JP5396826B2 - Passing train imaging device by image processing - Google Patents

Description

  The present invention relates to an apparatus for photographing a state of a traveling train on a roof, and in particular, detects an approach of a train from an image of a monitoring camera installed above a track, and detects a train passing through the field of view of the monitoring camera. The present invention relates to a passing train photographing device by image processing for photographing a state on a roof.

  As a device for photographing the state on the roof of the train passing through the field of view of the surveillance camera, a camera and lighting are installed above the track, while a recording device is installed indoors, and recording is performed in conjunction with the vehicle approach warning device. The apparatus to be implemented is generally used. Such a device, for example, performs illumination lighting and recording start based on a train approach signal generated by a vehicle approach warning device when a train approaches, and records an image during a period when the train approach signal is generated. In the meantime, I photographed the roof of a train passing through the field of view of the surveillance camera. Several proposals have been made for a vehicle approach warning device that generates a train approach signal that serves as a reference for the start and end of imaging.

  Patent Document 1 discloses a technique in which a vibration sensor is directly attached to a rail using a magnet to detect a vibration when a railway vehicle approaches and generate a train approach signal. In Patent Document 2, a train current detection sensor is attached to a rail via a magnet, the output of the sensor is connected to an amplifier built in the arithmetic processing unit, and an alarm is sounded based on the result processed by the arithmetic processing unit. Such a technique is disclosed. Further, Patent Document 3 includes a transmission unit and a reception unit. The transmission unit transmits an exploration signal to the rail, the reflected signal reflected by the approaching track vehicle is received by the reception unit, and the reflected signal of the reception unit is received. A technique is disclosed that issues an alarm in response to reception.

JP 2001-39304 A JP 2002-337687 A JP-A-9-132141

  However, the devices described in Patent Documents 1 to 3 described above are devices for notifying a construction worker working in the vicinity of a track of the approach of a train. Therefore, these vehicle approach warning devices are configured to detect a train at a position away from the work site by a predetermined distance so that a sufficient time can be secured for the construction worker to retreat. Therefore, in the passing train imaging device, when starting and ending recording based on the train approach signal generated by the vehicle approach warning device, the train will be photographed before the train actually passes through the field of view of the camera, There is a time for taking a picture other than the time when the train passes through the field of view of the surveillance camera.

  In such a case, there will be unnecessary image data that has not been photographed for the train, many image storage areas of the recording device will be used, and the work to confirm the state of the passing train on the roof from the stored image When performing, it is necessary to search the image portion where the train is reflected from the image data, which may lead to a decrease in workability.

  In view of the above, an object of the present invention is to provide a passing train photographing apparatus using image processing that can efficiently store an image obtained by photographing a train passing through the field of view of a surveillance camera.

A passing train imaging device based on image processing according to the first invention for solving the above-mentioned problems includes a vehicle approach alarm means for detecting the approach of a vehicle and outputting a signal, and an imaging means for photographing the roof of the vehicle. And an image processing means for processing a train photographed by the photographing means, wherein the image processing means detects a vehicle from the image photographed by the photographing means. Control means for starting and stopping the photographing means based on a train approach signal transmitted from the vehicle approach warning means, and from the vehicle detection means. based on the presence or absence of the vehicle to be transmitted and control means for controlling the start and stop of the recording of the image by said recording means, said vehicle detection means The input image photographed by the photographing means is divided into a number of small areas, the correlation between the background image acquired in advance and the small area corresponding to the input image is calculated, and the temporal correlation value obtained by the correlation calculation is calculated. Detecting the small area where the change is larger than a predetermined value as a small change area, combining the adjacent small change areas to obtain a change detection partial area, calculating a geometric feature amount of the change detection partial area; the vehicle is characterized by entrance detection means der Rukoto for detecting and enters when the geometric feature value is within the feature amount range corresponding to the train set in advance.

According to the passing train photographing apparatus using the image processing according to the first aspect of the present invention, the image photographed by the photographing means is recorded in the recording means only while the vehicle is photographed in the image photographed by the photographing means. Therefore, the image storage area of the recording means can be used only for storing the image data of the vehicle being photographed, and the storage area of the recording means can be used efficiently. For this reason, when performing the operation | work which confirms the state on the roof of a passing train from a storage image, the operation | work which searches the image part in which the train is reflected can be abbreviate | omitted, and the highly efficient operation | work can be implement | achieved. In addition, since it is not necessary to set feature data in advance in order to detect an approaching train, the time for on-site adjustment can be shortened, and workability is improved.

  Hereinafter, the details of a passing train photographing apparatus using image processing according to the present invention will be described with reference to the drawings.

  A first embodiment of a passing train photographing apparatus using image processing according to the present invention will be described with reference to FIGS. FIG. 1 is an explanatory view showing an installation example of a passing train photographing apparatus by image processing according to the present embodiment, FIG. 2 is a block diagram showing a configuration of the passing train photographing apparatus by image processing according to the present embodiment, and FIG. FIG. 4 is an explanatory diagram showing an example of setting a detection area in the present embodiment, and FIG. 5 is a flowchart showing a process flow in the approaching vehicle detection unit of the present embodiment. .

  As shown in FIG. 1 and FIG. 2, in this embodiment, the passing train photographing apparatus using image processing includes a sensor 1 a for detecting the approach of the vehicle 6, illumination 2 for illuminating the roof of the vehicle 6, and the roof of the vehicle 6. And an image processing device 4 as an image processing unit connected to the sensor 1 a, the illumination 2 and the monitoring camera 3.

  The sensor 1a measures, for example, vibrations of the rail 5 and is attached to the rail 5 and connected to the vehicle approach warning unit 1 as vehicle approach warning means. The vehicle approach warning unit 1 detects whether or not the vehicle 6 is approaching according to the output of the sensor 1a. When it is determined that the vehicle 6 is approaching, the train approach signal is sent as an alarm to the image processing device 4. Send to.

  The illumination 2 is installed ahead of the vehicle 6 in the approach direction with respect to the sensor 1a, and is configured to be turned on and off based on a signal from the image processing device 4. The surveillance camera 3 is installed at approximately the same position as the illumination 2 so as to capture an area illuminated by the illumination 2 on the roof of the vehicle 6, and the start and end of the capture are based on a signal from the image processing device 4. Configured to do. An image captured by the monitoring camera 3 is transmitted to the image processing device 4.

  The position where the illumination 2 and the monitoring camera 3 are installed is a position separated from the sensor 1a by a predetermined distance. Specifically, when the monitoring camera 3 is activated based on the output of the sensor 1a, the vehicle 6 It is set as a position where photographing by the monitoring camera 3 can be started before entering the visual field of the monitoring camera 3.

  The image processing apparatus 4 includes a process setting unit 41, a control unit 42 as a control unit, an approach vehicle detection unit 43 as an approach vehicle detection unit, and a recording unit 44 as a recording unit shown in FIG. Yes.

  The process setting unit 41 sets a detection area A at a predetermined position as shown in an example in FIG. 4 with respect to an image in which the vehicle 6 acquired from the approaching vehicle detection unit 43 is not photographed. A difference detection threshold value for determining whether or not the vehicle 6 has entered is set, and the setting data is stored as processing data.

  The control unit 42 receives the train approach signal transmitted from the vehicle approach warning unit 1 and transmits an activation signal to the illumination 2, the monitoring camera 3, the approaching vehicle detection unit 43, and the recording unit 44. Further, it receives a train detection signal from the approaching vehicle detection unit 43 and transmits a recording start signal to the recording unit 44. Further, upon receiving a train disappearance signal from the approaching vehicle detection unit 43, a recording stop signal is transmitted to the recording unit 44, and a stop signal is transmitted to the illumination 2, the monitoring camera 3, the approaching vehicle detection unit 43, and the recording unit 44. .

  The approaching vehicle detection unit 43 receives an activation signal from the control unit 42 and starts an approaching vehicle detection process, reads processing data from the process setting unit 41, and receives a stop signal from the control unit 42 to receive an approaching vehicle detection process. To stop. Further, during the approaching vehicle detection process, the presence or absence of entry of the vehicle 6 is detected by analyzing an image transmitted from the monitoring camera 3 (hereinafter referred to as an input image), and the vehicle 6 is detected in the input image. The train detection signal is transmitted to the control unit 42. Furthermore, when the vehicle 6 is no longer detected in the input image after detecting the entry of the vehicle 6, a train disappearance signal is transmitted to the control unit 42.

  More specifically, the approaching vehicle detection unit 43 is provided with an image input processing unit 43a, a reference image storage processing unit 43b, an absolute value difference calculation processing unit 43c, and a train approach determination processing unit 43d as shown in FIG. Yes.

  The image input processing unit 43a performs a process of inputting an image transmitted from the surveillance camera 3, and the reference image storage processing unit 43b is based on the detection area A shown in FIG. Process to store. Further, the absolute value difference calculation processing unit 43c performs an absolute value difference calculation between the detection area A of the input image input to the image input processing unit 43a and the reference image stored in the reference image storage processing unit 43b. The train entry determination processing unit 43d performs a process of determining whether a train has entered based on the result calculated by the absolute value difference calculation processing unit 43c and outputting a detection result.

  And in such an approach vehicle detection part 43, an approach vehicle detection process as shown in FIG. 5 is performed. That is, when the approaching vehicle detection unit 43 receives an activation signal from the control unit 42, the image input processing unit 43a first inputs an image transmitted from the monitoring camera 3 (step PA1).

  Subsequently, it is determined whether or not it is immediately after activation (step PA2), and if it is determined that it is immediately after activation (Yes), it is considered that the vehicle 6 has not been photographed in the input image, and processing data is processed from the processing setting unit 41. The preset detection area A and difference detection threshold are input (step PA3), and then the image in the detection area A of the input image is stored as a reference image in the reference image storage processing unit 43b (step PA4).

  On the other hand, when it is determined in step PA2 that it is not immediately after activation (No), the absolute value difference calculation processing unit 43c performs the absolute value difference calculation between the detection area A of the input image and the reference image (step PA5).

  Thereafter, when the absolute value difference calculation result is larger than the difference detection threshold value input from the process setting unit 41 in step PA3 in the train approach determination processing unit 43d, the vehicle 6 is captured in the input image, that is, the vehicle 6 Is determined to have entered (step PA6). In step PA6, when the absolute value difference calculation result is equal to or smaller than the difference detection threshold, the vehicle 6 is not photographed in the input image, that is, the vehicle 6 has not entered or the vehicle 6 has passed. Judge that there is.

  Subsequently, the presence / absence of the vehicle 6 determined in step PA6 is output to the control unit 42 as a determination result (step PA7). Specifically, while it is determined that the vehicle 6 is captured in the input image, a train detection signal is output to the control unit 42, and after the vehicle 6 is detected, the vehicle 6 is not captured in the input image. If determined, a train disappearance signal is output to the control unit 42. This process is repeated until a stop signal is received from the control unit 42.

  Through such processing, the approaching vehicle detection unit 43 detects whether the vehicle 6 has entered.

  The recording unit 44 receives a start signal from the control unit 42 to start the recording function, and receives a stop signal from the control unit 42 to stop the recording function. Also, recording of an input image is started in response to a recording start signal from the control unit 42, and recording is stopped in response to a recording stop signal from the control unit 42.

  According to such a passing train photographing apparatus using image processing according to the present embodiment, the input image can be recorded only when the vehicle is passing through the field of view of the surveillance camera, so that the vehicle 6 is photographed. The image storage area of the recording unit 44 can be used only for storing image data. Thereby, when performing the operation of confirming the state of the vehicle 6 on the roof from the stored image, the operation of searching for the image portion where the vehicle 6 is photographed can be omitted, and the efficiency of the operation can be realized. .

  Based on FIG.6 and FIG.7, the 2nd Example of the passing train imaging device by the image processing which concerns on this invention is described. FIG. 6 is a block diagram showing the configuration of the approaching vehicle detection unit of this embodiment, and FIG. 7 is a flowchart showing the flow of processing in the approaching vehicle detection unit of this embodiment.

  The present embodiment is different from the first embodiment described above in the processing in the processing setting unit 41 and the approaching vehicle detection unit 43. Other configurations are substantially the same as the configurations shown in FIGS. 1 to 5 and described above, and the description overlapping with the description in the first embodiment will be omitted, and different portions will be mainly described.

  In the present embodiment, the process setting unit 41 sets the detection area A for the input image in which the vehicle 6 acquired from the approaching vehicle detection unit 43 is not photographed, and sets the difference detection threshold set in the first embodiment. Instead, a correlation detection threshold value for determining whether or not the vehicle 6 has entered the detection area A is set, and these data are stored as processing data.

  Further, the approaching vehicle detection unit 43 is different from the first embodiment in the process of detecting whether or not the vehicle 6 has entered. Specifically, as shown in FIG. 6, in this embodiment, the approaching vehicle detection unit 43 includes an image input processing unit 43a, a reference image storage processing unit 43b, a normalized correlation calculation processing unit 43e, and a train approach determination processing unit 43d. Is provided.

  The image input processing unit 43a performs a process of inputting an image transmitted from the surveillance camera 3, and the reference image storage processing unit 43b is based on the detection area A shown in FIG. Process to store. The normalized correlation calculation processing unit 43e performs a normalized correlation calculation between the image input to the image input processing unit 43a and the reference image stored in the reference image storage processing unit 43b. The train entry determination processing unit 43d performs a process of determining whether or not the vehicle 6 has entered based on the result calculated by the normalized first publication calculation processing unit 43e and outputting a detection result.

  Hereinafter, the flow of the train detection process by this approach vehicle detection part 43 is demonstrated using FIG. As shown in FIG. 7, when the approaching vehicle detection unit 43 receives an activation signal from the control unit 42, first, the image input processing unit 43a inputs an image transmitted from the monitoring camera 3 (step PB1).

  Subsequently, it is determined whether or not it is immediately after activation (step PB2), and if it is determined that it is immediately after activation (Yes), it is considered that the vehicle 6 is before entering the field of view of the monitoring camera 3, and the process setting unit 41, the detection area A and the correlation detection threshold value set in advance as processing data are input (step PB3), and then the image in the detection area A of the input image is stored as a reference image in the reference image storage processing unit 43b. (Step PB4).

  On the other hand, if it is determined in step PB2 that it is not immediately after activation (No), the normalized correlation calculation processing unit 43e performs a normalized correlation calculation between the detection area A of the input image and the reference image (step PB5).

  Thereafter, in the train approach determination processing unit 43d, when the normalized correlation calculation result is larger than the correlation detection threshold value input from the processing setting unit 41 in step PA3, the vehicle 6 is captured in the input image. 6 is entered (step PB6). In step PA6, when the normalized correlation calculation result is equal to or smaller than the correlation detection threshold, the vehicle 6 is not photographed in the input image, that is, the vehicle 6 has not entered or the vehicle 6 has passed. Judge that there is.

  Subsequently, the presence / absence of the vehicle 6 determined in step PA6 is output to the control unit 42 as a determination result (step PB7). Specifically, while it is determined that the vehicle 6 is captured in the input image, a train detection signal is output to the control unit 42. After the train is detected, it is determined that the vehicle 6 is not captured in the input image. Then, a train disappearance signal is output to the control unit 42. This process is repeated until a stop signal is received from the control unit 42.

  With the above processing, the approaching vehicle detection unit 43 detects whether or not the vehicle 6 has entered.

  According to such a passing train photographing apparatus using image processing according to the present embodiment, in addition to the effects of the first embodiment, the comparison between the reference image and the input image is determined by the normalized correlation calculation, and the presence / absence of the train approach is determined. Therefore, the calculation is more complicated than in the first embodiment, but it is robust to uniform illuminance changes over the entire detection area, and is sensitive even when a wide range of illuminance changes that are shaded during the day occurs. The presence / absence of the vehicle 6 can be detected stably.

  A third embodiment of the passing train photographing apparatus using image processing according to the present invention will be described with reference to FIGS. FIG. 8 is a block diagram showing the structure of a passing train photographing apparatus using image processing according to this embodiment, FIG. 9 is a block diagram showing the configuration of the object shape inspection unit of this embodiment, and FIG. 10 shows an example of an object shape model. FIG. 11 is a flowchart showing the flow of object shape model creation in the process setting unit of this embodiment, and FIG. 12 is a flowchart showing the flow of processing in the object shape inspection unit of this embodiment.

  In the present embodiment, the configuration of the image processing apparatus 4 is different from that of the first embodiment. Specifically, the image processing apparatus 4 includes an object shape inspection unit 45 as an object shape inspection unit illustrated in FIG. 8 instead of the approaching vehicle detection unit 43 described in the first embodiment, and includes a process setting unit 41 and The processing in the control unit 42 is different from that in the first embodiment.

  Other configurations are substantially the same as the structure described in the first embodiment. Hereinafter, the same members as those described in the first embodiment are denoted by the same reference numerals, and redundant description is omitted. The difference will be mainly described.

  The passing train imaging device by image processing according to the present embodiment is configured to be able to set two processing modes of a shape model production mode and a train approach detection mode as device modes, and performs different processing depending on the device mode. The image processing apparatus 4 includes a process setting unit 41, a control unit 42, an object shape inspection unit 45, and a recording unit 44 shown in FIG. In the shape model production mode, a process for producing the object shape model M (see FIG. 10) of the track portion is performed, and in the train approach detection mode, a process for detecting whether or not the vehicle 6 has entered is performed.

  In the image processing apparatus 4 of the present embodiment, the process setting unit 41 sets “shape creation mode” or “train entry detection mode” as the apparatus mode. In addition, a detection area A is set at a predetermined position as shown in FIG. 4 for an input image in which the vehicle 6 acquired from the object shape inspection unit 45 is not photographed. Also, setting processing for obtaining an object shape model production image, setting feature extraction parameters, creating an object shape model M, and setting object shape inspection parameters is performed. These setting data are stored as processing data.

  The control unit 42 inputs the device mode from the process setting unit 41. When the device mode is the “shape model production mode”, the control unit 42 sends an activation signal to the illumination 2, the monitoring camera 3, the object shape inspection unit 45, and the recording unit 44. Send. When the apparatus mode is changed from the “shape model production mode” to the “train entry detection mode”, a stop signal is transmitted to the illumination 2, the monitoring camera 3, the object shape inspection unit 45, and the recording unit 44. In response to the train approach signal transmitted from the vehicle approach warning unit 1, an activation signal is transmitted to the illumination 2, the monitoring camera 3, the object shape inspection unit 45, and the recording unit 44. Further, it receives a train detection signal from the object shape inspection unit 45 and transmits a recording start to the recording unit 44. In addition, upon receiving a train disappearance signal from the object shape inspection unit 45, a recording stop signal is transmitted to the recording unit 44, and a stop signal is transmitted to the illumination 2, the monitoring camera 3, the object shape inspection unit 45, and the recording unit 44. To do.

  The object shape inspection unit 45 receives an activation signal from the control unit 42, inputs a device mode from the processing setting unit 41, starts a shape model production process or a train approach detection process according to the device mode, and the control unit 42 The shape model preparation process or the train approach detection process is stopped in response to the stop signal from.

  Hereinafter, the shape model preparation process and the train approach detection process will be described in more detail. As shown in FIG. 9, the object shape inspection unit 45 includes an image input processing unit 45a, a feature extraction processing unit 45b, an object shape model setting processing unit 45c, a feature data matching processing unit 45d, and a train approach determination processing unit 45e. It has been.

  The image input processing unit 45a performs a process of inputting an image transmitted from the monitoring camera 3, and the feature extraction processing unit 45b performs a process of extracting a feature in the input image acquired by the image input processing unit 45a. The object shape model setting processing unit 45c performs processing for setting an object shape model M shown in FIG. The feature data matching processing unit 45d matches the feature data of the object shape model M and the feature data of the input image input from the monitoring camera 3. The train entry determination processing unit 43d performs a process of determining whether or not the vehicle 6 has entered based on the result of collating the feature data by the feature data collating processing unit 45d and outputting the detection result.

  In such an object shape inspection unit 45, when the shape model production mode is transmitted from the process setting unit 41 as the apparatus mode, the process shown in FIG. 11 is performed.

  That is, when the object shape inspection unit 45 receives the activation signal from the control unit 42, the object shape model setting processing unit 45c first generates an object shape model from the processing setting unit 41 and from the input image in the processing setting unit 41. An object shape model preparation image selected in advance for use is acquired (step PC1).

  Subsequently, the object shape model creation image acquired in step PC1 is analyzed, and characteristic portions such as the contour and pattern of the object existing in the object shape model creation image are extracted by approximating a straight line or an arc. Extracted as feature data consisting of straight line features and arc features (step PC2).

  Subsequently, a straight line feature and an arc feature of the track portion where the vehicle 6 does not particularly exist are selected from the feature data (step PC3), and are registered as feature data of the object shape model M (step PC4).

  With this process, the object shape model M is produced in the object shape inspection unit 45.

  Moreover, when the train approach detection mode is transmitted from the process setting unit 41 as the apparatus mode, the process illustrated in FIG. 12 is performed.

  That is, when the object shape inspection unit 45 receives the activation signal from the control unit 42, it first determines whether or not it is immediately after activation (step PD1), and if it is determined that it is immediately after activation (Yes), it sets the processing. The detection area A in the image, the feature extraction parameter, the object shape model M, and the object shape inspection parameter set in advance as processing data are input from the unit 41 (step PD2). When it is determined in step PD1 that it is not immediately after activation (No) or following step PD2, an image transmitted from the monitoring camera 3 is input in the image input processing unit 45a (step PD3).

  In the feature extraction processing unit 45b, an input image is analyzed, and a linear feature and an arc feature extracted by approximating a characteristic portion such as a contour or a pattern of an object existing in the detection area A in the image to a straight line or an arc. (Step PD4), and the feature data registered in the object shape model M input from the process setting unit 41 in the feature data matching processing unit 45d is matched with the feature data extracted from the input image. (Step PD5).

  Subsequently, in the train approach determination processing unit 45e, when the feature data collated in the feature data collation processing unit 45d are similar within the range of the object shape inspection parameter set in advance in the process setting unit 41, two features are used. It is determined that the data is suitable and the vehicle 6 is photographed in the input image, that is, it is determined that the vehicle 6 has entered (step PD6). If the feature data collated by the feature data matching processing unit 45d are not similar within the range of the object shape inspection parameter set in advance by the processing setting unit 41, the two feature data are regarded as incompatible and the input image It is determined that the vehicle 6 has not been photographed, that is, the vehicle 6 has not entered or the vehicle 6 has passed.

  Subsequently, the presence / absence of the vehicle 6 determined in step PD6 is output to the control unit 42 as a determination result (step PD7). Specifically, while it is determined that the vehicle 6 is captured in the input image, a train detection signal is output to the control unit 42. After the train is detected, it is determined that the vehicle 6 is not captured in the input image. When it is done, a train disappearance signal is output to the control unit 42. This process is repeated until a stop signal is received from the control unit 42.

  By such processing, the train shape detection unit 45 performs train approach detection processing.

  According to such a passing train photographing apparatus using image processing according to the present embodiment, since the input image is recorded only while the vehicle 6 is photographed within the field of view of the monitoring camera 3, the image where the vehicle 6 is photographed. The image storage area of the recording unit 44 can be used only for storing data. Therefore, when performing the operation of checking the state of the vehicle 6 on the roof from the stored image, the operation of searching for the image portion where the vehicle 6 is photographed can be omitted, and the efficiency of the operation can be realized.

  In addition, since the determination is based on the features extracted from the input image, it is more robust to changes in illuminance than the methods of comparing the images themselves as in the first and second embodiments. In addition, stable train entry detection can be performed even when there is uneven night lighting.

  A fourth embodiment of the passing train photographing apparatus using image processing according to the present invention will be described with reference to FIGS.

  FIG. 13 is a block diagram showing the structure of a passing train photographing apparatus using image processing according to the present embodiment, FIG. 14 is a block diagram showing the configuration of the entry detection unit of this example, and FIG. 15 is by the entry detection unit of this example. FIG. 16 is an explanatory diagram illustrating an example of a processing result in the approach detection unit of the present embodiment.

  This embodiment is different from the first embodiment described above in the configuration of the image processing apparatus 4. Specifically, in place of the approach vehicle detection unit 43 of the image processing apparatus 4 in the first embodiment, an approach detection unit 46 as an approach detection unit shown in FIG. 13 is provided, and the process setting unit 41 and The processing in the control unit 42 is different. Other configurations are substantially the same as the structure described in the first embodiment. Hereinafter, the same members as those described in the first embodiment are denoted by the same reference numerals, and redundant description is omitted. The difference will be mainly described.

  The image processing apparatus 4 according to this embodiment includes a process setting unit 41, a control unit 42, an entry detection unit 46, and a recording unit 44 shown in FIG.

  The processing setting unit 41 sets a detection area A at a predetermined position as shown in FIG. 4 for an input image in which the vehicle 6 acquired from the approach detection unit 46 is not photographed. A range of geometric feature amounts for determining whether or not the vehicle 6 has entered is set, and these setting data are stored as processing data.

  The control unit 42 receives a train approach signal from the vehicle approach warning unit 1 and transmits an activation signal to the illumination 2, the monitoring camera 3, the approach detection unit 46, and the recording unit 44. Further, it receives a train detection signal from the approach detection unit 46 and transmits a recording start signal to the recording unit 44. In addition, a train stop signal is transmitted to the recording unit 44 in response to a train disappearance signal from the approach detection unit 46, and a stop signal is transmitted to the illumination 2, the monitoring camera 2, the approach detection unit 46, and the recording unit 44.

  The entry detection unit 46 starts the entry detection process in response to the activation signal from the control unit 42, reads the process data from the process setting unit 41, and stops the entry detection process in response to the stop signal from the control unit 42. . In addition, during the approach detection process, the input image is analyzed to detect the approach of the vehicle 6. When the vehicle 6 is detected in the input image, a train detection signal is transmitted to the control unit 42. Further, when the vehicle 6 is no longer detected in the input image after detecting the entry of the vehicle 6, a train disappearance signal is transmitted to the control unit 42.

  More specifically, as shown in FIG. 14, the approach detection unit 46 includes an image input processing unit 46a, a reference image storage processing unit 46b, a change detection partial region detection processing unit 46c, and a geometric feature amount inspection processing unit 46d. Is provided.

  The image input processing unit 46a performs a process of inputting an image transmitted from the surveillance camera 3, and the reference image storage processing unit 46b is based on the detection area A shown in FIG. Process to store. The change detection partial area detection processing unit 46c performs correlation calculation between the input image input to the image input processing unit 46a and the reference image stored in the reference image storage processing unit 46b. The geometric feature amount inspection processing unit 46d determines whether the vehicle 6 has entered based on the result calculated by the change detection partial region detection processing unit 46c, and outputs a detection result.

  In this approach detection part 46, an approach detection process is performed in the flow shown below. That is, in the approach detection unit 46, the reference image storage processing unit 46b acquires in advance an image (see FIG. 4) in which the vehicle 6 is not photographed, and stores this as reference image data.

  When the activation signal is received from the control unit 42, the detection area A and the geometric feature range in the input image are input as processing data from the processing setting unit 41, and from the monitoring camera 3 in the image input processing unit 46a. Enter the image to be sent. Subsequently, the change detection partial area detection processing unit 46c divides the reference image and the input image 7 into appropriate small areas 71 (see FIG. 15A).

  Thereafter, in the detection area A that is set in advance in the portion of the image where the vehicle 6 passes, the correlation calculation between the reference image data corresponding to each small area 71 and the input image data is performed, and the correlation value From the small area 71 set in the input image based on the temporal change, the small change area 72 in which the occurrence of the change is detected and the background candidate small area 73 with little change from the background are detected (FIG. 15). (See (b)), and then, the adjacent change small areas 72 are combined to obtain a change detection partial area 74 (see FIG. 15C).

  Subsequently, in the geometric feature amount inspection processing unit 46d, it is determined whether or not the geometric feature amount of the change detection partial area 74 is within the range of the geometric feature amount corresponding to the vehicle 6 set in advance. When the geometric feature amount of the change detection partial area 74 is within the range of the geometric feature amount corresponding to the vehicle 6 set in advance, the vehicle 6 is captured in the input image, that is, the vehicle It is determined that 6 has entered. If the geometric feature amount of the change detection partial area 74 is not within the predetermined range of the geometric feature amount corresponding to the vehicle 6 set in advance, it is assumed that the two feature data do not match and the vehicle is not included in the input image. 6 is not photographed, that is, it is determined that the vehicle 6 has not entered or that the vehicle 6 has passed.

  Subsequently, the presence / absence of entry of the vehicle 6 is output to the control unit 42. Specifically, while it is determined that the vehicle 6 is captured in the input image, a train detection signal is output to the control unit 42, and after the vehicle 6 is detected, the vehicle 6 is not captured in the input image. If determined, a train disappearance signal is output to the control unit 42. This process is repeated until a stop signal is received from the control unit 42.

  In such an approach detection process, even if a partial illuminance change as shown in FIG. 16 occurs and the areas 81 and 82 in FIG. 16 are extracted as change detection partial areas, such a change is observed in various places. Since it is divided, it is distinguished from the vehicle 6 by the determination based on the geometric feature amount corresponding to the vehicle 6 set in advance in the geometric feature amount inspection processing unit 46d.

  According to such a passing train photographing apparatus using image processing according to the present embodiment, in addition to the effects of the first embodiment and the second embodiment, it is robust against the illuminance change, and the dawn or partial deviation in which the partial illuminance change occurs. Even when night illumination is present, it is possible to perform stable train entry detection, and since it is not necessary to set the object shape model M compared to Example 3, even if there are many shooting points, It is not necessary to create the object shape model M at each point, and the field adjustment work can be greatly reduced.

  In addition, this invention is not limited to the Example mentioned above, It cannot be overemphasized that a various change is possible in the range which does not deviate from the meaning of this invention.

  The present invention is suitable for application to a passing train photographing apparatus using image processing.

It is explanatory drawing which shows the example of installation of the passing train imaging device by the image processing which concerns on this invention. It is a block diagram which shows the structure of the passing train imaging device by the image processing which concerns on Example 1 of this invention. It is a block diagram which shows the structure of the approaching vehicle detection part of the passing train imaging device by the image processing which concerns on Example 1 of this invention. It is explanatory drawing which shows the example of a setting of the detection area in Example 1 of this invention. It is a flowchart which shows the flow of the approach vehicle detection process in the passing train imaging device by the image process which concerns on Example 1 of this invention. It is a block diagram which shows the structure of the approach vehicle detection part of the passing train imaging device by the image processing which concerns on Example 2 of this invention. It is a flowchart which shows the flow of the approach vehicle detection process in the passing train imaging device by the image processing which concerns on Example 2 of this invention. It is a block diagram which shows the structure of the passing train imaging device by the image processing which concerns on Example 3 of this invention. It is a block diagram which shows the structure of the object shape test | inspection part of the passing train imaging device by the image processing which concerns on Example 3 of this invention. It is explanatory drawing which shows the example of the object shape model in Example 3 of this invention. It is a flowchart which shows the flow of the object shape model preparation process in the passing train imaging device by the image processing which concerns on Example 3 of this invention. It is a flowchart which shows the flow of the train approach detection process in the passing train imaging device by the image processing which concerns on Example 3 of this invention. It is a block diagram which shows the structure of the passing train imaging device by the image processing which concerns on Example 4 of this invention. It is a block diagram which shows the structure of the approach detection part of the passing train imaging device by the image processing which concerns on Example 4 of this invention. FIG. 15A is an explanatory diagram showing an example of dividing the reference image into small regions, FIG. 15B is an explanatory diagram showing an example of setting the change small region and the background candidate small region, and FIG. 15C. FIG. 7 is an explanatory diagram showing an example of setting a change detection partial area as a change small area; It is another explanatory drawing which shows the example of a setting of a change detection partial area | region.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Vehicle approach alarm part 1a Sensor 2 Illumination 3 Monitoring camera 4 Image processing apparatus 41 Process setting part 42 Control part 43 Approaching vehicle detection part 43a Image input process part 43b Reference | standard image storage process part 43c Absolute value difference calculation process part 43d Train entry judgment Processing unit 44 Recording unit 45 Object shape inspection unit 45a Image input processing unit 45b Feature extraction processing unit 45c Object shape model setting processing unit 45d Feature data matching processing unit 45e Train approach determination processing unit 46 Entrance detection unit 46a Image input processing unit 46b Reference Image storage processing unit 46c Change detection partial region detection processing unit 46d Geometric feature amount inspection processing unit 5 Rail 6 Vehicle 7 Input image 71 Small region 72 Small change region 73 Background candidate small region 74 Change detection partial region A Detection area M Object Shape model

Claims (1)

By image processing comprising vehicle approach alarm means for detecting approach of a vehicle and outputting a signal, photographing means for photographing the roof of the vehicle, and image processing means for processing an image photographed by the photographing means In the passing train imaging device,
The image processing means
Vehicle detection means for detecting a vehicle from an image photographed by the photographing means;
Recording means for recording an image photographed by the photographing means;
Based on the train approach signal transmitted from the vehicle approach warning means, the start and stop of the photographing means is controlled, and the recording of the image by the recording means is performed based on the presence or absence of the vehicle transmitted from the vehicle detection means. and control means for controlling the start and stop,
The vehicle detection means divides the input image photographed by the photographing means into a number of small areas, performs correlation calculation between the background image acquired in advance and the small area corresponding to the input image, and performs the correlation calculation. The small area where the temporal change of the correlation value is larger than a predetermined value is detected as a small change area, the adjacent small change areas are combined to obtain a change detection partial area, and the geometric shape of the change detection partial area image processing, characterized in that to calculate the feature amount, Ru entry detecting means der for the vehicle when the geometric feature value is within the feature amount range corresponding to the train set in advance is detected to have entered By passing train photographing device.

Images