Elevator door identification method and device

Through background modeling and historical integral diagram technology, the position of elevator doors is automatically identified, which solves the problem of poor accuracy and versatility of elevator door identification in the existing technology, and realizes efficient and accurate elevator door fault detection, reducing human resource consumption.

CN115258865BActive Publication Date: 2024-11-15CHENGDU PENGYE SOFTWARE
View PDF 16 Cites 0 Cited by

Patent Information

Application Number
CN202210942576.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-08-08
Publication Date
2024-11-15
Estimated Expiration
2042-08-08

AI Technical Summary

Technical Problem

In the detection of elevator door faults, it is difficult to accurately identify the position of elevator doors, resulting in poor detection accuracy and versatility, and requires a lot of human resources. The surveillance camera viewing angle deviation and elevator environment changes have a great impact.

Method used

Motion extraction is performed using background modeling, the elevator door area position is filtered out through outline size parameters, and the pixel movement times are counted using the historical integral graph. Combined with morphological opening operation and linear detection, the elevator door and door line positions are automatically determined to reduce manual annotation. and use tags.

Benefits of technology

It improves the accuracy and versatility of elevator door recognition, reduces human resource consumption, avoids the impact of surveillance camera perspective deviation and environmental changes on recognition, and reduces misidentification.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115258865B_ABST
    Figure CN115258865B_ABST
Patent Text Reader

Abstract

The present invention relates to the technical field of elevators, and discloses an elevator door recognition method and device, which aims to solve the problems of poor accuracy and high human resource consumption in existing elevator door area position recognition methods. The solution mainly includes: obtaining a monitoring image of the elevator; using background modeling to perform motion extraction to obtain an image mask corresponding to the monitoring image, and the image mask is used to represent the pixel change position in the monitoring image; performing contour extraction on the image mask to obtain at least one contour and its corresponding contour size parameters; screening out contours that meet the requirements according to the contour size parameters; and determining the elevator door area position according to the screened contours. The present invention improves the accuracy and versatility of elevator door area position recognition and reduces human resource consumption, and is particularly suitable for detecting the switch status of elevator doors.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method and device for identifying an elevator door. Background Art

[0002] Elevator failures can have a significant impact on the safety and experience of passengers, and elevator door failures are a very common type of elevator failure. At present, in order to detect elevator door failures, the commonly used method is to first detect the switch status of the elevator door, and then determine whether the elevator door has a fault based on the switch status of the elevator door. When detecting the switch status of the elevator door, the commonly used method is to use the elevator monitoring image and image recognition to determine the switch status of the elevator. Since the field of view of the surveillance camera is very large, it basically covers the entire elevator. And due to the influence of the installation angle, the position of the elevator door in the field of view cannot be determined, and the monitoring may only include part of the entire elevator door. Therefore, the image recognition-based method needs to first determine the door area position of the elevator, that is, the position of the elevator door in the image, otherwise it is impossible to accurately determine the switch status of the elevator door.

[0003] In order to determine the door area position of an elevator, the prior art usually adopts a method of presetting the door area position of the elevator in the monitoring screen. This method is not only complicated to operate and consumes a lot of human resources, but also the viewing angle of the monitoring camera may be offset during the maintenance of the elevator, resulting in the preset door area position of the elevator in the monitoring screen being inconsistent with the actual door area position, and its accuracy is poor. In addition, due to the differences in the use environment of different elevators and the installation location of the monitoring camera, this method needs to determine the door area position of the elevator separately for each elevator, and its versatility is poor.

[0004] Application publication number CN107187980A discloses a method and device for detecting the opening and closing of elevator doors in an elevator device. The main technical solution is to set a corresponding label on the elevator door, and use the motion area corresponding to the label in the image as the target area corresponding to the elevator door area position, and then perform corresponding image recognition processing on the target area, thereby realizing the detection of the opening and closing degree of the elevator door. This method requires manual setting of labels on the elevator door, and there is still a problem of high human resource consumption. In addition, during the use of the elevator, advertisements will be posted on the elevator door from time to time, and labels will be covered or fall off, causing the method to fail. Taking into account the influencing factors such as changes in the perspective of the elevator camera, elevator advertisements, and post-maintenance of the elevator in real scenes, the technology of automatic identification of the elevator door area becomes particularly important. Summary of the invention

[0005] The present invention aims to provide an elevator door recognition method and device, aiming to improve the accuracy and versatility of elevator door zone position recognition and reduce human resource consumption.

[0006] The technical solution adopted by the present invention to solve the above technical problems is:

[0007] On the one hand, a method for identifying an elevator door is provided, comprising the following steps:

[0008] Step 1: Obtain the monitoring image of the elevator;

[0009] Step 2: Using background modeling to perform motion extraction, obtain an image mask corresponding to the monitoring image, wherein the image mask is used to indicate a pixel change position in the monitoring image;

[0010] Step 3: extracting contours of the image mask to obtain at least one contour and its corresponding contour size parameters;

[0011] Step 4: Screen out contours that meet the requirements according to the contour size parameters;

[0012] Step 5: Determine the door area position of the elevator based on the screened outline.

[0013] Further, in step 5, determining the door zone position of the elevator according to the screened contour specifically includes:

[0014] Step 51, construct and initialize a gate area historical integral map, wherein the gate area historical integral map is used to count the number of movements of each pixel in the monitoring image;

[0015] Step 52, repeating steps 1 to 4, adding the contour screened out each time to the corresponding position of the gate area historical integral map, and increasing the movement times of the pixel point at the corresponding position in the gate area historical integral map by one;

[0016] Step 53: In the door zone historical integral map, if there is only one contour corresponding to a pixel point whose movement times is greater than a first preset value, then the contour is used as the door zone position of the elevator; if there are multiple contours corresponding to a pixel point whose movement times is greater than the first preset value, then among the multiple contours, the contour with the largest area is used as the door zone position of the elevator.

[0017] Furthermore, the monitoring image is the upper half of the elevator monitoring screen.

[0018] Furthermore, in step 3, before extracting the contour of the image mask, the following steps are further included:

[0019] A morphological opening operation is performed on the image mask to remove interference positions.

[0020] Furthermore, the background modeling method is: KNN algorithm, single Gaussian model, mixed Gaussian model, frame difference method, ViBe algorithm or CodeBook algorithm.

[0021] Furthermore, in step 4, the outline size parameters include at least: width, height, aspect ratio, perimeter, area, perimeter-to-area ratio and central moment of the outline circumscribed rectangle.

[0022] Furthermore, the step 2 further comprises:

[0023] Calculate the areas of all image masks, and determine whether the ratio of the area of ​​the image mask to the area of ​​the monitoring image is greater than a second preset value. If so, proceed to step 1; otherwise, proceed to step 3.

[0024] Furthermore, the method further comprises:

[0025] Step 6: Determine whether the elevator door in the current monitoring image is in a just-opened state according to the contour area corresponding to the door area position of the elevator. If so, proceed to step 7; otherwise, proceed to step 1.

[0026] Step 7: Perform straight line detection on the current monitoring image and select straight lines whose angles and lengths meet preset requirements;

[0027] Step 8: Determine the door line position of the elevator based on the selected straight line.

[0028] Further, in step 8, determining the door line position of the elevator according to the screened straight line specifically includes:

[0029] Step 81, construct and initialize a gate line historical integral map, wherein the gate line historical integral map is used to count the number of coincidences between the monitoring image and each pixel point in the screened straight line;

[0030] Step 82, repeating steps 6 to 7, adding the straight line selected each time to the corresponding position of the goal line historical integral map, and increasing the number of coincidences of the pixel points at the corresponding position in the goal line historical integral map by one;

[0031] Step 83, in the door line historical integral graph, if there is only one straight line corresponding to the pixel points whose number of overlaps is greater than the third preset value, then this straight line is used as the door line position of the elevator; if there are multiple straight lines corresponding to the pixel points whose number of overlaps is greater than the third preset value, then for these multiple straight lines, with each straight line as the axis, calculate the torque of the door area position on the straight line axis, and use the straight line corresponding to the minimum torque as the door line position of the elevator.

[0032] On the other hand, an elevator door recognition device is provided, comprising:

[0033] An acquisition unit, used for acquiring a monitoring image of the elevator;

[0034] A motion extraction unit, used for performing motion extraction by using background modeling to obtain an image mask corresponding to the monitoring image, wherein the image mask is used to indicate a pixel change position in the monitoring image;

[0035] A contour extraction unit, used to extract contours from the image mask to obtain at least one contour and its corresponding contour size parameters;

[0036] A screening unit, used for screening out contours that meet the requirements according to the contour size parameters;

[0037] The determination unit is used to determine the door zone position of the elevator according to the screened profile.

[0038] The beneficial effects of the present invention are as follows: the elevator door recognition method and device described in the present invention extracts motion from the monitoring image based on background modeling, and simultaneously screens the contour according to the contour size parameters, and then determines the door area position of the elevator according to the screened contour. The present invention can identify the door area and door line position of the elevator without manually marking the door area and door line position of the elevator on the monitoring screen, nor does it need to set corresponding labels on the elevator door, thereby reducing the complexity of manual operation and saving human resources. At the same time, the present invention can avoid the influence of the viewing angle offset of the monitoring camera, light changes and passenger movement on the recognition, thereby improving the recognition accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A schematic diagram of a flow chart of an elevator door recognition method according to an embodiment of the present invention;

[0040] Figure 2 It is a schematic diagram of the structure of the elevator door recognition device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0041] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0042] The elevator door recognition method and device provided by the elevator of the present invention mainly include: acquiring a monitoring image of the elevator by an acquisition unit; performing motion extraction by a motion extraction unit using background modeling to obtain an image mask corresponding to the monitoring image, wherein the image mask is used to represent the pixel change position in the monitoring image; performing contour extraction on the image mask by a contour extraction unit to obtain at least one contour and its corresponding contour size parameters; screening out contours that meet the requirements according to the contour size parameters by a screening unit; and determining the door area position of the elevator according to the screened contours by a determination unit.

[0043] Normally, in the monitoring image of the elevator, the elevator door is the largest movable component, and the elevator door is partially regular to a certain extent, and its height must be greater than its width. Based on this, the present invention performs motion extraction on the monitoring image by background modeling, determines the position where the pixel in the monitoring image changes, and performs contour extraction based on the position. The extracted contour may be caused by the switch of the elevator door, or it may be caused by the flickering of the lights or advertising display screens in the elevator, or the movement of passengers. Since the elevator door is partially regular in the monitoring image, and its height must be greater than its width. Therefore, the contour that meets the requirements can be screened out from at least one contour by contour size parameters, and used as the door area position of the elevator, thereby avoiding the influence of the flickering of the lights or advertising display screens in the elevator, or the pixel changes caused by the movement of passengers on the door area position recognition of the elevator.

[0044] Example

[0045] The elevator door recognition method according to the embodiment of the present invention is as follows: Figure 1 As shown, the following steps are included:

[0046] Step 1: Obtain the monitoring image of the elevator;

[0047] In this embodiment, the monitoring screen in the elevator car can be obtained through the monitoring camera. Since the lower half of the elevator is easily blocked by passengers, the upper half of the monitoring screen is used as the monitoring image of the elevator in this embodiment, and the subsequent door area position recognition is performed based on the monitoring image, so as to reduce the amount of data calculation and improve the recognition accuracy.

[0048] Step 2: Using background modeling to perform motion extraction, obtain an image mask corresponding to the monitoring image, wherein the image mask is used to indicate a pixel change position in the monitoring image;

[0049] Among them, background modeling is a method for segmenting all moving targets in a scene with a fixed camera. The background modeling method in this embodiment can be: KNN algorithm, single Gaussian model, mixed Gaussian model, inter-frame difference method, ViBe algorithm or CodeBook algorithm. In the actual application process, the corresponding method can be selected according to the actual situation, and this embodiment does not limit this.

[0050] This embodiment performs motion extraction on the monitoring image through background modeling to obtain a foreground mask corresponding to the monitoring image. The foreground mask can reflect the pixel change position in the monitoring image.

[0051] Step 3: extracting contours of the image mask to obtain at least one contour and its corresponding contour size parameters;

[0052] In order to further improve the accuracy of elevator door recognition, this implementation also calculates the area of ​​all image masks before performing contour extraction, and determines whether the ratio of the area of ​​the image mask to the area of ​​the monitoring image is greater than a second preset value. If so, it may be a false motion caused by the influence of lighting. In this case, the motion is determined to be invalid, and this frame is skipped to enter step 1. Otherwise, a morphological opening operation is performed on the image mask, and after eliminating the interference position, step 3 is entered to perform contour extraction to obtain at least one contour and its corresponding contour size parameter.

[0053] The second preset value may be set according to actual conditions, and this embodiment does not limit this.

[0054] In this embodiment, the outline size parameters include at least: the width, height, aspect ratio, perimeter, area, perimeter-to-area ratio and central moment of the outline circumscribed rectangle.

[0055] Step 4: Screen out contours that meet the requirements according to the contour size parameters;

[0056] It is understandable that the opening and closing of the elevator door, the flashing of the lights or advertising display screen in the elevator, and the movement of passengers will cause the pixels at the corresponding position of the monitoring image to change. Therefore, in practical applications, more than one contour may be extracted, and since the elevator door is partially regular in the monitoring image, and its height must be greater than its width. Based on this, this embodiment screens the contour by the width, height, aspect ratio, perimeter, area, perimeter-to-area ratio, and central moment of the circumscribed rectangle of the contour to obtain a contour that meets the requirements.

[0057] Step 5: Determine the door area position of the elevator based on the screened outline.

[0058] In this embodiment, the screened contour can be directly used as the door zone position of the elevator.

[0059] In order to further improve the accuracy of elevator door recognition, this embodiment can also determine the door zone position of the elevator by the following method:

[0060] Step 51, construct and initialize a gate area historical integral map, wherein the gate area historical integral map is used to count the number of movements of each pixel in the monitoring image;

[0061] Step 52, repeating steps 1 to 4, adding the contour screened out each time to the corresponding position of the gate area historical integral map, and increasing the movement times of the pixel point at the corresponding position in the gate area historical integral map by one;

[0062] Step 53: In the door zone historical integral map, if there is only one contour corresponding to a pixel point whose movement times is greater than a first preset value, then the contour is used as the door zone position of the elevator; if there are multiple contours corresponding to a pixel point whose movement times is greater than the first preset value, then among the multiple contours, the contour with the largest area is used as the door zone position of the elevator.

[0063] It is understandable that since there may be passengers carrying larger regular objects through the elevator, this will also cause the pixels in the monitoring image to change, and the corresponding contours may also meet the contour screening requirements. Only one recognition may misidentify it as the door area position, thereby reducing the accuracy of door area position recognition. In order to solve the above problem, this embodiment repeatedly executes steps 1-4 multiple times, and at the same time counts the number of movements of each pixel point in the monitoring image through the constructed door area historical integral map, and determines the door area position of the elevator based on the number of movements of each pixel point and the contour area.

[0064] Specifically, after each execution of steps 1-4, the selected contour is added to the corresponding position of the door area historical integral map, and the movement times of the pixel points at the corresponding position in the door area historical integral map are increased by one, and so on. After repeated execution for multiple times, the door area position of the elevator is determined according to the movement times of each pixel point in the door area historical integral map and the contour area.

[0065] It can be understood that the number of pixel movements caused by passengers carrying larger regular objects through the elevator is relatively small. Based on this, this implementation eliminates the interference of passengers carrying larger objects on door area recognition by screening out pixel points and their corresponding contours with a number of movements greater than the first preset value; if there is only one contour determined, it means that there is no light interference such as advertising display screens, and the contour is directly used as the door area position of the elevator; if there are multiple contours determined, it means that it may be due to the interference caused by the continuous playback of the advertising display screen in the elevator car, and the contour area corresponding to this part should be smaller than the contour area corresponding to the door area position. At this time, the contour with the largest contour area is used as the door area position of the elevator; when the door area position of the elevator is stable and no longer changes within the preset time, it is determined that the door area learning converges.

[0066] The first preset value may be determined according to the number of times steps 1-4 are repeated, and the number of times steps 1-4 are repeated may be set according to actual conditions, which is not limited in this embodiment.

[0067] In order to detect the door line position of the elevator, this embodiment also includes:

[0068] Step 6: Determine whether the elevator door in the current monitoring image is in a just-opened state according to the contour area corresponding to the door area position of the elevator. If so, proceed to step 7; otherwise, proceed to step 1.

[0069] Specifically, after the door area position of the elevator is determined by the above method, if the contour area corresponding to the door area position of the elevator is greater than zero but less than a preset value, it is determined that the elevator door in the current monitoring image is in a just-opened state, and the contour is a straight line, that is, there is a straight line in the current monitoring image to represent the door line position, and step 7 is entered at this time.

[0070] Step 7: Perform straight line detection on the current monitoring image and select straight lines whose angles and lengths meet preset requirements;

[0071] According to the above analysis, there is a straight line in the current monitoring image used to represent the door line position, but because there may be other straight lines in the elevator, there may be more than one straight line obtained by performing straight line detection on the current monitoring image. Under normal circumstances, the elevator door line should be approximately vertical in the monitoring image and its length is relatively long. Based on this, in order to avoid the influence of other straight lines on the door line position recognition of the elevator, this embodiment selects straight lines that may be door lines by angle and length.

[0072] Step 8: Determine the door line position of the elevator based on the selected straight line.

[0073] Since there may be advertising stickers with straight lines on the elevator door, when the advertising stickers move with the elevator door, the pixels at the corresponding position in the monitoring image will also change, thereby generating a straight line profile. Therefore, more than one straight line may be screened out. In order to avoid the influence of the straight lines on the advertising stickers on the elevator door on the elevator door door line recognition, this embodiment can calculate the torque of the door area position on the straight line axis with each straight line as the axis for the multiple straight lines screened out, and use the straight line with the smallest torque as the elevator door line position. This further improves the accuracy of door line position recognition.

[0074] Similarly, in order to further improve the accuracy of elevator door line recognition, this embodiment can also determine the elevator door line position by the following method:

[0075] Step 81, construct and initialize a gate line historical integral map, wherein the gate line historical integral map is used to count the number of coincidences between the monitoring image and each pixel point in the screened straight line;

[0076] Step 82, repeating steps 6 to 7, adding the straight line selected each time to the corresponding position of the goal line historical integral map, and increasing the number of coincidences of the pixel points at the corresponding position in the goal line historical integral map by one;

[0077] Step 83, in the door line historical integral graph, if there is only one straight line corresponding to the pixel points whose number of overlaps is greater than the third preset value, then this straight line is used as the door line position of the elevator; if there are multiple straight lines corresponding to the pixel points whose number of overlaps is greater than the third preset value, then for these multiple straight lines, with each straight line as the axis, calculate the torque of the door area position on the straight line axis, and use the straight line corresponding to the minimum torque as the door line position of the elevator.

[0078] It is understandable that since there may be passengers carrying straight objects through the elevator, and the angle and length of the objects may also meet the preset requirements, there may be door line misidentification through only one recognition, and the accurate door line position cannot be obtained. Based on this, this embodiment repeatedly executes steps 6-7 for many times, and at the same time, the number of coincidences between the monitoring image and each pixel point in the screened straight line is counted through the constructed door line history integral graph, and the door line position of the elevator is determined based on the number of coincidences and torque of each pixel point.

[0079] Specifically, after each execution of steps 6-7, the selected straight line is added to the corresponding position of the door line history integral map. If the selected straight line coincides with a pixel point in the monitoring image, the number of coincidences of the pixel point at the corresponding position in the door line history integral map is increased by one, and so on. After repeated execution for multiple times, the door line position of the elevator is determined according to the number of coincidences of each pixel point in the door line history integral map and the torque.

[0080] It can be understood that the number of overlaps corresponding to the pixel points caused by passengers carrying objects with straight lines through the elevator is small. Based on this, this implementation eliminates the interference of passengers carrying objects with straight lines on door line recognition by screening out the pixel points and their corresponding straight lines with the number of overlaps greater than the third preset value; if there is only one straight line determined, it means that there are no other straight lines in the elevator that interfere with the door line recognition, and the straight line is directly used as the door line position of the elevator; if there are multiple straight lines determined, it means that it may be due to the interference of advertising stickers with straight lines on the elevator door, which interferes with the door line recognition. The straight line should not be in the center line position of the door area. At this time, each straight line is used as the axis, and the torque of the door area position on the straight line axis is calculated, and the straight line corresponding to the minimum torque is used as the door line position of the elevator; when the door line position of the elevator is stable and no longer changes within the preset time, it is determined that the door line learning converges.

[0081] The third preset value may be determined according to the number of times steps 6-7 are repeatedly executed, and the number of times steps 6-7 are repeatedly executed may be set according to actual conditions, which is not limited in this embodiment.

[0082] In summary, the elevator door recognition method provided in this embodiment performs motion extraction on the monitoring image based on background modeling, and at the same time, screens the contour according to the contour size parameter, and then determines the door area position of the elevator according to the screened contour. After determining the door line position, the monitoring image corresponding to the elevator door that has just been opened is subjected to straight line detection, and the straight line is screened according to the angle and length, and then the door line position of the elevator is determined according to the screened straight line. In this embodiment, there is no need to manually mark the door area and door line position of the elevator on the monitoring screen, nor is there a need to set a corresponding label on the elevator door, so that the door area and door line position of the elevator can be identified, which reduces the complexity of manual operation and saves human resources. At the same time, this embodiment can avoid the influence of the viewing angle offset, light changes and passenger movement of the monitoring camera on the door area recognition of the elevator, and can also avoid the influence of the presence of a large part of the regular objects in the elevator on the door area and door line recognition of the elevator, and can also avoid the influence of the advertising stickers with straight lines on the elevator door on the door line recognition of the elevator, thereby improving the accuracy of elevator door recognition.

[0083] Based on the above technical solution, this embodiment also proposes an elevator door recognition device, such as Figure 2 As shown, including:

[0084] An acquisition unit, used for acquiring a monitoring image of the elevator;

[0085] A motion extraction unit, used for performing motion extraction by using background modeling to obtain an image mask corresponding to the monitoring image, wherein the image mask is used to indicate a pixel change position in the monitoring image;

[0086] A contour extraction unit, used to extract contours from the image mask to obtain at least one contour and its corresponding contour size parameters;

[0087] A screening unit, used for screening out contours that meet the requirements according to the contour size parameters;

[0088] The determination unit is used to determine the door zone position of the elevator according to the screened profile.

[0089] It can be understood that since the elevator door recognition device described in the embodiment of the present invention is a device for implementing the elevator door recognition method described in the embodiment, for the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method.

Claims

1. The elevator door recognition method is characterized by: The following steps are involved: Step 1: Obtain the monitoring image of the elevator; Step 2: Using background modeling to perform motion extraction, obtain an image mask corresponding to the monitoring image, wherein the image mask is used to indicate a pixel change position in the monitoring image; Step 3: extracting contours of the image mask to obtain at least one contour and its corresponding contour size parameters; Step 4: Screen out contours that meet the requirements according to the contour size parameters; Step 5: Determine the door area position of the elevator according to the screened contour; In step 5, determining the door area position of the elevator according to the screened contour specifically includes: Step 51, construct and initialize a gate area historical integral map, wherein the gate area historical integral map is used to count the number of movements of each pixel in the monitoring image; Step 52, repeating steps 1 to 4, adding the contour screened out each time to the corresponding position of the gate area historical integral map, and increasing the movement times of the pixel point at the corresponding position in the gate area historical integral map by one; Step 53: In the door zone history integral map, if there is only one contour corresponding to the pixel point whose movement times is greater than the first preset value, the contour is used as the door zone position of the elevator; if there are multiple contours corresponding to the pixel point whose movement times is greater than the first preset value, the contour with the largest area among the multiple contours is used as the door zone position of the elevator; Step 6: Determine whether the elevator door in the current monitoring image is in a just-opened state according to the contour area corresponding to the door area position of the elevator. If so, proceed to step 7; otherwise, proceed to step 1. Step 7: Perform straight line detection on the current monitoring image and select straight lines whose angles and lengths meet preset requirements; Step 8: Determine the door line position of the elevator based on the selected straight line.

2. The elevator door recognition method according to claim 1, characterized in that: The monitoring image is the upper half of the elevator monitoring screen.

3. The elevator door recognition method according to claim 1, characterized in that: In step 3, before extracting the contour of the image mask, the following steps are also included: A morphological opening operation is performed on the image mask to remove interference positions.

4. The elevator door recognition method according to claim 1, characterized in that: The background modeling method is: KNN algorithm, single Gaussian model, mixed Gaussian model, frame difference method, ViBe algorithm or CodeBook algorithm.

5. The elevator door recognition method according to claim 1, characterized in that: In step 4, the outline size parameters include at least: the width, height, aspect ratio, perimeter, area, perimeter-to-area ratio and central moment of the outline circumscribed rectangle.

6. The elevator door recognition method according to claim 1, characterized in that: The step 2 also includes: Calculate the areas of all image masks, and determine whether the ratio of the area of ​​the image mask to the area of ​​the monitoring image is greater than a second preset value. If so, proceed to step 1; otherwise, proceed to step 3.

7. The elevator door recognition method according to claim 1, characterized in that: In step 8, determining the door line position of the elevator according to the screened straight line specifically includes: Step 81, construct and initialize a gate line historical integral map, wherein the gate line historical integral map is used to count the number of coincidences between the monitoring image and each pixel point in the screened straight line; Step 82, repeating steps 6 to 7, adding the straight line selected each time to the corresponding position of the goal line historical integral map, and increasing the number of coincidences of the pixel points at the corresponding position in the goal line historical integral map by one; Step 83, in the door line historical integral graph, if there is only one straight line corresponding to the pixel points whose number of overlaps is greater than the third preset value, then this straight line is used as the door line position of the elevator; if there are multiple straight lines corresponding to the pixel points whose number of overlaps is greater than the third preset value, then for these multiple straight lines, with each straight line as the axis, calculate the torque of the door area position on the straight line axis, and use the straight line corresponding to the minimum torque as the door line position of the elevator.

8. An elevator door recognition device, characterized in that: The device is used to implement the elevator door recognition method according to any one of claims 1 to 7, comprising: An acquisition unit, used for acquiring a monitoring image of the elevator; A motion extraction unit, used for performing motion extraction by using background modeling to obtain an image mask corresponding to the monitoring image, wherein the image mask is used to indicate a pixel change position in the monitoring image; A contour extraction unit, used to extract contours from the image mask to obtain at least one contour and its corresponding contour size parameters; A screening unit, used for screening out contours that meet the requirements according to the contour size parameters; The determination unit is used to determine the door zone position of the elevator according to the screened profile, specifically for: Construct and initialize a gate area historical integral map, wherein the gate area historical integral map is used to count the number of movements of each pixel in the monitoring image; Repeat the contour screening step, add the contour screened each time to the corresponding position of the gate area historical integral map, and increase the number of movements of the pixel point at the corresponding position in the gate area historical integral map by one; In the door area historical integral map, if there is only one contour corresponding to a pixel point whose movement times is greater than a first preset value, the contour is used as the door area position of the elevator; if there are multiple contours corresponding to a pixel point whose movement times is greater than the first preset value, among the multiple contours, the contour with the largest area is used as the door area position of the elevator.

Citation Information

Patent Citations

  • Detecting method and device for opening and closing of elevator door in elevator device

    CN107187980A

  • Elevator door opening and closing detection method based on camera images

    CN106986248A

  • Elevator door state line detection method based on video images

    CN108584586A

  • Object contour extraction method based on mask-RCNN

    CN108898610A

  • Method and system for recognizing moving field marker based on image processing

    CN109271937A