image memo 'J9b, and the value of the difference is saved as a work memo! Each pixel is stored in Jllb. In this way, the number of sampling times N in 18 seconds
Repeat until the end of the test. This process is performed in steps Fil and F.
13. F15. It is F17. Regarding the difference values obtained N times, the arithmetic processor 10 calculates, for each pixel, the maximum difference value (that is, the one with a large luminance change), the minimum difference value j, and the difference value of the sample after IIT from the work memory 11b. Calculate the maximum value distribution of changes in and the average value. This process is step F19. If the image of the platform changes from the standard state during this registration mode, cancel the above operation and
Run registration mode again. This operation requires the operator to
Monitor the home status in registration mode.

If it is determined that there is an abnormality, this is achieved by pressing the registration push button 161 again. That is, the interruption caused by this button operation causes the steps to be re-executed in sequence from step F1 in FIG.

If the result of the difference at a certain pixel is as shown in FIG. 5, the arithmetic processor 10 examines samples 1 to N in order, and the maximum value is 10. The minimum value is 1
-3 for the second time, the maximum value of change between samples is 2 from the first time
Absolute value when changed to 10. The average value is obtained by calculating the sum of the values from the first to N times and dividing this by N. Expressed as a formula, it is the average! 11 Ay is, where: Si: Each difference value N: Number of samples Also, as for the distribution, as shown in FIG. 6, the frequency distribution of the difference values is examined to find the portion with high frequency. Furthermore, an image brightness change vector is determined. In other words, when the changes shown in Figure 5 are shown in a graph, as shown in Figure 7, by comparing the average value for each minute sample interval of 8 samples and the average value for 68 samples, it is possible to determine whether the luminance change is increasing or decreasing. You can know. In the example of FIG. 7, it can be seen that there is an increasing trend as indicated by the two-dot chain line.

In this way, the characteristics of the image are captured and the reference image is stored in the image memory 9b for calculation. That is, the average value of each pixel is stored as a standard value.

Furthermore, the maximum value of the change between a sample and the next sample is stored in the work memory 11b as an instantaneous luminance change tolerance range, and is used as one of the monitoring data.

Further, the maximum value and minimum value for 18 seconds are similarly stored in order to be used as a permissible unit time luminance change value. This process corresponds to the process of step F21.

Using the reference image obtained in this way, moving objects (particularly people) in the station configuration are monitored. The start of monitoring is started by pressing the start button 16b of the operation unit 16 (see FIG. 1) and transmitting a start signal to the arithmetic processor 10 via the input/output circuit. That is, in response to this start signal, the arithmetic processor 10 reads out the monitoring program stored in the memory 111, and
Start executing the programs one after another.

Next, the monitoring process started by pressing the push button 16b will be explained. In this embodiment, this process is executed according to the procedure shown in FIG. Next, this process will be explained in detail using Figure 11.

First, in step F30, it is checked whether a start signal has been input from start push button 1 (ib). When the start signal is input manually, the processes from step F32 onwards are executed one after another.

In the monitoring mode, step F32. As shown at F34, raw images from the television camera 4 are sent to the current image memory 9a via the image switching circuit 8 at regular intervals. In step F36, the arithmetic processor 10 compares new data with the arithmetic image memory 9b every time new data is written into the current image memory 9a. If the difference 8ab between the image memory 9a for the current image and the image memo 'J9b for calculation is larger than the instantaneous brightness change tolerance value, step F40
, examine the brightness change state of the adjacent pixels on the upper, lower, left and right sides, and check whether each adjacent pixel has a resolution of at least 5.
If the pixels exceed the allowable value, the process proceeds to step F42, where they are registered as abnormal state candidate surfaces. In the next sample, if an abnormality candidate surface appears in the same or nearby pixels as the abnormal state candidate surface, and an abnormality candidate surface appears in the vicinity of the previous abnormality candidate surface K times in a row, it is input/output as an abnormality. The alarm device 15 is turned on via the circuit element C. If this is the case,
Step F44. This is the process of F48.

Then, the surrounding 2-pixel image memory 9m surrounding the abnormality candidate surface
The white level value is written to the address , and the abnormality candidate surface is displayed on the monitor television 14 via the display control section 13 . This is the process of step F50.

In step F461, even if the instantaneous brightness change permissible value continues to be within the specified range, if it exceeds the unit time brightness change permissible value for 18 seconds, the process proceeds to step F48 and issues an alarm. In step F52, it is determined whether it is finished,
If the process has not ended, the process returns to step F32.

In the monitoring mode, the currently obtained difference data is stored at the address of the oldest content in the difference memory shown in Figure 5, and the ring pointer method is used, so that the current time is always T.
Since it uses values from N seconds ago, it has the characteristic that it is not affected by continuous changes in the surroundings.

Furthermore, after an alarm is issued, if the operator checks the condition and finds that it is normal, by pressing the registration button 16a, the value at the time the alarm was issued is newly adopted as the allowable value. . In this way.

Through learning, machines can be equipped with the ambiguity judgment mechanism that humans traditionally have. This is essential for optimizing the monitoring system, and is extremely effective in that it can easily and reliably improve reliability.

In addition, if a person falls onto the tracks, please use IS from the platform side.
Since there is a directional vector that moves toward the road, only when the abnormality candidate surface moves from the platform toward the railway, it is considered an abnormality and a warning is issued.

It may be possible to increase certainty.

That is, when the screen as shown in FIG. 8(a) is monitored, the abnormal state candidate surface 17 changes from candidate surface 17 m to li m ii 17 for each sampling as shown in FIG. 8(b).
When moving to c, its center of gravity coordinates are as shown in Figure 1fi9.

17a (Xa, Va) to 17'(XC# Yc)
Move to.

Here, pay attention to the component in the X direction, and only when Xa - Xc ) has a positive value, it is determined to be abnormal.

(However, the origin is set at the upper left of Figure 1.) In order to improve processing speed, only the area from the edge of the platform to the wall is monitored, and the rest is masked (as shown in Section 8 (c)). (hatched area in the figure).

In other words, in the image memory storage procedure shown in Figure 1M4, select addresses that do not need to be accessed in advance! Record it! , it is possible to perform processing without accessing during comparison, thereby improving speed.

This access-unnecessary data is obtained in advance from the image address and stored as a pattern in the ROM 11m! Record it.

〔Effect of the invention〕

According to one aspect of the present invention, as described in detail above.

Achieving a highly reliable method and device for monitoring moving objects that is unaffected by weather and lighting fluctuations.

[Brief explanation of drawings]

Fig. 1 is a diagram showing the hardware configuration of a station premises monitoring system that is an embodiment of the present invention, Fig. 2 is a diagram showing the arrangement of television cameras in the station premises according to an example of the present invention, and Fig. 3 is
Figure 34 is a diagram for explaining the concept of WJ division, Figure 34 is a diagram for explaining the concept of memory configuration, Figure aS is a diagram showing the state in which difference values are stored in memory, ll6g! l is a diagram showing an example of luminance distribution, I! Figure 7 shows examples of trends in difference changes, Figure 8C1), Figure 8(b), and! 8(c) is a diagram showing an example of a display on a monitoring monitor television, and 39all is a diagram showing the status of an abnormality candidate surface. FIG. 10 is a flowchart showing the standard image 51 recording process among the diagrams showing the operation flow in one embodiment of the present invention. FIG. 11 is a flow diagram showing the monitoring process among the diagrams showing the operation flow in the -wm example of the present invention. 4... Television (TV) camera, 6... Image input section, 7... Oscillator, 8... Switching circuit, 9
m, fib... image memory, W... arithmetic processor, U-. 11 b...Memory, 12...Input/output circuit, 13...Music display control section, 14...Monitor TV, IS..."" 1141111, 16...・Operation part ゛・、：＊・・ 1 fig. 3 H 770 figures // figure

Claims (1)

[Claims] 1. Digitize an image signal from a camera that images an area where a moving object may exist, store the digitized image signal in an image memory, and store the stored image signal. Calculate the temporal average value of the luminance of 1. A method for monitoring a moving object, comprising: detecting an intrusion of a moving object into the area. 2. A camera that images an area where a moving object may exist, a signal converter that digitizes the output signal of the camera, an image memory that stores the digitized image signal, and an image memory that stores the digitized image signal; A calculation function that calculates the temporal average value of the brightness of the image signal and calculates the difference between the current input image signal and the temporal average value, and the difference value is equal to or greater than a specified value and is equal to or greater than the specified value. 1. A moving object monitoring device comprising: a processing device having a determination function that determines that a moving object has entered the area when there is a spatial change in the area.