CN115116191B - Machine vision-based long and large three-dimensional area non-blind area foreign matter monitoring device and method - Google Patents

Abstract

The invention discloses a device and a method for monitoring a large three-dimensional area non-blind area foreign matter based on machine vision; dividing a large three-dimensional area to be detected into one or more hexahedral non-blind area detection areas, wherein the hexahedrons have two opposite planes which are parallel rectangles and have the same size; a set of monitoring devices are arranged in each non-blind area detection area in the hexahedron; the monitoring device comprises a detection device, a detection output device and a control processing device; the detection device comprises two groups of background references and a visual sensor; two groups of background references are respectively positioned on two opposite parallel rectangular surfaces of the non-blind area detection area in each hexahedron; at least two diagonal points of each group of background references are provided with visual sensors, and the connecting lines of the two diagonal points of the two groups of background references, which are provided with the two visual sensors, are different-plane straight lines for respectively acquiring images of the two groups of background references; after the control processing device processes the image, whether foreign matters enter the detection area or not is judged, and whether the detection output device gives an alarm or not is controlled.

Description

Machine vision-based long and large three-dimensional area non-blind area foreign matter monitoring device and method

Technical Field

The invention relates to the field of security protection and foreign matter monitoring, in particular to a device and a method for monitoring a long and large three-dimensional area non-blind area foreign matter based on machine vision.

Background

In various industrial applications of modern society, there are often some areas needing to be controlled with emphasis, such as stations of subways and high-speed rail stations, nuclear power stations, museums, underground passages, vaults, dangerous goods warehouses and other oversized three-dimensional space scenes, in the areas needing to be controlled with emphasis, in order to ensure personal safety, equipment safety, property safety and normal operation of the system, a safety protection system is required to be established, and no invasion of foreign matters and people is ensured in a specified three-dimensional area on the ground and in any place in the air within a specified time; meanwhile, the concealment and the installation position of the monitoring equipment are strictly limited, and the space where the sensor of the monitoring equipment is located is required to be in the monitored three-dimensional space region.

The existing safety protection system is almost used for monitoring the intrusion of the boundary, but is used for detecting a discontinuous three-dimensional region, and only when foreign matters pass through the boundary, the detection blind area exists in the safety protection system, and the risk of being deliberately avoided exists, for example, an intruder enters a monitoring region at the time of system shutdown such as short power failure, system maintenance upgrading, non-monitoring time and the like, and then the system can not work again to detect the condition of the intruder.

For the foreign matter monitoring of the continuous large-and-large three-dimensional area, the existing perimeter safety protection system has one or more defects in detection areas, detection ranges, detection blind areas, detection precision and visibility, and the common perimeter safety protection system mainly comprises the following components in terms of implementation principle: video monitoring, infrared correlation detection, laser detection, microwave detection, vibration sensing optical cables, laser radars, combinations of these, and the like.

The video monitoring is the most commonly used method, and has the advantages that the visual and detection area is a three-dimensional space, but the monitoring range of the large space of a single camera has a large dead angle, is easy to intentionally avoid, and can obtain clear images of long and short distances only by matching with a zoom lens; meanwhile, the influence of weather and light is large, and the detection accuracy can be greatly reduced in the conditions of weather with low visibility, at night and the like. The simple video detection has obvious defects in the foreign matter monitoring of the continuous long and large three-dimensional area.

The infrared correlation detection mainly utilizes infrared rays to form a warning line between a transmitter and a receiver, adopts a shielding alarm mode, and forms a protection area which is a single beam or a reticular protection area formed by multiple beams. Since the light spot of the infrared light gradually increases along with the increase of the detection distance, the infrared light spot is not suitable for long-distance detection, and laser correlation detection can be used when long-distance detection is needed. The detection range of the system is usually one surface, three-dimensional detection cannot be carried out, a large detection blind area exists, and the system has obvious defects in visibility.

The microwave detection cable is characterized in that a microwave emitter and a microwave receiver are arranged on the cable at intervals, the emitter emits microwaves to adjacent receivers to form an arc detection area, and when the microwave receiver cannot receive nearby microwaves, the microwave receiver judges that foreign matters exist. The detection area formed by the method is a nearby area along the direction of the detection optical cable, a detection blind area exists in the internal area of the detection area, three-dimensional detection cannot be carried out, and the detection result is invisible.

The vibration sensing optical cable analyzes disturbance light information by monitoring the influence of the surrounding vibration or stress change of the perimeter on the light signal in the optical fiber. So that the intrusion behavior can be identified and sensed for this disturbance. The scheme has the advantages of being free from the influence of terrains and climates, and the sensing optical cable is passive, low in power consumption, low in use cost, long in monitoring distance and capable of monitoring the perimeter in real time. However, the detection area formed by the method is only a nearby area along the direction of the detection optical cable, so that the safety is low, the detection blind area is large, the three-dimensional detection cannot be performed, and the detection result is invisible.

The laser radar emits a beam of laser through the group laser transmitters, the laser beam returns to the laser receiver after encountering an object through diffuse reflection, and the radar module calculates the distance between the transmitters and the object according to the time of transmitting and receiving signals, so as to detect whether invasion exists or not, and a triangular detection surface or a three-dimensional cone laser detection interval is obtained through a scanning method. The technical advantages are that the detection distance is long, the resolution ratio is high, the interference from the environment is small, and the like, and the defects are that the target range of single-line laser radar scanning is limited, and the multi-line laser radar has the problems of high cost, overlarge energy consumption, difficult assembly and adjustment and difficult mass production, generates larger detection blind areas like video monitoring, and cannot form a cuboid non-blind area detection area limited in a cuboid.

In summary, the existing perimeter security protection systems generally have one or more defects in terms of three-dimension detection, detection range, detection blind area, detection precision and visibility, and most perimeter security protection systems cannot simultaneously achieve the detection of blind area-free, long-distance, visible and full-region three-dimension detection, so that the detection requirement of continuous long and large three-dimension regions cannot be met.

Disclosure of Invention

The invention provides a long and large three-dimensional area non-blind area foreign matter monitoring device based on machine vision, which aims to overcome the defects in the prior art and has the advantages of three-dimensional detection, non-blind area, long distance, high precision, strong visibility and strong reliability.

The second object of the invention is to provide a machine vision-based long and large three-dimensional area non-blind area foreign matter monitoring method, which can realize three-dimensional real-time detection of a detection area and realize the three-dimensional detection of a long distance, visible and full area without blind areas.

The technical scheme for solving the technical problems is as follows:

the utility model provides a three-dimensional area non-blind area foreign matter monitoring devices grows up based on machine vision, includes the monitoring devices who sets up the non-blind area monitoring region in the three-dimensional area that grows up, wherein, the non-blind area monitoring region in the three-dimensional area that grows up is formed by the combination of one or more hexahedron non-blind area detection region, and wherein hexahedron exists two opposite faces and is parallel rectangle, and the size is equal, is provided with a set of monitoring devices in every hexahedron non-blind area detection region, wherein, monitoring devices includes detection device, detection output equipment and control processing equipment, wherein, detection device is used for detecting whether there is the foreign matter in every hexahedron non-blind area detection region, including background reference thing and vision sensor, wherein, background reference thing is two sets of, and two sets of background reference thing are located respectively in every hexahedron the two opposite parallel rectangle faces of hexahedron non-blind area detection region; the length and width of each set of background references are equal to the length and width of the face, respectively; the two groups of background references are respectively a background reference ref1 and a background reference ref2; the visual sensors are at least four groups, and lens focuses of two groups of visual sensors are positioned on two diagonal points of the background reference object ref 1; the lens focuses of the other two groups of vision sensors are positioned on two opposite angles of the background reference ref2, wherein the connecting lines of the lens focuses of the vision sensors positioned on the two pairs of corner points of the background reference ref1 and the connecting lines of the lens focuses of the vision sensors positioned on the two pairs of corner points of the background reference ref2 are intersected on the vertical projection line of the background reference ref1 or the background reference ref2; wherein, the vision sensor located on the diagonal point of the background reference object ref1 is used for acquiring the image of the background reference object ref2, and the vision sensor located on the diagonal point of the background reference object ref2 is used for acquiring the image of the background reference object ref 1; the control processing equipment is used for receiving the image data returned by the visual sensor, judging whether foreign matters enter the detection area without blind areas in the hexahedron through processing and comparing the image information in the areas of the background reference object ref1 and the background reference object ref2, and controlling the alarm and result output of the detection output equipment when detecting that the foreign matters enter the detection area without blind areas in the hexahedron.

Preferably, the visual sensors in each hexahedral non-blind area detection area are four groups, namely a visual sensor sen1, a visual sensor sen2, a visual sensor sen3 and a visual sensor sen4; the corner points of the background reference object ref1 are A (101), B (102), C (103) and D (104) respectively; the corner points of the background reference object ref2 are E (105), F (106), G (107) and H (108) respectively; wherein lens foci of the visual sensor sen1 and the visual sensor sen2 are located at the corner a (101) and the corner D (104) of the background reference ref1, respectively, and lens foci of the visual sensor sen3 and the visual sensor sen4 are located at the corner F (106) and the corner G (107) of the background reference ref1, respectively; or the lens focuses of the vision sensor sen1 and the vision sensor sen2 are respectively located at the corner B (102) and the corner C (103) of the background reference ref1, and the lens focuses of the vision sensor sen3 and the vision sensor sen4 are respectively located at the corner E (105) and the corner H (108) of the background reference ref 1.

Preferably, the visual sensors in each hexahedral non-blind area detection area are eight groups; the corner points of the background reference object ref1 are A (101), B (102), C (103) and D (104) respectively; the corner points of the background reference object ref2 are E (105), F (106), G (107) and H (108) respectively; the lens focuses of the four groups of vision sensors are respectively positioned on four opposite angles of the background reference object ref1, and the lens focuses of the four groups of vision sensors are respectively positioned on four opposite angles of the background reference object ref 2.

Preferably, the background reference is set to emit light or/and is a marker with high color contrast with the surrounding environment, wherein the light emitted by the light-emitting marker is visible light or/and invisible light.

Preferably, the detection output device is used for displaying detection results, and comprises one or more of an audible and visual alarm device, a display and a background server.

Preferably, one or more sets of control processing equipment and detection output equipment are shared among the sets of monitoring devices in the non-blind area detection areas in the hexahedrons.

Preferably, each group of vision sensors is a monocular camera or a binocular camera, and a visible light camera and/or an infrared camera is adopted.

Preferably, the visual sensor, the control processing device and the detection output device are connected through a wired connection or/and a wireless connection.

The long and large three-dimensional area to be monitored is divided into one or more hexahedral non-blind area detection areas, and a set of monitoring devices are arranged in each hexahedral non-blind area detection area; wherein the foreign matter detection includes the steps of:

S1, detecting and collecting image information of a non-blind area detection area in a hexahedron through vision sensors, wherein two groups of vision sensors positioned on opposite angles of a background reference object ref1 are used for shooting images of the background reference object ref2, and two groups of vision sensors positioned on opposite angles of the background reference object ref2 are used for shooting images of the background reference object ref1 and transmitting the image information to control processing equipment;

S2, the control processing equipment processes the image information collected in each hexahedral non-blind area detection area, determines the position of a background reference object in the image, identifies the boundary condition of the background reference object image, and retains the image information of the background reference object without foreign matters for information comparison;

S3, controlling processing equipment to detect a change value between a real-time image of a background reference object in a detection area without blind areas in the hexahedron and an image in a boundary of the background reference object when the background reference object is not blocked by foreign objects, if the change value is larger than a set threshold value, judging that the foreign objects invade the detection area without blind areas in the hexahedron, and entering step S4; otherwise, enter step S5;

S4, the control processing equipment outputs the processed information to the detection output equipment, and an image of a foreign object position is displayed in the display for manually confirming whether the false alarm is generated or not, and meanwhile, the audible and visual alarm device sends out alarm sounds to inform the invasion position of the foreign object; simultaneously starting a video recording function, and storing live videos and partial images; then re-entering step S3;

And S5, the control processing equipment outputs the processed information to the detection output equipment, the display displays the image information of the detection area, if the detection area is too many and can not be displayed completely, the display is performed in a carousel mode, and then the step S3 is restarted.

Compared with the prior art, the invention has the following advantages and beneficial effects:

1. The machine vision-based long and large three-dimensional area non-blind area foreign matter monitoring device method provided by the invention realizes continuous monitoring of a three-dimensional interval in a continuous long and large three-dimensional detection area, has better detection continuity, three-dimensional detection performance and a larger detection range compared with the existing monitoring method, and can realize 1m-1000m or longer-distance internal foreign matter monitoring.

2. The method for monitoring the large and large three-dimensional area non-blind area foreign matter based on the machine vision detects images of the background reference object ref1 and the background reference object ref2 through the vision sensor, and compared with a traditional machine vision detection method, the method is low in environmental interference, higher in robustness, stronger in anti-interference performance and lower in false/missing detection rate. If an infrared luminous background reference object is adopted, the background reference object is hidden and cannot cause light pollution to people.

3. The invention provides a method for monitoring a large three-dimensional area non-blind area foreign matter based on machine vision, which provides a non-blind area hexahedral monitoring device, wherein at least four groups of visual sensors are used for monitoring in the hexahedral monitoring area, and through special diagonal setting, thereby mutually supplementing visual angle blind areas, further completely eliminating detection blind areas therein, and realizing detection without blind areas in the hexahedron.

4. Compared with other methods, the method for monitoring the long and large three-dimensional area non-blind area foreign matter based on machine vision has higher adaptability and flexibility, when the condition of a complex three-dimensional detection area is faced, such as irregular area shape, oversized area, overlong area, area block division and the like, the three-dimensional detection area can be approximately decomposed into one or more hexahedral non-blind area detection areas, and then the monitoring device is utilized for free splicing, so that the non-blind area monitoring of the whole three-dimensional detection area is ensured, and the characteristic of flexible customization can be realized to enable the three-dimensional detection area to face a vast number of application scenes.

5. The method for monitoring the foreign matters in the large three-dimensional area without the blind areas based on the machine vision adopts the vision sensor, and when the foreign matters are found, the information such as images, videos, time and the like can be automatically saved, and the log is saved so as to be convenient for evidence collection. Compared with other methods, the method has better visibility, and the information is more comprehensive and visual.

6. According to the method for monitoring the large three-dimensional area non-blind area foreign matter based on the machine vision, the connecting lines between the lens focus of each group of vision sensors and four corner points of a background reference object opposite to the vision sensors form tetrahedrons, the four groups of vision sensors can form four tetrahedrons, and the range formed by the four tetrahedrons encloses a hexahedral space formed by the hexahedral non-blind area detection area, so that non-blind area detection can be realized, and in addition, the hexahedral space formed by the hexahedral non-blind area detection area is enlarged and lengthened no matter how; the monitoring device can realize the non-blind area detection of the hexahedral space formed by the hexahedral non-blind area detection area through four groups of visual sensors (namely, meeting the setting requirements), so that the detection cost can be reduced while the non-blind area detection is realized.

7. The method for monitoring the foreign matters in the large and large three-dimensional area without the blind areas based on the machine vision comprises the steps of shooting images of background references in shooting pictures through a visual sensor, detecting whether invasion or foreign matters exist or not through an image recognition technology, and transmitting monitoring results and live real-time images in real time. As long as a detection area without blind areas in the hexahedron detects the foreign matters, the detection area without blind areas in the large three-dimensional area is indicated to have the foreign matters. When the foreign matter is found, the information such as images, time and the like is automatically saved, the alarm is rapidly given, the log is saved, the evidence is conveniently obtained, no blind area is formed in the long and large three-dimensional detection area, the detection range is continuous, the three-dimensional detection of the whole area is realized, the positioning and the detection result of the invasion foreign matter are visible.

Drawings

Fig. 1 is a schematic diagram of an installation position of a machine vision-based long and large three-dimensional area non-blind area foreign matter monitoring device according to the present invention (i.e., a first embodiment).

Fig. 2 is a schematic view of the detection range of the vision sensor in fig. 1.

Fig. 3 is a system structural block diagram of the machine vision-based long and large three-dimensional area non-blind area foreign matter monitoring device.

Fig. 4 is a schematic diagram showing a combination of a plurality of machine vision-based large three-dimensional area non-blind area foreign matter monitoring devices according to the present invention (i.e., a second embodiment).

Detailed Description

The present invention will be described in further detail with reference to the following examples and drawings, but the embodiments of the present invention are not limited thereto.

Referring to fig. 1-4, the machine vision-based long-sized and large-sized three-dimensional area non-blind area foreign matter monitoring device comprises a monitoring device arranged in a long-sized and large-sized three-dimensional area non-blind area monitoring area, wherein the long-sized three-dimensional area non-blind area monitoring area is formed by combining one or more hexahedral non-blind area detection areas, two opposite surfaces of the hexahedral non-blind area detection areas are parallel rectangles, the sizes of the hexahedral non-blind area detection areas are equal, a set of monitoring device is arranged in each hexahedral non-blind area detection area, the monitoring device comprises a detection device, a detection output device and a control processing device, the detection device is used for detecting whether foreign matters exist in each hexahedral non-blind area detection area or not, and the detection device comprises background references and a visual sensor, wherein the background references are two groups, and the two groups of background references are respectively positioned on two opposite parallel rectangular surfaces of each hexahedral non-blind area detection area; the length and width of each set of background references are equal to the length and width of the face, respectively; the two groups of background references are respectively a background reference ref1 and a background reference ref2; the visual sensors are at least four groups, and lens focuses of two groups of visual sensors are positioned on two diagonal points of the background reference object ref 1; the lens focuses of the other two groups of vision sensors are positioned on two opposite angles of the background reference ref2, wherein the connecting lines of the lens focuses of the vision sensors positioned on the two pairs of corner points of the background reference ref1 and the connecting lines of the lens focuses of the vision sensors positioned on the two pairs of corner points of the background reference ref2 are intersected on the vertical projection line of the background reference ref1 or the background reference ref2; wherein, the vision sensor located on the diagonal point of the background reference object ref1 is used for acquiring the image of the background reference object ref2, and the vision sensor located on the diagonal point of the background reference object ref2 is used for acquiring the image of the background reference object ref 1; the control processing equipment is used for receiving the image data returned by the visual sensor, judging whether foreign matters enter the detection area without blind areas in the hexahedron through processing and comparing the image information in the areas of the background reference object ref1 and the background reference object ref2, and controlling the alarm and result output of the detection output equipment when detecting that the foreign matters enter the detection area without blind areas in the hexahedron.

In this embodiment, the two surfaces of the hexahedral non-blind area detection area for mounting the background reference object are parallel to each other, the appearance is rectangular, and the sizes of the two surfaces are equal. In addition, each group of vision sensors is a monocular camera or a binocular camera, and a visible light camera and/or an infrared camera is adopted.

Referring to fig. 1 to 3, the visual sensors in each non-blind area detection area in the hexahedron are four groups, namely a visual sensor sen1, a visual sensor sen2, a visual sensor sen3 and a visual sensor sen4; the corner points of the background reference object ref1 are A (101), B (102), C (103) and D (104) respectively; the corner points of the background reference object ref2 are E (105), F (106), G (107) and H (108) respectively; wherein lens foci of the visual sensor sen1 and the visual sensor sen2 are located at the corner a (101) and the corner D (104) of the background reference ref1, respectively, and lens foci of the visual sensor sen3 and the visual sensor sen4 are located at the corner F (106) and the corner G (107) of the background reference ref1, respectively; or the lens focuses of the vision sensor sen1 and the vision sensor sen2 are respectively located at the corner B (102) and the corner C (103) of the background reference ref1, and the lens focuses of the vision sensor sen3 and the vision sensor sen4 are respectively located at the corner E (105) and the corner H (108) of the background reference ref 1. Through the arrangement, the connection lines between the lens focus of each group of vision sensors and the four corner points of the background reference object opposite to the vision sensors form tetrahedrons, the four groups of vision sensors can form four tetrahedrons, and the range of the four tetrahedrons covers the hexahedral space in the hexahedral non-blind area detection area, so that the non-blind area detection can be realized, and in addition, the hexahedral space formed by the hexahedral non-blind area detection area is increased and lengthened no matter how the hexahedral space formed by the hexahedral non-blind area detection area is increased; the monitoring device can realize the non-blind area detection of the hexahedral space formed by the hexahedral non-blind area detection area through four groups of visual sensors, so that the detection cost can be reduced while the non-blind area detection is realized.

That is to say: the lens focuses of the four vision sensors and the three-dimensional cone formed by connecting the four corner points of the background references shot by the corresponding vision sensors can be combined to form a continuous hexahedral region formed by connecting eight corner points of two rectangular background references (namely, a background reference ref1 and a background reference ref 2); and the above-mentioned hexahedral non-blind area detection region is a continuous hexahedral region formed by connecting eight corner points of rectangular background references, i.e., background reference ref1 and background reference ref 2); the method for detecting the foreign matters in the hexahedron comprises the steps of comparing images of four sides of a background reference object (namely a background reference object ref1 or a background reference object ref 2) in a central projection area of a visual sensor with differences of the central projection images of the background reference object in the visual sensor to judge whether the foreign matters exist in the continuous hexahedron area; in the plurality of hexahedral non-blind area detection areas, as long as foreign matters are detected in one hexahedral non-blind area detection area, the detection of foreign matters in the area monitored by the non-blind area in the large three-dimensional area is indicated.

As a preferable scheme, the visual sensors in each hexahedral non-blind area detection area are eight groups; the corner points of the background reference object ref1 are A (101), B (102), C (103) and D (104) respectively; the corner points of the background reference object ref2 are E (105), F (106), G (107) and H (108) respectively; the lens focuses of the four groups of vision sensors are respectively positioned on four opposite angles of the background reference object ref1, and the lens focuses of the four groups of vision sensors are respectively positioned on four opposite angles of the background reference object ref 2. When the arrangement mode is adopted, the non-blind area detection can be completely ensured to be carried out on the non-blind area detection area in the hexahedron, the principle is as above, namely, eight groups of visual sensors can form eight tetrahedrons, the composition range of the eight tetrahedrons can ensure that the hexahedron space in the non-blind area detection area in the hexahedron is completely enclosed, thus the non-blind area detection can be realized,

Referring to fig. 1-3, the background reference is configured to emit light or/and is a marker with high color contrast with the surrounding environment, wherein the light emitted by the light-emitting marker is visible light or/and invisible light; the method aims to reduce the dependence degree of foreign matter monitoring on the external environment, so that the robustness and the anti-interference performance of the system are improved, and the detection recognition rate is improved. The background reference in this embodiment is an invisible light emitting surface, including an infrared lamp panel.

Referring to fig. 1 to 3, the control processing apparatus is an AI control processing apparatus; the monitoring device further comprises a detection output device, wherein the detection output device is used for displaying detection results and comprises one or more of an audible and visual alarm device, a display and a background server, and the visual sensor, the control processing device and the detection output device are connected through a wire or/and a wireless connection (such as WiFi).

In this embodiment, the plurality of hexahedral non-blind area detection areas may share one or more sets of AI control processing devices and detection output devices, and perform image processing and alarm output, respectively.

Referring to fig. 1-4, the method for monitoring the foreign matters in the large three-dimensional area based on the machine vision comprises the steps of firstly dividing the large three-dimensional area to be monitored into one or more hexahedral non-blind area detection areas, and arranging a set of monitoring device in each hexahedral non-blind area detection area; wherein the foreign matter detection includes the steps of:

S1, detecting and collecting image information of a non-blind area detection area in a hexahedron through vision sensors, wherein two groups of vision sensors positioned on opposite angles of a background reference object ref1 are used for shooting images of the background reference object ref2, and two groups of vision sensors positioned on opposite angles of the background reference object ref2 are used for shooting images of the background reference object ref1 and transmitting the image information to control processing equipment;

S2, the control processing equipment processes the image information collected in each hexahedral non-blind area detection area, determines the position of a background reference object in the image, identifies the boundary condition of the background reference object image, and retains the image information of the background reference object without foreign matters for information comparison;

S3, controlling processing equipment to detect a change value between a real-time image of a background reference object in a detection area without blind areas in the hexahedron and an image in a boundary of the background reference object when the background reference object is not blocked by foreign objects, if the change value is larger than a set threshold value, judging that the foreign objects invade the detection area without blind areas in the hexahedron, and entering step S4; otherwise, enter step S5;

S4, the control processing equipment outputs the processed information to the detection output equipment, and an image of a foreign object position is displayed in the display for manually confirming whether the false alarm is generated or not, and meanwhile, the audible and visual alarm device sends out alarm sounds to inform the invasion position of the foreign object; simultaneously starting a video recording function, and storing live videos and partial images; then re-entering step S3;

And S5, the control processing equipment outputs the processed information to the detection output equipment, the display displays the image information of the detection area, if the detection area is too many and can not be displayed completely, the display is performed in a carousel mode, and then the step S3 is restarted.

Example 1

The embodiment mainly describes a specific embodiment of the monitoring device in a single hexahedral non-blind area detection area, and the surface of the hexahedral non-blind area detection area for installing the background reference object is rectangular with equal size, namely, rectangle, and is hereinafter called as a "hexahedral non-blind area detection area" as a "cuboid non-blind area detection area".

The application scenario of the embodiment mainly includes cuboid spaces requiring non-blind area real-time intrusion detection or other spaces which can be approximately covered by a cuboid, such as long channels, corridor, rail traffic platform gap, museums and other cuboid spaces requiring monitoring. The machine vision-based long and large three-dimensional area non-blind area foreign matter monitoring device is arranged in the three-dimensional space to perform non-blind area detection. In the long and large three-dimensional area non-blind area foreign matter monitoring device based on machine vision, the background reference adopts the rectangular LED infrared lamp panel, and infrared light generated by the LED infrared lamp panel is invisible to human eyes, so that the device has concealment and does not generate light pollution; the vision sensor adopts an infrared camera, so that the difference between the captured image of the LED infrared lamp panel and the environment is obvious.

The following description is made with reference to the accompanying drawings:

As shown in fig. 1-3, a rectangular background reference object ref1 and a rectangular background reference object ref2 are arranged according to a cuboid detection area to be detected; because the system can monitor longer distances, the background reference object is generally arranged on two planes which are far apart in a cuboid space. The four corner points of the background reference object ref1 are 101, 102, 103 and 104, and a visual sensor sen1 and a visual sensor sen2 are arranged on the corner points 101 and 104, and the lens of the visual sensor sen1 and the visual sensor sen2 shoots the background reference object ref2. The four corner points of the background reference object ref2 are 105, 106, 107 and 108, respectively, and a visual sensor sen3 and a visual sensor sen4 are arranged on the corner points 106 and 107, and the lens of the visual sensor sen4 shoots the background reference object ref1. Each vision sensor can obtain an image of a background reference object with clear opposite sides, wherein the sight coverage range of the vision sensor is as shown in fig. 2, so that no detection blind area in a rectangular detection area can be ensured. The control processing equipment is an embedded microprocessor control board for processing and identifying the image, and is connected with each visual sensor and each detection output equipment, so that a non-blind area cuboid monitoring system is obtained.

The long and large three-dimensional area non-blind area foreign matter monitoring device based on machine vision performs the following detection steps during foreign matter detection:

(1) Acquiring image information of the cuboid detection area through a vision sensor, capturing images of a background reference object ref1 and a background reference object ref2, and transmitting the image information to a control processing device;

(2) The control processing equipment processes the collected images in a concentrated mode, positions of a background reference object ref1 and a background reference object ref2 in the images are determined, the background reference object ref1 and the background reference object ref2 are adaptively focused, the conditions of the background reference object ref1 and the background reference object ref2 in a cuboid detection area are confirmed, and image information of the background reference object without foreign matters is reserved for information comparison;

(3) The control processing equipment transmits the real-time image of the cuboid detection area to a display in the detection output equipment;

(4) Detecting a change value between a real-time image of a background reference object ref1 and a real-time image of a background reference object ref2 in a rectangular detection area and a complete image of the background reference object ref1 and the background reference object ref2 when the real-time image is not blocked by foreign objects, judging that the foreign objects exist if the change value is larger than a threshold value set by a user, and entering a step (5); otherwise, entering step (6);

(5) The control processing equipment outputs the processed information to the detection output equipment, a foreign object image is displayed in the display, and meanwhile, the audible and visual alarm device displays the foreign object and gives out alarm sound; meanwhile, a video recording function is started, on-site videos and partial images are stored, and evidence collection is facilitated;

(6) And the control processing equipment outputs the processed information to the detection output equipment, the display displays normal information such as images of the cuboid detection area, and the audible and visual alarm does not give an alarm.

The main characteristics of the embodiment are as follows:

The long three-dimensional area non-blind area foreign matter monitoring device based on machine vision can realize three-dimensional real-time monitoring of long-distance continuous long areas, but not boundary monitoring; the four vision sensors of the long and large three-dimensional area non-blind area foreign matter monitoring device based on machine vision are mutually covered in view, when one or more vision sensors are damaged, other vision sensors can capture related images of the damage vision sensors of an invader, and a certain deterrent force is provided, so that the responsibility following is realized. The long and large three-dimensional area non-blind area foreign matter monitoring device based on machine vision provided by the invention has no monitoring blind area in a continuous long and large three-dimensional monitoring area; the real-time monitoring picture can be displayed, the monitoring video is reserved, and excellent visibility is achieved; the background reference object and the vision sensor are respectively selected from an infrared LED lamp panel and an infrared camera, so that the information of the background reference object (namely the infrared LED infrared lamp panel) can be effectively extracted in a complex light environment, and the stability of the vision system in complex environment light is improved; meanwhile, the infrared light emitted by the infrared LED infrared lamp panel is invisible to human eyes, so that the infrared LED infrared lamp panel has concealment and cannot cause light pollution to the environment. In addition, no matter how the hexahedral space formed by the non-blind area detection area in the hexahedron is increased, the hexahedron is lengthened, or the hexahedron is a rhombic hexahedron; the monitoring device can realize the non-blind area detection of the hexahedral space formed by the hexahedral non-blind area detection area through four groups of visual sensors, so that the detection cost can be reduced while the non-blind area detection is realized, and the monitoring device has higher flexibility.

Example 2

The embodiment mainly describes a specific embodiment of the monitoring device in the plurality of hexahedral non-blind area detection areas, and the surface of each hexahedral non-blind area detection area for installing the background reference object is rectangular, which is hereinafter referred to as a "hexahedral non-blind area detection area" as a "cuboid non-blind area detection area".

The application scenario of the embodiment mainly includes a space which needs no blind area real-time intrusion detection and cannot be covered by a single rectangle or an irregular space, such as a circular corridor, an arc area space, a triangle space and the like which need to be monitored. Any one curve can be approximated by an infinite number of increments, according to the principle of integration. Theoretically, for an irregular solid region of limited volume, it is also possible to perform infinite approximation for the solid region with a plurality of rectangular solids. Because the volume of the equipment is not infinitesimal, 100% approximation cannot be realized, but effective monitoring of most areas can be realized.

In the embodiment 2, i large three-dimensional area non-blind area foreign matter monitoring devices based on machine vision are used for splicing, so that the irregular three-dimensional area is monitored.

The following description is made with reference to the accompanying drawings: the machine vision-based large three-dimensional area non-blind area foreign matter monitoring device is arranged in the large three-dimensional space to perform non-blind area detection. In the long and large three-dimensional area non-blind area foreign matter monitoring device based on machine vision, a background reference object adopts a rectangular LED infrared lamp panel, the LED infrared lamp panel can be obviously distinguished from the surrounding environment, and a vision sensor adopts an infrared camera, so that the difference between a captured LED infrared lamp panel image and the environment is obvious. By splicing the i large-sized three-dimensional area non-blind area foreign matter monitoring devices based on machine vision, the irregular three-dimensional area is monitored, so that the non-detection blind area in the detection area of each large-sized three-dimensional area non-blind area foreign matter monitoring device based on machine vision is realized, and the detection area is shown in a figure 4.

According to the irregular space to be monitored, a plurality of non-blind zone cuboid monitoring devices are designed, and background references should be placed on planes far apart in a three-dimensional area preferentially during design. The large three-dimensional area non-blind area foreign matter monitoring device based on machine vision is installed in a mode of embodiment 1. For the nth long three-dimensional area non-blind area foreign matter monitoring device (n E [0, i ]) based on machine vision, the specific installation mode is as follows: setting up a background reference object ref1n, wherein four corner points are respectively 101n, 102n, 103n and 104n, and a visual sensor sen1n and a visual sensor sen2n are arranged on the corner point 101n and the corner point 104n, and the lens of the visual sensor sen1n shoots the background reference object ref2n. The background reference ref2n is set up, four corner points are 105n, 106n, 107n and 108n, a visual sensor sen3n and a visual sensor sen4n are arranged on the corner point 106n and the corner point 107n, and the lens of the visual sensor sen4n shoots the background reference ref1n. The foreign matter detection period of the machine vision-based large three-dimensional area non-blind area foreign matter monitoring according to the present embodiment is implemented according to the foreign matter detection step of embodiment 1, as shown in fig. 4.

As an embodiment, the plurality of the large three-dimensional area non-blind area foreign matter monitoring devices may share one or more control processing apparatuses;

as another embodiment, the multiple large and large three-dimensional area non-blind area foreign matter monitoring devices may share one or more detection output devices, and when videos of all detection areas cannot be displayed at the same time, a carousel mode may be used;

The main characteristics of the embodiment are as follows: by splicing the plurality of the long and large three-dimensional area non-blind area foreign matter monitoring devices, the long and large three-dimensional area non-blind area foreign matter monitoring device is not limited to the cuboid detection area any more, and the non-blind area detection of the detection area with any shape can be realized, so that the adaptability of the system is improved.

The above is only a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and the visual sensor positions as in embodiment 1 and embodiment 2 may be provided with additional diagonal points, such as visual sensor sen1, visual sensor sen2, visual sensor sen3, visual sensor sen4 may be provided at the corner points 102, 103, 105, 108, respectively. For reliability and redundancy, visual sensors may also be provided at all corner locations.

It should be noted that, for the technical field, several improvements and modifications can be made, such as using two opposite background references, realizing the detection without blind areas by using four visual sensors or more, or modifying the shape of the background references, such as parallelogram, diamond, trapezoid, circle, hexagon, trapezoid, etc. in case of allowing a certain detection blind area, and realizing a larger detection coverage by diagonally installing the visual sensors, so as to achieve the purpose of better fitting the required detection space, which should also be regarded as the protection scope of the present invention.

Claims (8)

1. The device is characterized by comprising a monitoring device arranged in a large three-dimensional area non-blind area monitoring area, wherein the large three-dimensional area non-blind area monitoring area is formed by combining one or more hexahedral non-blind area detection areas, two opposite surfaces of the hexahedron are parallel rectangles, the sizes of the hexahedrons are equal, a set of monitoring device is arranged in each hexahedral non-blind area detection area, the monitoring device comprises a detection device, a detection output device and a control processing device, the detection device is used for detecting whether the foreign object exists in each hexahedral non-blind area detection area, and the detection device comprises background references and a visual sensor, wherein the background references are two groups, and the two groups of background references are respectively positioned on the two opposite parallel rectangular surfaces of each hexahedral non-blind area detection area; the length and width of each set of background references are equal to the length and width of the face, respectively; the two groups of background references are respectively a background reference ref1 and a background reference ref2; the visual sensors are at least four groups, and lens focuses of two groups of visual sensors are positioned on two diagonal points of the background reference object ref 1; the lens focuses of the other two groups of vision sensors are positioned on two opposite angles of the background reference ref2, wherein the connecting lines of the lens focuses of the vision sensors positioned on the two pairs of corner points of the background reference ref1 and the connecting lines of the lens focuses of the vision sensors positioned on the two pairs of corner points of the background reference ref2 are intersected on the vertical projection line of the background reference ref1 or the background reference ref2; wherein, the vision sensor located on the diagonal point of the background reference object ref1 is used for acquiring the image of the background reference object ref2, and the vision sensor located on the diagonal point of the background reference object ref2 is used for acquiring the image of the background reference object ref 1; the control processing equipment is used for receiving the image data returned by the visual sensor, judging whether foreign matters enter the detection area without blind areas in the hexahedron through processing and comparing the image information in the areas of the background reference object ref1 and the background reference object ref2, and controlling the alarm and result output of the detection output equipment when detecting that the foreign matters enter the detection area without blind areas in the hexahedron.

2. The machine vision-based long and large three-dimensional area non-blind area foreign matter monitoring device according to claim 1, wherein the number of the vision sensors in each hexahedral non-blind area detection area is four, namely a vision sensor sen1, a vision sensor sen2, a vision sensor sen3 and a vision sensor sen4; the corner points of the background reference object ref1 are A (101), B (102), C (103) and D (104) respectively; the corner points of the background reference object ref2 are E (105), F (106), G (107) and H (108) respectively; wherein lens foci of the visual sensor sen1 and the visual sensor sen2 are located at the corner a (101) and the corner D (104) of the background reference ref1, respectively, and lens foci of the visual sensor sen3 and the visual sensor sen4 are located at the corner F (106) and the corner G (107) of the background reference ref1, respectively; or the lens focuses of the vision sensor sen1 and the vision sensor sen2 are respectively located at the corner B (102) and the corner C (103) of the background reference ref1, and the lens focuses of the vision sensor sen3 and the vision sensor sen4 are respectively located at the corner E (105) and the corner H (108) of the background reference ref 1.

3. The machine vision-based long and large three-dimensional area non-blind area foreign matter monitoring device of claim 1, wherein the visual sensors in each hexahedral non-blind area detection area are eight groups; the corner points of the background reference object ref1 are A (101), B (102), C (103) and D (104) respectively; the corner points of the background reference object ref2 are E (105), F (106), G (107) and H (108) respectively; the lens focuses of the four groups of vision sensors are respectively positioned on four opposite angles of the background reference object ref1, and the lens focuses of the four groups of vision sensors are respectively positioned on four opposite angles of the background reference object ref 2.

4. The machine vision-based large three-dimensional area non-blind area foreign matter monitoring device of claim 1, wherein the background reference is set to emit light or/and is a marker with high color contrast with the surrounding environment, wherein the light emitted by the light-emitting marker is visible light or/and invisible light.

5. The machine vision-based long and large three-dimensional area non-blind area foreign matter monitoring device according to claim 1, wherein the detection output device is configured to display a detection result, and the detection output device comprises one or more of an audible and visual alarm device, a display and a background server.

6. The machine vision-based long and large three-dimensional area non-blind area foreign matter monitoring device of claim 5, wherein one or more sets of control processing equipment and detection output equipment are shared among a plurality of sets of monitoring devices in the plurality of hexahedral non-blind area detection areas.

7. The machine vision-based large three-dimensional area non-blind area foreign matter monitoring device according to claim 1, wherein each group of vision sensors is a monocular camera or a binocular camera, and a visible light camera and/or an infrared camera are/is adopted.

8. A machine vision based long and large three-dimensional area non-blind area foreign matter monitoring method of the machine vision based long and large three-dimensional area non-blind area foreign matter monitoring device according to any one of claims 1 to 7, characterized in that the long and large three-dimensional area to be monitored is divided into one or more hexahedral inner non-blind area detection areas, and a set of monitoring devices is arranged in each hexahedral inner non-blind area detection area; wherein the foreign matter detection includes the steps of:

S1, detecting and collecting image information of a non-blind area detection area in a hexahedron through vision sensors, wherein two groups of vision sensors positioned on opposite angles of a background reference object ref1 are used for shooting images of the background reference object ref2, and two groups of vision sensors positioned on opposite angles of the background reference object ref2 are used for shooting images of the background reference object ref1 and transmitting the image information to control processing equipment;

S2, the control processing equipment processes the image information collected in each hexahedral non-blind area detection area, determines the position of a background reference object in the image, identifies the boundary condition of the background reference object image, and retains the image information of the background reference object without foreign matters for information comparison;

S3, controlling processing equipment to detect a change value between a real-time image of a background reference object in a detection area without blind areas in the hexahedron and an image in a boundary of the background reference object when the background reference object is not blocked by foreign objects, if the change value is larger than a set threshold value, judging that the foreign objects invade the detection area without blind areas in the hexahedron, and entering step S4; otherwise, enter step S5;

S4, the control processing equipment outputs the processed information to the detection output equipment, and an image of a foreign object position is displayed in the display for manually confirming whether the false alarm is generated or not, and meanwhile, the audible and visual alarm device sends out alarm sounds to inform the invasion position of the foreign object; simultaneously starting a video recording function, and storing live videos and partial images; then re-entering step S3;

And S5, the control processing equipment outputs the processed information to the detection output equipment, the display displays the image information of the detection area, if the detection area is too many and can not be displayed completely, the display is performed in a carousel mode, and then the step S3 is restarted.