JP5653093B2 - Railway vehicle door safety confirmation device - Google Patents

Description

  The present invention relates to a railway vehicle door safety confirmation apparatus, and more particularly to a railway vehicle door safety confirmation apparatus that confirms safety from the three-dimensional shape of an object sandwiched when the railway vehicle door is closed.

  2. Description of the Related Art Conventionally, a technique for detecting that an object has been caught between railcar doors has been widely put into practical use. When a state where an object is caught in a door is detected on the train, the driver is notified and the train is operated safely. When an object is caught in the door of the train entrance / exit, it is detected by a sensor or the like that the door is not completely closed. The elevator apparatus also has a mechanism in which when an object is caught in the door, it is detected by a sensor and the door does not close.

  In recent years, in order to increase the safety of trains and passengers getting on and off trains at homes at railway stations, installation of home door devices mainly having double doors for the left and right doors is also proceeding. For example, Japanese Patent Application Laid-Open No. 2010-7456 discloses a technique for avoiding a door pinching state by detecting the so-called door pinching state with an electrical sensor, for example, an overcurrent detection sensor, even in a home door device. Yes.

JP 2010-7456 A (paragraph number 0003)

  However, the detection of the conventional door pinching state uses an electric sensor to detect that an object is being pinched, so that the pinched object does not interfere with train operation. However, there was a problem that it was detected as a door pinched state. On the other hand, even if the sandwiched object has a hindrance to the train operation, it may not be detected by an electrical sensor. SUMMARY OF THE INVENTION An object of the present invention is to provide a railroad vehicle door safety confirmation device that can recognize an object sandwiched by image processing and confirm safety when the door is closed.

According to one aspect of the present invention, a distance that is generated based on at least two image signals obtained by imaging with at least two or more image sensors and each pixel is a value indicating a distance or a position in a three-dimensional space. The distance image generation unit that generates an image, the distance image generated by the distance image generation unit, and the distance image of a predetermined background image including a door of a railcar stored in advance are generated by the comparison. A background image separating unit that separates the predetermined background image portion from the distance image , and a distance image that is left after the predetermined background image portion is separated by the background image separation unit. a distance image extracting unit extracts a range image of the object, from the time series changes in the cross-sectional shape of the largest distance value of the object distance image extracting unit distance in the image of the extracted the product in a predetermined And determining the deformation degree determining section deformation degree of the product during the time, the volume of the said product range image extracting unit the product obtained from the distance image of the extracted the product in, shape classification, and the degree of deformation and parameter information including a parameter of the degree of deformation which is determined in the determination section, the volume, the shape classification and, based on the based have been defined rule information including the deformation degree, the safety confirmation regarding said door It is possible to provide a safety confirmation device for a railway vehicle door having a safety confirmation determination unit that performs the determination.

  According to the present invention, it is possible to recognize an object sandwiched by image processing and confirm safety when the door of the railway vehicle is closed.

It is a figure for demonstrating the monitoring system 1 which image | photographs the platform of a station with the stereo camera based on embodiment of this invention. 1 is an external view of a stereo camera according to an embodiment of the present invention. It is a block diagram which shows the structure of the stereo camera SC which concerns on embodiment of this invention. It is a figure for demonstrating the example of the distance image obtained by stereo camera SC which concerns on embodiment of this invention. It is a flowchart which shows the content of the process in the recognition process part 34 of each stereo camera SC based on embodiment of this invention. It is a flowchart which shows the flow of the volume calculation process which concerns on embodiment of this invention. It is a figure for demonstrating the volume calculation which concerns on embodiment of this invention. It is a flowchart which shows the flow of the classification process of the shape category which concerns on embodiment of this invention. It is a figure which shows the example of the classification | category of the predetermined shape category which concerns on embodiment of this invention. It is a flowchart which shows the flow of the deformation | transformation degree determination process which concerns on embodiment of this invention. It is a figure for demonstrating the deformation | transformation of the thing based on embodiment of this invention.

Embodiments of the present invention will be described below with reference to the drawings. Here, an apparatus for confirming safety when a door of a railway vehicle is closed using a stereo camera installed on the platform end of a railway station will be described as an example.
FIG. 1 is a diagram for explaining a monitoring system 1 that photographs a station platform with a stereo camera. FIG. 1 is a view of the platform viewed from the direction of travel of the train. In FIG. 1, only one stereo camera SC is shown, but a plurality of stereo cameras are installed on the ceiling along the edge of the long home H on the ground G where the railroad tracks are laid. The whole is monitored from the vertical direction. Each stereo camera is installed at a position where an object captured by the train door can be imaged when the train T stops.

  The stereo camera SC, which is a monitoring camera that monitors the edge of the home H, has two image sensors, and the optical system of each image sensor is arranged so that the optical axes are parallel to each other with a predetermined interval. Has been placed. As shown in FIG. 1, the stereo camera SC is fixedly installed so that the imaging direction thereof is a direction orthogonal to the floor surface at the end of the platform H on the track side.

  The surveillance system 1 includes a plurality of stereo cameras SC (only one stereo camera is shown in FIG. 1), a personal computer (hereinafter abbreviated as PC) 11 connected to the stereo camera SC and a communication cable CB, a station, Or it is comprised including the control apparatus 12 of a train.

For example, the PC 11 and the control device 12 are arranged in a monitoring room of a station, and a monitor such as a station staff displays a moving image acquired by each stereo camera SC on a monitor connected to the PC 11 or the control device 12. You can see the station platform.
The PC 11 communicates with each stereo camera SC, instructs each stereo camera SC to execute safety confirmation processing, receives a detection signal from each stereo camera SC, and outputs it to the control device 12.

  The control device 12 is a device that receives an instruction from a station staff or the like and transmits the instruction to the PC 11. In the following example, when a train door is closed, a safety confirmation instruction signal instructing execution of safety confirmation is output to the PC 11. When receiving the safety confirmation instruction signal, the PC 11 outputs a predetermined command instructing each stereo camera SC to execute the safety confirmation process. Note that the PC 11 may transfer the safety confirmation instruction signal to each stereo camera SC as it is to execute the safety confirmation process.

  When the stereo camera SC receives the command, the stereo camera SC executes safety confirmation processing, determines whether an object having a predetermined three-dimensional shape is sandwiched between doors by image processing, and outputs the determination result to the PC 11 as a detection signal. When the control device 12 receives the detection signal, it notifies the monitor that a predetermined three-dimensional object is or is not sandwiched between the doors by monitor display, audio output, or the like.

(Configuration of stereo camera SC)
FIG. 2 is an external view of the stereo camera. On one side surface of the case 21 of the stereo camera SC, for example, two objective optical systems 22 are provided so that their optical axes are parallel to each other by a predetermined distance. The two objective optical systems 22 are objective lenses, and an imaging element such as a CMOS sensor is provided behind each objective lens in the housing 21. The stereo camera SC is installed on the platform H of the station so that the surface on which the two objective optical systems 22 are provided faces the floor of the platform H.

  Here, each stereo camera SC will be described as an example having two image sensors for stereo imaging, but may have three or more image sensors for stereo imaging, It suffices to have at least two image sensors.

Further, a communication cable CB for connecting to the PC 11 is connected to the stereo camera SC.
FIG. 3 is a block diagram showing the configuration of the stereo camera SC. The stereo camera SC has two CMOS sensors 31, two image capturing units 32, a stereo processing unit 33, a recognition processing unit 34, and a communication unit 35 as cameras, and is mounted on a substrate (not shown) in the housing 21. Is provided. In the present embodiment, the two image capturing units 32, the stereo processing unit 33, and the recognition processing unit 34 are configured by a DSP.

  Each CMOS sensor 31 receives light from the objective optical system 22 and outputs an imaging signal to the corresponding image capturing unit 32.

  Each image capturing unit 32 is a processing unit that captures an imaging signal from the CMOS sensor 31, generates an image signal from the imaging element, and supplies the image signal to the stereo processing unit 33.

  The stereo processing unit 33 is a processing unit that generates a distance image as information regarding a three-dimensional shape based on the input image signals of two moving images. The distance image is image information in which each pixel value is a distance value. The distance image may be information indicating the position of the pixel value in the three-dimensional space. That is, the stereo processing unit 33 configures a three-dimensional shape information generation unit that generates information regarding a three-dimensional shape based on two image signals obtained by imaging with two imaging elements.

  The recognition processing unit 34 recognizes the three-dimensional shape based on the information of the distance image obtained by the stereo processing unit 33, and indicates that the safety when the door is closed is confirmed or indicates that it is not safe. A processing unit that outputs a signal.

  That is, the recognition processing unit 34 recognizes the three-dimensional shape of the object sandwiched between the closed doors in the image based on the information of the distance image generated by the stereo processing unit 33, and the three-dimensional shape is predetermined. This is a processing unit that outputs a detection signal indicating that it is safe or unsafe when the conditions are not met or met.

  When the communication unit 35 receives a predetermined command from the PC 11, the communication unit 35 gives an instruction to execute a safety confirmation process to the recognition processing unit 34, and when receiving a detection signal from the recognition processing unit 34, It is a processing unit that transmits a detection signal to the PC 11. The communication unit 35 is connected to the communication cable CB via an interface circuit and a connector (not shown).

  Further, the communication unit 35 is also connected to the image capturing unit 32 so that a moving image image signal obtained from both or one of the CMOS sensors 31 can be transmitted to the PC 11 at a predetermined timing. It has become. Therefore, the PC 11 can display the image signal from each stereo camera SC on the monitor.

FIG. 4 is a diagram for explaining an example of a distance image obtained by the stereo camera SC.
When the train T stops at the station, passengers get on and off, and all the doors for getting on and off the train are closed, each stereo camera SC confirms safety based on the obtained image based on the above command. Execute the process.

  Since the stereo camera SC is fixed and the positions of the train T and the home H do not move, the areas of the train T and the home H are fixed in the generated distance screen IR. FIG. 4 is an image of an area captured by the stereo camera SC and the distance value of each pixel is generated. In the distance image IR of FIG. 4, the distance values and ranges of predetermined areas indicated by diagonal lines, that is, the area T of the train T, the area GR of the ground G, and the area HR of the home H are known in advance. It is an area. The area TR of the train T includes a door.

  That is, the ranges and distance values of these three regions TR, GR, and HR are known in advance and are constant. Accordingly, in these areas, the pixel area having a predetermined distance value is a pixel area of the background image portion, and in these areas, the pixel area not having the predetermined distance value is a three-dimensional object sandwiched between doors. It becomes a pixel region of shape.

FIG. 4 shows that the person M and the object SB exist in the image IR. In the distance image IR, the standing person M has the shortest distance to the top of the head, and the distance from the top to the periphery gradually increases. Furthermore, the distance value of the shoulder portion is a distance value corresponding to the height of the shoulder within a predetermined range from the top of the head and within a predetermined range from the top of the head. Therefore, if such a distance value is obtained, the standing person M can be recognized in the distance image IR.
In addition, an object sandwiched between two doors can be recognized as an object that exists at the edge of the home H and in the vicinity thereof and has a distance different from the distance values of the areas TR, GR, and HR.

  FIG. 4 shows that there is a person M in the home H, and a bar-like object SB is present at the end of the home H and in the vicinity thereof. In FIG. 4, there is normally no object within the construction limit indicated by the two-dot chain line (here, the end of the home H and its vicinity). When there is an object at the edge of the platform H and in the vicinity thereof when the train door is closed, there may be an object SB sandwiched between the doors.

  In FIG. 4, a region 41 indicated by a dotted line is set in advance as a target region for the safety confirmation process. The region 41 is a region where there is a possibility that there is a portion where objects are caught and jumped out of the train when the train door is closed. Since the stop position of the train is determined in advance, the area that includes the range where the train may stop and the part that has jumped out of the train of the sandwiched object may exist. Is set as area 41.

  Since the areas and distances of the train T, the ground G, and the home H are known in advance, the pixel information of the areas TR, GR, and HR of the train T, the ground G, and the home H in the area 41 is stored in the DSP. Pre-stored in the area. Therefore, as will be described later, the recognition processing unit 34 performs the background separation process by comparing the distance image of the region 41 with the pixel information of each region TR, GR, HR stored in advance. . A safety confirmation determination process is performed based on the three-dimensional shape of the object separated and extracted.

Based on the distance image generated by the stereo processing unit 33, the recognition processing unit 34 determines whether or not there is an object at the end of the home H and in the vicinity thereof, and that the object hinders the train operation. It is determined whether or not there is , and a detection signal is output as a determination result.

(Content of processing)
FIG. 5 is a flowchart showing the contents of processing in the recognition processing unit 34 of each stereo camera SC.

  Before the train door closes and departs, a safety check is performed to confirm that no object is caught in the door, or that even if it is caught, there is no obstacle to train operation. The safety check is executed in each stereo camera SC. That is, each stereo camera SC is a safety confirmation device for a railcar door. Trains cannot be started unless safety is confirmed for all doors. The instruction for the safety confirmation is performed by a safety confirmation instruction signal.

  For example, the safety confirmation instruction signal may be input to the control device 12 when a button installed in the station is pressed, or input to the control device 12 when a button installed in the monitoring room of the station is pressed. You may be made to do. When the safety confirmation instruction signal is input, the control device 12 transfers the safety confirmation instruction signal to the PC 11. When receiving the safety confirmation instruction signal, the PC 11 transfers the command to each stereo camera SC.

  When a predetermined command is received from the PC 11, the communication unit 35 outputs a predetermined signal to the recognition processing unit 34 so as to execute a safety confirmation process. When the recognition processing unit 34 receives the predetermined signal, the recognition processing unit 34 performs the process of FIG. 5 for a predetermined time. That is, the recognition process is executed for a certain time after receiving a predetermined signal.

  The processing in FIG. 5 is executed by the recognition processing unit 34 included in the DSP. First, a background separation process is executed within a predetermined area 41 in the obtained distance image (step (hereinafter abbreviated as S) 1). As described above, in the background separation process, since the areas and distances of the train T, the ground G, and the home H are known in advance, the pixels that match the information on the areas and the corresponding distances are used as background pixel portions. This is a process of separating from the distance image IR. Information on the pixels in each of the areas TR, GR, and HR of the train T, the ground G, and the home H is stored in advance in a storage area in the DSP. In the background separation process, the background can be separated by comparing the pixel information of each of the regions TR, GR, and HR stored in advance with each pixel of the generated distance image. As a result, a distance image of only an image that is not a background is extracted in the region 41.

  Therefore, S1 constitutes a background image separation unit that separates a part of a predetermined background image including a door from a distance image that is information on the generated three-dimensional shape, and a part of the predetermined background image is separated and remains The three-dimensional object extraction unit is configured to extract information regarding the three-dimensional shape of the object sandwiched between the doors from the distance image that is information regarding the three-dimensional shape.

Next, a volume calculation process is executed (S2). The volume calculation process will be described with reference to FIGS. FIG. 6 is a flowchart showing the flow of the volume calculation process. FIG. 7 is a diagram for explaining the volume calculation.
In FIG. 5 , a distance image, which is information relating to the three-dimensional shape of the object SB, is extracted by S1, and first, a cross-sectional area corresponding to each distance value of the distance image of the object SB is calculated (S11). As shown in FIG. 7, since the object SB is imaged from above the home H, the shape of the lower side of the object SB is unknown. FIG. 7 shows an umbrella as an example of the object SB. FIG. 7 shows an example in which an umbrella protrudes horizontally from a door. When the object SB is an umbrella, since the object has a substantially conical shape, the distance image is data composed of distance values of the upper surface.

  Therefore, in the volume calculation, the cross-sectional area of each distance value is calculated from the upper distance image data without considering the lower shape (indicated by a two-dot chain line) of the object SB. First, for example, as shown in FIG. 7, the area S1 of a region (shown by hatching) surrounded by the pixel PR1 having the distance value on the outermost peripheral side of the extracted object SB is calculated. In FIG. 7, the direction indicated by the arrow A is the viewing direction of the stereo camera SC. The pixel value at the outer edge of the pixel PR1, that is, the distance value is the same. However, here, as shown in FIGS. 4 and 7, since the pixel value of the door side portion of the object SB cannot be obtained, here, the line connecting the end points of the curve drawn by the pixel PR1 having the same distance value The area inside the region surrounded by is the cross-sectional area SS1. In FIG. 7, the area surrounded by the curve (or straight line) formed by the pixels PR1 having the same distance value and the line TL connecting the two end points T of the curve corresponds to the pixel PR1. Cross-sectional area.

  Next, the area SS2 of the region surrounded by the pixel PR2 having the next largest distance value than the outermost distance value is calculated in the same manner. Subsequently, the area SS3 of the region surrounded by the pixel PR3 having the next largest distance value than the distance value PR2 corresponding to the area SS2 is calculated. Such calculation of the cross-sectional area is performed for all distance values related to the object SB. As a result of S11, a cross-sectional area corresponding to each distance value of the object SB is obtained.

Next, each cross-sectional area obtained in S11 is normalized based on the distance value, and the volume of the object SB is calculated and estimated (S12).
Each cross-sectional area obtained in S11 differs depending on the distance from the stereo camera SC to the object SB. Therefore, it is necessary to normalize the cross-sectional area according to the distance to the object SB. The distance from the stereo camera SC to the object SB is, for example, the minimum distance value in the distance image of the object SB or the average value of the distance values in the distance image of the object SB. The cross-sectional area is normalized by multiplying each cross-sectional area obtained in S11 by the square of the distance value from the stereo camera SC to the object SB. Note that another method may be used to determine the distance from the stereo camera SC to the object SB.

Then, the volume L of the object SB is calculated from each normalized cross-sectional area (S12). The volume L can be obtained by taking the sum of the cross-sectional areas.

  Returning to FIG. 5, after S2, the shape category classification processing is executed (S3). FIG. 8 is a flowchart showing the flow of shape category classification processing. The shape category classification processing will be described with reference to FIGS.

  First, the aspect ratio of the cross section of the distance image is calculated and obtained (S21). This aspect ratio is the ratio of the vertical length to the horizontal length of the object SB. In FIGS. 4 and 7, for example, when the direction orthogonal to the surface of the door is the horizontal direction, The length in the horizontal direction and the length in the vertical direction perpendicular to the horizontal direction are obtained, and the ratio between the vertical and horizontal lengths is obtained.

For Figure 7, the object since SB is a umbrella, the shape of the cross-sectional area (SS1) of the largest distance value in the distance image, the aspect ratio obtained from the sideways direction length LR and the longitudinal length LL (Vertical / Horizontal) becomes smaller. Moreover, in the case of a thing like a bag, an aspect ratio becomes large compared with an umbrella.

Then, the classification of the shape category is determined from the obtained aspect ratio (S22). The shape category is previously classified into a plurality according to the aspect ratio. FIG. 9 is a diagram showing an example of classification of the predetermined shape category. The table data of FIG. 9 is stored in advance in a predetermined storage area in the DSP.
In the case of FIG. 9, the shape category is classified into 10 according to the aspect ratio. In the case of FIG. 9, an object having an aspect ratio of 1 is classified into a classification C1 in the shape category. A thing with an aspect ratio of 0.5 is classified into classification C2 in the shape category. A product having an aspect ratio of 0.1 is classified into classification C10 in the shape category. An object belonging to the classification C1 has an object shape in the distance image that is close to a square or close to a circle. An object belonging to the classification C10 has a bar shape or a bar-like shape in the distance image.

  Based on the shape category information of FIG. 9 stored in a predetermined storage area in the DSP, the recognition processing unit 34 determines the shape category based on which aspect ratio of FIG. Determine the classification. In other words, shape classification information is obtained from the aspect ratio of the object SB through S22. For example, the aspect ratio is determined by rounding off the second decimal place of the aspect ratio obtained in S21, and the classification is determined by comparing the determined aspect ratio with the aspect ratio of FIG. For example, in the case of FIG. 7, if the aspect ratio of the object SB is 0.1, the object SB is classified into the classification C10.

  Returning to FIG. 5, after S3, a deformation degree determination process is executed (S4). FIG. 10 is a flowchart showing the flow of the deformation degree determination process. FIG. 11 is a diagram for explaining the deformation of an object. The deformation degree determination process will be described with reference to FIGS.

  First, the recognition processing unit 34 detects time-series conversion of the cross section of the distance image (S31). It is detected how much the cross-sectional shape having the largest distance value in the distance image has changed within a predetermined time, for example, within one second. For example, all the pixels constituting the cross section at a certain timing are compared with all the pixels after a predetermined time has elapsed. As a result of the comparison, the ratio of the changed pixels to the total pixels is obtained.

  FIG. 11 shows that the cross-sectional shape of the object SB in the region 41 has changed from the shape indicated by the solid line to the shape indicated by the broken line during a predetermined time. In FIG. 11, the hatched area is the changed pixel portion.

  Accordingly, the recognition processing unit 34 can determine the degree of deformation of the object SB based on the ratio of the changed portion of the pixel (S32).

  For example, if the ratio of the changed pixels indicated by diagonal lines to all the pixels of the object SB before the change is 50% or more, the object SB is determined to be a soft object such as a cloth. If the ratio is less than 50%, the object SB is determined to be a hard object. Therefore, according to the deformation degree determination process, the object SB is determined to be “hard” or “soft” according to the deformation degree of the object SB.

Note that the degree of deformation is determined in two stages of “ hard” or “soft” in the above-described example, but may be determined in more than three stages.

  Returning to FIG. 5, next to S4, a safety determination process is executed (S5). This safety determination processing is performed by rule-based processing based on the results obtained in S2, S3, and S4. Therefore, this determination processing is performed by the inference engine unit in the recognition processing unit 34 using rules stored in advance in a storage area in the DSP. Here, the rule is defined based on the three parameter information of the volume, shape classification, and deformation degree of the object SB obtained from the distance image that is information on the above-described three-dimensional shape.

Here, the determination based on the predetermined rule is performed based on whether or not an object sandwiched between the train doors is an obstacle to train operation. For example, since the extracted object SB is very small if its volume L is equal to or less than a predetermined volume Lth, there is no problem in train operation regardless of the classification and deformation degree of the shape category. Therefore, in such a case, the safety determination process in S5 is performed by detecting the parameter SB obtained from the distance image and a predetermined signal based on the parameter information. Signal indicating no problem).

Also, for example, if the volume L of the extracted object SB is equal to or greater than the predetermined volume Lth, the aspect ratio is 0.1, the classification is C10, and the degree of deformation exceeds 50%, Since SB is a long and soft object, the rules are set so that the safety judgment process of S5 generates a detection signal (a signal indicating no problem) indicating that the object SB is an object that does not have a problem with operation. The

Furthermore, for example, when the volume L of the extracted object SB is equal to or larger than the predetermined volume Lth, the aspect ratio is 1.0, the classification is C1, and the degree of deformation is less than 50%, the object SB is Since it is a bulky object having a size larger than a certain volume, the safety judgment process of S5 generates a detection signal (signal indicating that there is a problem) that the object SB is an obstacle to operation. As such, the rules are set.

Furthermore, for example, when the volume L of the extracted object SB is equal to or greater than the predetermined volume Lth, the aspect ratio is 0.1, the classification is C10, and the degree of deformation is less than 50%, the object SB is Since the object is slender but hard, the rule is set so that the safety judgment process of S5 generates a detection signal (signal indicating that there is a problem) that indicates that the object SB is an obstacle to operation. The

  As described above, the safety determination process is performed based on a rule set based on whether or not there is an obstacle to train operation. By setting various rules in addition to the examples given above and setting them in advance as rules used in rule-based processing, the safety confirmation determination process can be performed in accordance with a desired purpose. .

  As described above, S5 relates to the door based on a plurality of predetermined parameter information obtained from the distance image that is information regarding the three-dimensional shape of the extracted object SB and a predetermined rule based on the predetermined parameter information. A determination unit for determining safety confirmation is configured.

Rules can be created based on past accident cases to improve detection accuracy when there is a problem. For example, the rule may be changed according to the time zone.
Rules can be added and changed by storing them in a rewritable storage area in the DSP.

  Furthermore, in the above-described example, the parameter information used for the rule serving as a determination criterion is three, that is, volume, shape classification, and deformation degree, but a rule that does not use all three may be used. Further, parameters other than these three may be used as predetermined parameters relating to the rule, or may be used in addition to the above three parameters.

  As described above, since safety confirmation processing is performed on a rule basis, a desired object can be determined by creating and adding parameters or rules for detecting the shape of various objects.

The recognition processing unit 34 executes output processing for outputting the determination result of S5 to the communication unit (S6). Therefore, the determination result by the recognition processing unit 34 is transmitted to the PC 11 and the control device 12 by communication. The PC 11 or the control device 12 can perform a warning by a necessary voice or output a control signal so that the train door opens based on the received determination result. If there is a problem with train operation, the train can be prevented from leaving.

  In the above-described embodiment, the stereo camera SC as the door safety confirmation device performs the safety confirmation process, but the process of the recognition processing unit 34 may be performed by the PC 11. That is, even if the stereo camera SC outputs a distance image as the information of the three-dimensional shape and executes the recognition process in the PC 11 or the control device 12, the same effect as the above-described embodiment can be obtained. . Alternatively, even if the stereo camera SC outputs two or more image signals, and the PC 11 or the control device 12 executes distance image generation and recognition processing, the same effects as those of the above-described embodiment can be obtained. be able to. In this case, the PC 11 constitutes a part of the door safety confirmation device.

As described above, according to the above-described railroad vehicle door safety confirmation device according to the present embodiment, it is possible to recognize a pinched object by image processing and confirm safety when the door is closed. it can.
The present invention is not limited to the above-described embodiments, and various changes and modifications can be made without departing from the scope of the present invention.

DESCRIPTION OF SYMBOLS 1 Monitoring system, 11 PC, 12 Control apparatus, 21 Case, 22 Objective optical system, 31 CMOS sensor, 32 Image capture part, 33 Stereo processing part, 34 Recognition processing part, 35 Communication part

Claims (3)

A distance image generation unit that generates a distance image that is generated based on at least two image signals obtained by imaging with at least two image sensors and each pixel is a value indicating a distance or a position in a three-dimensional space ; ,
By comparing the distance image generated by the distance image generation unit and the distance image of a predetermined background image including a railcar door stored in advance, the predetermined background image is compared with the generated distance image. A background image separation part that separates the parts;
An object distance image extracting unit that extracts a distance image of an object sandwiched between the doors from a distance image remaining after the predetermined background image is separated in the background image separating unit;
Deformation degree determination that determines the degree of deformation of the object during a predetermined time from the time-series change in the cross-sectional shape of the largest distance value in the object distance image extracted by the object distance image extraction unit And
And parameter information including the volume of the said product range image extracting unit the product obtained from the distance image of the extracted the product in, shape classification, and the parameters of the deformation degree, which is determined in the deformation degree determination unit, the volume, and the shape classification and, based on the based have been defined rule information including the deformation degree, the safety confirmation judging unit for judging safety confirmation regarding said door,
A safety confirmation device for a door for a railway vehicle, comprising:   Obtaining the aspect ratio of the cross-section of the distance image of the object, having a shape category classification determination unit that determines the classification of the shape category that has been pre-classified from the aspect ratio,
  The safety confirmation device for a railcar door according to claim 1, wherein the shape classification is a classification of a shape category determined by the shape category classification determination unit. Before Kill Lumpur, safety confirmation device door for rail vehicles according to claim 1 or 2, characterized in that additions are possible and modifications.

Images