JP6389087B2 - Boom collision avoidance device for work equipment - Google Patents

Description

  The present invention relates to a boom collision avoidance device for a work machine that is used in a work machine having a work boom and prevents a boom from colliding with an obstacle in order to prevent the boom from interfering with the obstacle during work.

  Some working machines having a working boom include a device that restricts the position and posture of the boom in order to prevent the boom from interfering with an obstacle during work. For example, the height limiting device described in Patent Document 1 geometrically positions the crane boom based on detection values such as a boom angle and a boom length in order to prevent the boom from interfering with an obstacle located above during operation. The control is performed so that the boom position does not exceed the designated work regulation range.

JP 2003-267666 A

  However, in the technique described in Patent Document 1, since the position of the boom of the crane is calculated geometrically, in order to set the work regulation range, for example, it is suspended between two power poles that become obstacles. Distance data from the obstacles such as the electric wire, such as how far the electric wire stands from the top of the crane boom, in which direction it is suspended, and how high it is (Environmental data) must be accurate.

  In addition, the height of the top of the boom changes each time depending on the extension of the crane's outrigger. If the position information is not grasped, the relative position of the boom top of the crane with respect to the obstacle such as an electric wire cannot be grasped and the operation is not performed accurately. Therefore, there is a problem that it is difficult to set a complicated work regulation range. In addition, when the work environment data is an input error so-called “poca mistake” or the work environment data itself is incorrect information, the height limit setting value is set to a position higher than the obstacle, The boom may collide with an obstacle.

  Here, the technique described in Patent Document 1 geometrically calculates the position of the boom with respect to the obstacle. However, in order to avoid the obstacle, as another measure that does not input the work environment data, It is also possible to use sound waves or laser scanning. For example, in the case of using ultrasonic waves, the distance from the time until the ultrasonic waves are transmitted and reflected and returned to the obstacle can be measured. Further, for example, in the case of laser scanning, the distance to the obstacle can be measured by irradiating the surroundings with a laser and how to shift the irradiation point.

However, when using an ultrasonic sensor, if the ultrasonic directivity is wide, pinpoint measurement cannot be performed and the detection accuracy becomes rough. On the other hand, if the ultrasonic directivity is narrow, a wide range is measured. Has the problem of requiring a large amount of sensors. In the case of laser scanning, if the obstacle is easy to absorb light, such as black, it becomes difficult to detect, and it is necessary to increase the laser intensity to perform measurement under sunlight irradiation conditions. Therefore, there is a problem that it is dangerous when there is a person in the vicinity.
Therefore, the present invention has been made in view of such problems, and it is not necessary to set an artificial work regulation range, and there is a person in the vicinity of a complex-shaped obstacle quickly and nearby. It is another object of the present invention to provide a boom collision avoidance device for work implements that can start work safely.

  In order to solve the above problems, a boom collision avoidance device for a working machine according to an aspect of the present invention is used in a working machine having a boom, and the boom is configured to prevent the boom from interfering with an obstacle during work. A boom collision avoidance device for a work machine for avoiding a collision with an obstacle, wherein the video camera is provided at the tip of the boom so as to photograph the boom and the background, and the boom of the boom photographed by the video camera Or an image processing unit that tracks a natural feature point of an obstacle existing in an image when a turning motion is performed, and generates information on an obstacle that may collide with the boom based on the tracking result of the natural feature point; And a collision avoidance control unit that performs collision avoidance control for avoiding a collision between the boom and the obstacle based on the information on the obstacle generated by the image processing unit.

“Natural feature points” refer to points at which the corners and contours of an object in a photographed image have been extracted from characteristic portions where the contrast or hue difference is clear with respect to the background.
Here, in the boom collision avoidance device for a work machine according to an aspect of the present invention, the image processing unit performs a background based on a tracking result of the natural feature point when the boom is raised or lowered. Recognizing an object that moves differently from the movement of the natural feature point as an obstacle, calculating an estimated position of the natural feature point of the obstacle after a certain period of time as the obstacle information, and the collision avoidance control unit Based on the information on the estimated position, it is preferable to perform collision avoidance control such as warning to the operator or boom deceleration, stop, or restriction of the operation direction.

  Further, in the boom collision avoidance device for work implements according to one aspect of the present invention, the image processing unit may cause a collision with the boom after a predetermined time based on the estimated position of the natural feature point of the obstacle. Classification processing to classify obstacles according to the degree of danger of contacting the boom with dangerous obstacles that are judged to collide with the boom after a certain period of time and non-contact obstacles that are not likely to collide with the boom after a certain period of time The collision avoidance control unit continues normal boom operation when determined as the non-contact obstacle based on the result of the classification process, and when determined as the caution obstacle, It is preferable to output a boom operation stop signal in the direction of approaching the obstacle when an alarm is given and the danger obstacle is determined.

  Further, in the boom collision avoidance device for work implements according to an aspect of the present invention, the image processing unit tracks a natural feature point of an object captured by the video camera when the boom is raised or lowered. It is preferable to have a natural feature point tracking means for obtaining the moving direction and the moving speed thereof, and an obstacle extracting means for extracting a natural feature point belonging to the obstacle from the natural feature points tracked by the natural feature point tracking means. .

The obstacle extracting means is configured to detect a natural feature point of the object located on the front side of the turning center axis with respect to the turning center axis of the boom, and a natural feature point of the object located on the back side of the turning center axis. It is determined whether it is a natural feature point that exists in the foreground or a natural feature point that exists in the back from the difference in motion from the feature point, and the natural feature point that is determined to be in front of the turning center axis is determined to be an obstacle It is preferable to extract as a natural feature point belonging to.
Further, the natural feature point tracking means searches for a natural feature point of an object that can be tracked from the current camera image, and then searches for the natural feature point that can be tracked between the camera image of the previous sample and the current camera image. It is preferable to obtain the moving speed of the natural feature point by performing the association and further dividing the moving distance of the natural feature point on the camera image by one sample time.

  According to the boom collision avoidance device for a work machine according to an aspect of the present invention, the configuration includes a video camera and an image processing unit, and performs collision avoidance control based on image processing. Whether or not there is a possibility of collision can be determined based on the image information. Since the video camera and the image processing unit function as a distance sensor, it is possible to eliminate human error such as a mistake in setting the work regulation range. In addition, since there is no need for prior settings for inputting work environment data, the work can be started quickly and safely.

Further, since the collision determination is based on image processing, image information can be acquired even if the obstacle has a complicated shape, and even a small obstacle can be detected. In addition, since the video camera can shoot only with ambient light, it is safer than the case of laser scanning.
Moreover, since it is easy to recognize an obstacle other than glass or a mirror surface, it can be applied with high versatility to the situation at the work site. For example, as a situation of the work site that can be used of the boom collision avoidance device for a work machine according to the present invention, if the work machine is a crane, when carrying out a load work near a utility pole in an urban area or in a track, Contact between the boom and the utility pole can be prevented. Further, for example, when carrying out a hanging work at a construction site, it is possible to carry in a load while avoiding contact between the boom and the scaffold.

It is a typical perspective view explaining the vehicle-mounted crane which is one Embodiment of the working machine carrying the boom collision avoidance device for working machines which concerns on 1 aspect of this invention. It is a flowchart of the collision avoidance process (main process) which a boom collision avoidance apparatus performs. It is a flowchart of the obstacle detection process which a boom collision avoidance apparatus performs in the collision avoidance process. It is the typical explanatory view which looked down at the crane main part from the upper part. It is explanatory drawing of the camera image which looked at the boom and column of the crane main body from the visual field of a video camera (lower side), when performing the turning operation | movement of a boom. FIG. 6 is a diagram corresponding to FIG. 5, illustrating the natural feature point tracking process and the obstacle extraction process in FIG. 3. FIG. 6 is a diagram corresponding to FIG. 5, illustrating the natural feature point tracking process and the obstacle extraction process in FIG. 3.

Hereinafter, a vehicle-mounted crane that is an embodiment of a work machine equipped with a boom collision avoidance device for a work machine according to an aspect of the present invention will be described with reference to the drawings as appropriate. This embodiment will explain an example of preventing contact (collision) between a boom and a power pole when a load-carrying work is performed with a vehicle-mounted crane near a power pole in an urban area or a railway track.
As shown in FIG. 1, this vehicle-mounted crane (hereinafter, also simply referred to as “crane body”) 1 is mounted between a driver's seat 9 a and a loading platform 9 b of a vehicle 9 such as a truck, and the chassis of the vehicle 9. The base 2 is fixed on a frame (not shown). A pair of outriggers 8 that can project in the vehicle width direction are provided on the left and right sides of the base 2. A column 3 is erected on the base 2 at a position in the center of the vehicle width so as to be turnable. A boom 4 is pivotally supported at the upper end of the column 3 so as to be able to expand and contract. The column 3 is provided with a winch (not shown), and the wire rope 5 from the winch is hung on the hook 6 of the tip 4t of the boom 4.

  The crane 1 is provided with a plurality of actuators (not shown) corresponding to the boom 4 undulation, expansion / contraction, turning, and driving of the winch. The plurality of actuators are provided in the vicinity of the base 2 and are driven by the control of the controller 30 that controls the boom operation, so that the boom operation including undulation, expansion and contraction, and turning operation is possible.

When crane operation is performed, when an operator inputs a desired crane operation from a position away from the crane body 1 by operating each selection switch and speed lever of a remote controller (not shown), a signal corresponding to the desired crane operation is obtained. Is input to the controller 30 by wireless communication, and the controller 30 outputs an operation command for operating each actuator of the crane body 1 to start the operation of a desired actuator.
Here, in the crane body 1, two video cameras 10 and an image processing device 20 are attached to the tip 4 t of the boom 4. And the said controller 30 has a function as a collision avoidance control part, and can perform a collision avoidance process based on the estimated position (after-mentioned) of the obstruction output from the image processing apparatus 20 which is an image processing part. It is configured.

  Specifically, as shown in FIG. 1, the two video cameras 10 are provided on the lower side and the upper side of the boom tip 4 t so as to be able to shoot including the column 3 of the crane body 1 and the background. . The image processing apparatus 20 captures a camera image taken by the video camera 10 and executes an obstacle detection processing program described later to perform an object that exists in the camera image when the boom 4 is turned or undulated. When a target object that moves differently from the background motion is recognized based on the tracking result of the natural feature point, the target is recognized as an obstacle, and the obstacle to the boom 4 is detected. When an approach is detected, the obstacle information is sent to the controller 30. In the example of the present embodiment, the work is performed between a plurality of electric poles H (5 poles A to E for explanation in the example shown in FIG. 4) that are provided separately. This is an example in which each power pole H is an object of an obstacle.

Then, the controller 30 as the collision avoidance control unit executes a program for collision avoidance processing, based on the result of the estimated position of the obstacle with respect to the boom 4 after a predetermined time, an alarm to the operator, or deceleration of the crane, Collision avoidance control such as stopping or regulating the operation direction is performed.
Specifically, when the collision avoidance process is executed by the collision avoidance control unit of the controller 30, as shown in FIG. 2, first, the process proceeds to step S <b> 1, and the preset alarm grace time, stop grace time, image The processing mask is read and initialization is performed.

  Here, the “warning grace time” is a reference time used for determining that it is necessary to call attention when the detected obstacle may approach or collide with the boom 4 after a certain time has elapsed. The estimated position of the obstacle after the lapse of the “warning grace time” is set to issue an alarm when there is a possibility of collision with the boom 4. The “stop grace time” is a reference time used to determine that the detected obstacle collides with the boom 4 after a predetermined time has elapsed, and the estimated position of the obstacle after the lapse of the “stop grace time” is This is set to stop the crane when passing through the boom 4. Further, the “image processing mask” is a mask (range) for separating a portion where image processing is performed from a portion where image processing is not performed (see FIG. 6). A “crane collision mask” described later is a mask (range) for determining contact (collision) between a crane and an obstacle (see FIG. 7).

In the subsequent step S2, the image processing apparatus 20 acquires an image from the video camera 10 and sets it as the first reference image (reference image acquisition). In the subsequent step S3, the image processing apparatus 20 performs FIG. The “obstacle detection process” is executed, information related to the detection of the obstacle including information on the estimated position of the obstacle is generated by a series of image processing, and the process proceeds to step S4.
Then, the controller 30 performs branch processing after step S4 on the estimated position of the natural feature point of the obstacle in the image processing apparatus 20 of step S3, that is, classification of a caution obstacle, a dangerous obstacle, and a non-contact obstacle. Collision avoidance control is performed based on the processing result.

  Here, the “attention obstacle” refers to a case where the estimated position of the natural feature point related to the obstacle after the passage of the “warning grace time” overlaps with the mapping of the boom 4 captured by the video camera 10, that is, An obstacle that may collide with the boom 4 after the warning grace period has elapsed. In addition, the “dangerous obstacle” means that the estimated position of the natural feature point related to the obstacle after the passage of the “stop grace time” passes through the mapping area of the boom 4 captured by the video camera 10. In other words, an obstacle that is judged to collide with the boom 4 after the stop grace time has elapsed. In addition, the “non-contact obstacle” refers to an obstacle that does not meet the conditions of the “caution obstacle” and the “dangerous obstacle” and does not collide with the boom 4.

  Specifically, when the image processing apparatus 20 proceeds to step S3, the obstacle detection processing program shown in FIG. 3 is executed. First, the image processing apparatus 20 proceeds to step S31, and two videos provided above and below the boom tip 4t. The boom 4 and background images captured by the camera 10 are acquired as camera images, and the process proceeds to step S32. In step S32, as pre-processing for natural feature point tracking, filtering processing such as smoothing processing for noise removal and histogram flattening is performed on the camera image acquired in step S31.

Here, an image with a biased histogram is an “image in which a portion not exposed to light is crushed in black” or “an image in which a portion that is strongly exposed to light is white”, but the histogram is flattened. As a result, the image of the portion that has been crushed in black or the portion that has been blown out in white becomes clear, and the obstacle can be detected more reliably.
In addition, noise occurs in images taken with a video camera due to various factors. When an image containing noise is processed as it is, it is possible to misrecognize the noise as an obstacle even though there is no obstacle in the shooting range. High nature. Although it is possible to separate obstacles and noise by image processing using information such as the size of the projected area of the object on the image, if a large amount of noise is determined by image processing, the time required for noise removal will be enormous . Therefore, before performing the obstacle determination, smoothing processing is performed in advance to remove noise.

  In subsequent step S33, as pre-processing for tracking natural feature points, masking processing is performed to remove data of portions that do not require processing, such as a portion where the boom 4 is present, from the captured camera image. FIG. 6 is an image image of the video camera 10 below the boom tip 4t, and is an example in which the boom periphery is set as an image processing mask unit. In the figure, the masking process superimposes the camera image filtered in step S32 and the image processing mask, and sets the pixel value of the portion overlapping the image processing mask to 0.

  In step S34, a natural feature point tracking process is performed when the boom 4 is raised or lowered, and the moving direction and moving speed of the object captured by the video camera 10 are obtained. In the natural feature point tracking process, first, a natural feature point of an object that can be tracked is searched from the current camera image after masking in step S33. Next, a process for associating the camera image of the previous sample with the current camera image that can be tracked (tracking how each feature point has moved from the camera image of the previous sample) is performed. Further, the moving speed of the natural feature point is obtained by dividing the moving distance of the natural feature point on the camera image by one sample time. Thereby, it is recognized in which direction and at what speed the natural feature point on the screen is moving.

  In the subsequent step S35, an obstacle extraction process for extracting natural feature points belonging to the obstacle from the natural feature points tracked in step S34 is executed. This obstacle extraction processing is performed with respect to the object located on the near side (camera side) of the turning center axis with respect to the turning center axis of the boom 4 shown in FIG. Determine whether it is a natural feature point that is in front (obstacle) or a natural feature point in the back (background) from the difference in motion with the object located on the side) A natural feature point determined to exist in front is extracted as a natural feature point belonging to an obstacle.

As an example of the natural feature point tracking, consider a case where the right turn operation of the boom 4 shown in FIG. 4 is performed. FIG. 5 shows an image of the camera image of the video camera 10 when the boom is turned to the right. In the example shown in FIG. 5, an image is shown when the pillar indicated by a broken line on the camera image is recognized as an object so as to be positioned on the solid line by turning the boom 4.
When the boom 4 is turned right now, as shown in FIG. 5, the object closer to the video camera 10 than the turning center axis moves in the right direction as the turning direction on the camera image. In addition, the arrow shown to the figure has shown the image of the moving direction of the target object on a camera image (following, it is the same also in FIGS. 5-7). On the other hand, the object and the background farther from the video camera 10 than the turning center axis move to the left, which is the opposite direction to the turning direction on the camera image. Therefore, if the turning direction of the boom 4 is known, the natural feature point of the background and the natural feature point of the obstacle can be easily distinguished. For example, the movement of the control lever of the crane or the movement of the control valve turning spool and hoisting spool is sensed by a sensor such as a differential transformer, and the controller 30 receives the information to determine the operating direction of the boom 4. Whether it is a background or an obstacle can be read from the vector direction of the natural feature points on the screen.

  In this embodiment, the operation direction of the boom 4 is determined by the controller 30 by reading the movement of the operation lever of the crane with a limit switch. Then, the controller 30 recognizes the natural feature point moving in the same direction on the screen as the natural feature point based on the background with respect to the actual boom operation direction, while moving the natural feature point in the direction opposite to the boom operation direction. The point is recognized as a natural feature point of the obstacle.

  As another determination processing method, it is possible to compare and determine the difference in the number of natural feature points regardless of the turning direction of the boom 4. That is, normally, the number of natural feature points extracted as a background is overwhelmingly larger than the number of natural feature points extracted as an obstacle, so that the controller 30 has one natural feature point during boom operation. When moving in the direction, it is possible to determine the one with the larger absolute number as the “background” and the smaller one as the obstacle.

  Here, obstacles that may collide with the boom 4 when the boom 4 is turned are present in the space up to the turning center axis of the crane body 1 in the range captured by the video camera 10 image. It is a target object. In addition, when the boom 4 is turned to the right, there is a possibility that an object existing on the right side of the boom 4 collides with the boom 4 (see an area indicated by a two-dot chain line in FIG. 4).

  In the example shown in FIG. 5, the left half of the screen on the camera image corresponds to this. On the other hand, objects existing below the boom 4 and on the right half of the screen do not have a possibility of colliding with the boom 4. In other words, when the right turn operation of the boom 4 is performed, an object that is an object on the left side of the screen on the camera image and moves in the right direction on the camera image may collide with the boom. It can be determined that it is a thing.

  Now, as shown in FIGS. 4 and 5, consider a case where there are five pillars A to E as objects around the crane body 1. In this case, when the boom 4 turns right, the pillars A and B are recognized as the background because the natural feature points of the pillars A and B move in the left direction on the camera image, and there is no possibility of colliding with the boom 4. Can be determined. In addition, since the natural feature point moves to the right in the right half of the screen, the pillar E can be determined as an object that has no possibility of colliding with the boom. The pillars C and D may have a possibility of colliding with the boom 4 because their natural feature points move to the right in the left half of the screen. However, since the natural feature point of the column D is located below the region of the boom 4, it can be determined that there is no possibility of colliding with the boom 4. On the other hand, since the natural feature point of the pillar C is located above the region of the boom 4, it can be determined as an obstacle that may collide with the boom 4. This obstacle determination method by tracking the natural feature points can be applied to the case where the boom 4 is turned to the left or the undulation operation is performed based on the same concept.

  Further, from the moving speed of the object on the camera image, the estimated position of the object relative to the turning center axis of the crane body 1 and the video camera 10 after a predetermined time can be obtained by calculation. That is, in the subsequent step S36 shown in FIG. 3, the estimated position after the alarm delay time (position corresponding to the caution obstacle) and the estimated position after the stop delay time (dangerous) of the object determined as an obstacle in the above determination. The position corresponding to the obstacle is calculated. The obstacle estimated position calculation is performed on all natural feature points belonging to the obstacle extracted in step S35.

As shown in FIG. 7, the estimated position of the obstacle after a predetermined time can be obtained by adding a velocity vector to the current position. Specifically, the estimated position of an obstacle (attention obstacle, dangerous obstacle) is obtained by the following equation.
Speed vector = (current position-position before 1 sample) / time required for one sample Attention obstacle estimated position = current position + speed vector x alarm delay time Danger obstacle estimated position = current position + speed vector x stop delay time continued In step S37, obstacle classification processing is performed. The obstacle classification process is a process of classifying the natural feature points tracked in step S35 according to the risk of contact with the boom 4 based on the calculation result of step S36. Here, the natural feature points belonging to the obstacle move to the attention obstacle estimated position after the alarm grace time.

  That is, when the estimated position of the cautionary obstacle overlaps with the mapping of the boom 4 captured by the video camera 10, it can be determined that this obstacle may collide with the boom 4 after the lapse of the alarm grace time. In addition, even when the estimated position passes through the mapping of the boom 4 captured by the video camera 10, it can be determined that the obstacle collides with the crane within the alarm delay time. From this, it is determined whether or not the vector connecting the current position of the obstacle and the estimated position of the caution obstacle intersects the area of the mapping of the boom 4 captured by the video camera 10, that is, the crane collision mask shown in FIG. By calculating, it is possible to determine the possibility that the obstacle will collide with the boom 4 within the alarm grace period. The same applies to the estimated dangerous obstacle position. In step S37, with respect to the obstacle tracked in step S35 using this determination method, the following three obstacles (dangerous obstacle, attention obstacle, non-contact obstacle) for all natural feature points belonging to the obstacle. ).

In step S38, in order to use the state of the crane currently captured by the video camera 10 as a reference image in the next sample, a reference image update process is performed to overwrite the reference image with the current camera image, which is the main process. Return to collision avoidance process.
Then, after returning to the main process, the process proceeds to step S4 shown in FIG. 2, and in step S4, the presence or absence of a substantial “obstacle” is determined based on the result of the classification process. That is, if no natural feature point that can exist before the turning center axis is extracted in step S35 (No), the process returns to step S3, and the natural feature point existing before the turning center axis is found. When even one is extracted (Yes), the process proceeds to step S5.

In step S5, when it is determined that there is a natural feature point other than the “non-contact obstacle” classified in step S37, that is, the natural feature point related to either “attention obstacle” or “dangerous obstacle”. When the presence is determined, it is determined that there is a caution obstacle (Yes), it is determined that the distance between the boom 4 and the obstacle is short, and the process proceeds to step S6.
Otherwise, among the obstacles determined in step S4 above, when there is no natural feature point related to the attention obstacle and the dangerous obstacle (No), that is, classified as “non-contact obstacle” in step S37. For the object, it is determined that the distance between the boom 4 and the obstacle is sufficiently large, and the process returns to step S3. In the subsequent step S6, since the boom 4 is close to an obstacle, a signal for issuing an alarm is output and the process proceeds to step S7. As a result, an alarm such as a buzzer is sounded to alert the operator.

In step S7, when it is determined that there is a natural feature point related only to the “dangerous obstacle” classified in step S37, it is determined that there is a dangerous obstacle (Yes), and the boom 4 and the obstacle are It is determined that the distance is very close, and the process proceeds to step S8.
Otherwise, that is, when there is no natural feature point related to only the dangerous obstacle (No), it is determined that the distance between the boom 4 and the obstacle is somewhat apart, and the process returns to step S3. In step S8, in order to prevent the boom 4 from coming into contact with the obstacle, an operation stop signal for the boom 4 in the approaching direction to the obstacle (boom raising, lying down, turning left and right, and on the hook) is output to step S3. Return processing. In this case, the actual crane is automatically stopped in step S8 while the alarm in step S6 is sounded.

Next, operation | movement of the collision avoidance apparatus of this vehicle-mounted crane 1 and its effect | action and effect are demonstrated.
When carrying out the crane work with the vehicle-mounted crane 1, the operator inputs a desired crane operation from the position away from the crane body 1 by operating each selection switch and speed lever of the remote controller. As a result, a signal corresponding to a desired crane operation is input to the controller 30 by wireless communication, and the controller 30 outputs an operation command for operating each actuator of the crane body 1 to operate the desired actuator. Thereby, at the time of crane work, the operator can perform boom operation including raising / lowering, extending / contracting, and turning operation of the boom 4 from a position away from the crane body 1.

Here, as shown in FIG. 1, there is a concern that the boom 4 and the utility pole H may come into contact with each other when the hanging work is performed near the utility pole H in an urban area or a railway track as shown in FIG.
On the other hand, the vehicle-mounted crane 1 of the present embodiment has a collision avoidance device that includes the video camera 10, the image processing device 20, and the controller 30. When the turning operation is performed, the collision avoidance process shown in FIGS. 2 and 3 described above is performed to track the natural feature point, and the collision avoidance control is performed based on the tracking result of the natural feature point. Based on the image information acquired according to the situation, it is possible to determine whether or not the boom 4 may collide with the utility pole H that is an obstacle.

According to the collision avoidance device, since the video camera 10 and the image processing device 20 are used to function as a distance sensor, it is possible to eliminate human error such as a mistake in setting the work regulation range. In addition, since there is no need for prior settings for inputting work environment data, the work can be started quickly and safely.
In addition, according to this collision avoidance device, since the collision determination is based on the natural feature point of the object processed from the image information of the video camera 10, it is an object such as a utility pole, and an object having a complicated shape. Can also acquire image information and detect it. In addition, since the apparatus can be configured from images acquired by the video camera 10 as in the present embodiment, the number of sensor arrangement locations (arrangement number) is small, so that the collision avoidance apparatus can be easily manufactured and the local remodeling can be easily performed. Further, since the video camera 10 can shoot only with ambient light, it is safer than the case of laser scanning.

Further, according to this collision avoidance device, since recognition is easy if the object is other than glass or a mirror surface, it can be applied with high versatility to the situation at the work site.
In addition, the boom collision avoidance device for a working machine according to the present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the spirit of the present invention.
For example, in the above-described embodiment, an example in which the boom collision avoidance device for a work machine according to one aspect of the present invention is mounted on a vehicle-mounted crane is described. However, the present invention is not limited to this, and the work machine has a work boom. The present invention can be applied if there is any.

In the above embodiment, the example in which the image processing unit and the collision avoidance control unit are separately provided in the image processing device 20 and the controller 30 has been described. However, the present invention is not limited thereto, and for example, the image processing unit and the collision avoidance control are provided. Can also be provided in the controller 30.
In the above-described embodiment, the example in which the two video cameras 10 are provided has been described. However, the present invention is not limited to this. For example, even with a single video camera, the booms taken or swung by the video camera are operated. It is possible to track a natural feature point existing in the image when it is performed and determine an obstacle that may collide with the boom based on the tracking result of the natural feature point.

  In the above embodiment, as an example of the situation of the work site that can be used according to the present invention, an example of preventing contact with a utility pole when carrying a suspended work near a utility pole in an urban area or a railway line is described. However, the present invention is not limited to this, and the present invention can be widely applied to uses for limiting the position and posture of the boom in order to prevent the boom from interfering with an obstacle during work. For example, the present invention can be applied to a purpose of carrying a load while avoiding contact with a scaffold or the like when performing a suspended load work at a construction site.

1 Vehicle-mounted crane (crane body)
2 base 3 column 4 boom 5 wire rope 6 hook 7 hoisting cylinder 8 outrigger 9 vehicle 10 video camera 20 image processing device (image processing unit)
30 controller (collision avoidance control unit)
H Telephone pole (obstacle)
S electric wire

Claims (6)

A boom collision avoidance device for a work machine that is used in a work machine having a boom and that avoids a collision between the boom and an obstacle in order to prevent the boom from interfering with an obstacle during work,
The video camera provided to photograph the boom and the background at the tip of the boom, and the corners and contours of the object, etc., present in the image when the boom is moved up and down or turned by the video camera. Information on obstacles that may collide with the boom based on the result of tracking natural feature points, which are points where characteristic parts with clear contrast and hue difference are extracted from the background. And a collision avoidance control unit that performs collision avoidance control for avoiding a collision between the boom and the obstacle based on the obstacle information generated by the image processing unit. Boom collision avoidance device for work machines. The image processing unit recognizes an object that moves differently from the movement of the natural feature point of the background as an obstacle based on the tracking result of the natural feature point when the boom is raised or lowered. And calculating the estimated position of the natural feature point of the obstacle after a certain period of time as the information of the obstacle,
2. The collision avoidance control unit according to claim 1, wherein the collision avoidance control unit performs a collision avoidance control such as an alarm to an operator or a boom deceleration, stop, or restriction of an operation direction based on the information on the estimated position. Boom collision avoidance device for work equipment. The image processing unit, based on the estimated position of the natural feature point of the obstacle, a cautionary obstacle that may collide with the boom after a lapse of a certain time, and a dangerous obstacle that is judged to collide with the boom after a lapse of a certain time, , Performing a classification process that classifies the non-contact obstacle that does not collide with the boom after a certain period of time according to the risk of contact with the boom,
When the collision avoidance control unit determines that the non-contact obstacle is based on the result of the classification process, the normal collision operation is continued. The boom collision avoidance device for a work machine according to claim 2, wherein when the danger obstacle is determined, a boom operation stop signal in a direction of approaching the obstacle is output.   The image processing unit includes a natural feature point tracking unit that tracks a natural feature point of an object captured by the video camera when a boom is raised or lowered, and obtains a moving direction and a moving speed thereof. The work machine according to any one of claims 1 to 3, further comprising obstacle extraction means for extracting a natural feature point belonging to an obstacle from the natural feature points tracked by the natural feature point tracking means. Boom collision avoidance device.   The obstacle extraction means includes a natural feature point of an object located on the near side of the turning center axis with respect to the turning center axis of the boom, and a natural feature point of the object located on the back side of the turning center axis. It is determined whether it is a natural feature point that exists in front or a natural feature point that exists in the back, and the natural feature point that is determined to exist in front of the turning center axis belongs to the obstacle It extracts as a natural feature point, The boom collision avoidance apparatus for work machines of Claim 4 characterized by the above-mentioned.   The natural feature point tracking means searches for a natural feature point of a target object that can be tracked from a current camera image, and then associates the trackable natural feature point of the previous camera image with the current camera image. 5. The work machine boom collision according to claim 4, further comprising: obtaining the moving speed of the natural feature point by dividing the moving distance of the natural feature point on the camera image by one sample time. Avoidance device.

Images