JP2023035434A - Object detection device, system, method, and program - Google Patents

Abstract

To provide an object detection device capable of contributing to identification and detection of an object even when information on the object in a monitoring area is partial or sparse.SOLUTION: An object detection device includes: a monitoring area specification unit for specifying a monitoring area relative to three-dimensional point group data related to a site; a reference surface determination unit for determining a reference surface based on a point group in the monitoring area; and a detection unit for selectively detecting a detection target object matching conditions information based on the point group on the reference surface from among the point group in the monitoring area. The detection unit includes: a detection conditions specification unit for specifying detection conditions of a detection target object candidate; a detection processing unit for detecting a detection target object candidate satisfying the detection conditions based on the point group in the monitoring area; an exclusion conditions specification unit for specifying exclusion conditions including a specific form of part of detection exclusion objects; and an exclusion processing unit for excluding the detection exclusion object matching the exclusion conditions from the detection target object candidate based on the point group related to the detection target object candidate.SELECTED DRAWING: Figure 7

Description

The present invention relates to an object detection device, system, method, and program.

In railway companies, there is a need to realize safety management assistance during railway track construction by checking for workers or the like entering the track under inspection, especially for objects on the track and its surroundings when a train is passing. Objects to be checked include not only people, but also work vehicles (eg, rail vehicles). For trains, of course, we want to allow them to enter the tracks and exclude them from the objects we want to check. In order to meet such needs, there are techniques for detecting an object within a monitored area (monitored space) using a three-dimensional sensor (see Patent Documents 1 and 2, for example). Furthermore, there is also a technique that can switch the necessity of detection depending on conditions such as the size, volume, and presence or absence of movement of a detected object (see, for example, Patent Document 3).

JP 2005-214718 A JP 2018-90099 A JP 2021-60784 A

The following analysis is given by the inventors.

However, when the entire object does not exist within the monitoring area (monitored space) (for example, when only part of the object enters the monitoring area), or when the object is far from the 3D sensor and is imaged, When the information about the object in the monitoring area is partial or sparse, such as when the point cloud has a large sparse part, the techniques described in Patent Documents 1 to 3 can detect similar objects in the monitoring area. (e.g. rail vehicles and trains) are difficult to discriminate and detect.

A main object of the present invention is to provide an object detection apparatus, system, method, and program that can contribute to identifying and detecting an object even when information about the object in the monitoring area is partial or sparse. to provide.

An object detection apparatus according to a first viewpoint includes an input unit configured to receive a user's operation, and a monitoring area for three-dimensional point cloud data related to a site based on designation information from the input unit. a monitoring area specifying unit configured to specify a monitoring area; a reference plane determining unit configured to determine a reference plane based on a point group within the monitoring area in the three-dimensional point cloud data; and the three-dimensional point a detection unit configured to selectively detect a detection target object that matches the condition information from the input unit based on the point group on the reference plane among the point groups in the monitoring area in the group data; , wherein the detection unit includes a detection condition specifying unit configured to specify a detection condition for the detection target object candidate based on the condition information; and a point cloud within the monitoring region in the three-dimensional point cloud data. a detection processing unit configured to detect a detection target object candidate that satisfies the detection condition based on; and an exclusion condition designating unit configured to exclude detection-excluded objects that match the specified exclusion condition from the detection-target object candidates based on the detected point group related to the detection-target-object candidates. an exclusion processor configured to determine an object to be detected.

An object detection system relating to a second viewpoint includes an object detection device relating to the first viewpoint, and a three-dimensional sensor communicably connected to the object detection device and capturing an image of a scene.

An object detection method relating to a third viewpoint is an object detection method for detecting an object using hardware resources. determining a reference plane based on a point group within the monitoring area in the three-dimensional point cloud data; and selectively detecting a detection target object that matches the input condition information based on the above point cloud, wherein the detection step includes detection conditions for detection target object candidates based on the condition information. a step of detecting a detection target object candidate that satisfies the detection condition based on the point group in the monitoring region in the three-dimensional point cloud data; and a step of detecting a detection exclusion object based on the condition information designating an exclusion condition including a part of a specific form; and removing a detection exclusion object that matches the specified exclusion condition from the detection target object candidate based on the point group related to the detected detection target object candidate. and determining the object to be detected by exclusion processing.

A program related to a fourth viewpoint is a program that causes a hardware resource to execute processing for detecting an object, and designates a monitoring area for 3D point cloud data related to the site based on input designation information. a process of determining a reference plane based on the point group within the monitoring area in the three-dimensional point cloud data; and a process of selectively detecting a detection target object that matches input condition information based on the point cloud; A process of designating detection conditions for candidates; a process of detecting detection target object candidates that satisfy the detection conditions based on the point group within the monitoring region in the three-dimensional point cloud data; and based on the condition information, A process of designating an exclusion condition including a part of the peculiar form of the detection exclusion object, and detecting the detection exclusion object that matches the specified exclusion condition based on the point cloud related to the detected detection target object candidate. and a process of determining the detection target object by excluding it from target object candidates.

Note that the program can be recorded in a computer-readable storage medium. The storage medium can be non-transient such as semiconductor memory, hard disk, magnetic recording medium, optical recording medium, and the like. The present disclosure may also be embodied as a computer program product. A program is input to a computer device via an input device or an external communication interface, is stored in a storage device, drives a processor in accordance with predetermined steps or processes, and stages the results of processing including intermediate states as necessary. can be displayed via a display device, or can be communicated with the outside via a communication interface. A computer device for this purpose, as an example, typically includes a processor, a storage device, an input device, a communication interface, and optionally a display device, all of which are connectable to each other by a bus.

According to the first to fourth viewpoints, it is possible to contribute to identifying and detecting an object even when information about the object within the monitoring area is partial or sparse.

FIG. 2 is an image diagram showing an example of detecting an object using the object detection system according to Embodiment 1; 1 is a block diagram schematically showing the configuration of an object detection system according to Embodiment 1; FIG. 4 is a block diagram schematically showing the configuration of an exclusion processing unit in the object detection system according to Embodiment 1; FIG. 4 is a flowchart schematically showing the operation of the object detection device in the object detection system according to Embodiment 1; 7 is a flowchart schematically showing details of exclusion processing of an object detection device of the object detection system according to the first embodiment; FIG. 10 is an image diagram showing an example of detecting an object using an object detection system according to a comparative example; 2 is a block diagram schematically showing the configuration of an object detection device according to Embodiment 2; FIG. 3 is a block diagram schematically showing the configuration of hardware resources; FIG.

Hereinafter, embodiments will be described with reference to the drawings. It should be noted that when reference numerals are attached to the drawings in this application, they are solely for the purpose of helping understanding, and are not intended to limit the embodiments shown in the drawings. Moreover, the following embodiments are only examples, and do not limit the present invention. Also, connection lines between blocks in drawings and the like referred to in the following description include both bidirectional and unidirectional connections. The unidirectional arrows schematically show the flow of main signals (data) and do not exclude bidirectionality. Furthermore, in the circuit diagrams, block diagrams, internal configuration diagrams, connection diagrams, etc. disclosed in the present application, an input port and an output port exist at the input end and the output end of each connection line, respectively, although not explicitly shown. The input/output interface is the same. The program is executed via a computer device, and the computer device includes, for example, a processor, a storage device, an input device, a communication interface, and optionally a display device. It is configured to be able to communicate with external devices (including computers), whether wired or wireless.

[Embodiment 1]
An object detection system according to Embodiment 1 will be described with reference to the drawings. FIG. 1 is an image diagram showing an example of detecting an object using the object detection system according to the first embodiment. FIG. 2 is a block diagram schematically showing the configuration of the object detection system according to Embodiment 1. FIG. 3 is a block diagram schematically showing the configuration of an exclusion processing unit in the object detection system according to the first embodiment; FIG.

The object detection system 1 selectively excludes predetermined objects (objects excluded from detection; trains 23 and 24 in FIG. It is a system for detecting objects other than objects (objects to be detected; worker 21 and road-rail vehicle 22 in FIG. 1) (see FIGS. 1 and 2). If the object detection system 1 excludes the trains 23 and 24, in the field of safety surveillance, it is possible to detect an object (for example, a passerby) intruding into a railroad crossing (monitoring area) where the barrier has exited. , excluding trains 23, 24 in the detection of stranded objects (eg, automobiles) in railroad crossings. In addition, the object detection system 1 separates general vehicles from buses and excludes only buses, so that it can detect general vehicles entering a bus terminal (monitoring area) in the field of transportation infrastructure and issue a warning. can be used. The object detection system 1 includes an object detection device 200 and a 3D sensor 300 .

The object detection device 200 selects predetermined objects (objects excluded from detection; trains 23 and 24 in FIG. 1) in the monitoring area 20 based on the three-dimensional point cloud data 100 of the shooting scene 10 photographed by the three-dimensional sensor 300. It is a device for detecting objects (objects to be detected; worker 21 and road-rail vehicle 22 in FIG. 1) other than predetermined objects in the monitoring area 20 by physically excluding them (see FIGS. 1 and 2). The object detection device 200 can be configured not to detect an object (non-detection target object; worker 25 in FIG. 1) existing in an area other than the monitoring area 20 at the shooting site 10 . For the object detection device 200, a terminal (computer terminal) having functional units (eg, a processor, a storage device, an input device, a communication interface, and a display device) constituting a computer can be used. , a server, a tablet terminal, a smartphone, or the like can be used. The object detection device 200 is directly or indirectly connected to the three-dimensional sensor 300 so as to be communicable (wireless or wired), and can communicate with the three-dimensional sensor 300 via a network (not shown). may be connected to By executing a predetermined program, the object detection apparatus 200 includes a preprocessing unit 210, a monitoring area designating unit 220, a reference plane determining unit 230, a detecting unit 240, a result generating unit 260, and a user interface unit. 270 and .

The preprocessing unit 210 is a functional unit that performs preprocessing (data shaping) on the input three-dimensional point cloud data 100 (see FIG. 2). The 3D point cloud data 100 processed by the preprocessing unit 210 can be displayed and output by the output unit 271 of the user interface unit 270 . The preprocessing section 210 includes a format conversion section 211 and a noise removal section 212 .

The format conversion unit 211 is a functional unit that converts the format of the input three-dimensional point cloud data 100 into a common format commonly used in the object detection device 200 as preprocessing (see FIG. 2). The format conversion unit 211 outputs the converted common format three-dimensional point cloud data 100 to the noise removal unit 212 . Note that if the format of the three-dimensional point cloud data 100 is originally a common format, the format conversion processing in the format conversion unit 211 may be skipped.

The noise removal unit 212 is a functional unit that removes noise (point groups unnecessary for detection) from point groups in the three-dimensional point group data 100 from the format conversion unit 211 as preprocessing (see FIG. 2). The noise removal unit 212 outputs the noise-removed three-dimensional point cloud data 100 to the monitoring region designation unit 220 , the reference plane determination unit 230 , and the user interface unit 270 . Noise removal methods include, for example, smoothing processing, filtering (eg, moving average filter processing, median filter processing, etc.), outlier removal processing (eg, outlier removal processing by chi-square test), and the like. It should be noted that if there is almost no noise, the noise removal processing in the noise removal unit 212 may be skipped.

The monitoring area designation unit 220 is a functional unit that designates the monitoring area 20 in the three-dimensional point cloud data 100 relating to the shooting site (10 in FIG. 1) (see FIG. 2). A plurality of monitoring areas 20 may be designated and switched. The switching may be performed not only manually but also automatically at a specific time or by a timer. The monitoring area designation unit 220 designates the monitoring area 20 based on the designation information from the input unit 272 of the user interface unit 270 input by the user's operation. The designated monitoring area 20 is stored. If the entire area of the three-dimensional point cloud data 100 is the monitoring area 20, designation of the monitoring area 20 can be skipped. The monitoring area 20 specified by the monitoring area specifying section 220 can be displayed and output by the output section 271 of the user interface section 270 . The monitoring area specifying section 220 includes a range specifying section 221 and an angle specifying section 222 .

The range specifying unit 221 is a functional unit that specifies the range (width, depth, and height) of a partial area (monitoring area 20) to be monitored in the captured panoramic view of the three-dimensional point cloud data 100 (see FIG. 2). ). The range designating section 221 designates the range of the monitoring area 20 based on the designation information from the input section 272 of the user interface section 270 input by the user's operation. The range specifying unit 221 outputs information about the specified range of the monitoring area 20 to the reference plane determining unit 230 and the detecting unit 240 .

The angle specifying unit 222 is a functional unit that specifies an angle for rotating the point group of the three-dimensional point cloud data 100 about a predetermined coordinate point (or any specified coordinate point) (see FIG. 2). The angle designation unit 222 designates the angle for rotating the point cloud based on the designation information from the input unit 272 of the user interface unit 270 input by the user's operation. The specification of the angle for rotating the point group by the angle specification unit 222 assists the specification of the range of the monitoring area 20 by the range specification unit 221 . If rotation of the point group is unnecessary, the angle specifying section 222 may be omitted or the angle specification may be skipped. The angle designating section 222 outputs information about the designated angle to the reference plane determining section 230 and the detecting section 240 . Information about the specified angle is independent of information about the range of the specified monitoring area 20 .

The reference plane determination unit 230 is a functional unit that determines the reference plane (ground 14 in FIG. 1) in the monitoring area 20 (see FIG. 2). A plurality of reference planes may be determined and switched. The switching may be performed not only manually, but also automatically at a specific time or by a timer. The reference plane may be set not only on a horizontal ground or floor, but also on a slope or in the air. This makes it possible to monitor not only cars and people on the ground, but also birds and drones. The determined reference plane is stored. The reference plane determined by the reference plane determining section 230 can be displayed and output by the output section 271 of the user interface section 270 . The reference plane determination section 230 includes a plane detection section 231 , a reference plane selection section 232 and a reference plane storage section 233 .

The plane detection unit 231 has a function of detecting one or more planes (for example, a plane having an area equal to or larger than a preset area; the ground 14 in FIG. 1) based on the point cloud within the monitoring region 20 in the three-dimensional point cloud data 100. part (see FIG. 2). The plane detection unit 231 acquires the 3D point cloud data 100 preprocessed by the preprocessing unit 210 and acquires information (range, angle) related to the monitoring area 20 specified by the monitoring area specifying unit 220 . The plane detection unit 231 sets the monitoring region 20 in the 3D point cloud data 100 within the specified range, and rotates the point cloud of the 3D point cloud data 100 (including the monitoring region 20) at the specified angle. You can create an image. The plane detection unit 231 detects (searches) one or more planes from point groups within the monitoring region 20 in the three-dimensional point group data 100 . The plane detection unit 231 presets the types of planes to be detected (for example, horizontal planes, vertical planes, inclined planes, boundary planes within the range of the monitoring area 20, etc.), and selectively detects planes of the set types. You may make it Depending on the characteristics of the three-dimensional sensor 300, the actual plane may become a wavy point group in imaging due to variations in the errors in the coordinates that can be acquired. By setting the specified range of the error threshold between each point group to be large, the plane can be detected with less influence of waviness.

The reference plane selection unit 232 is a functional unit that selects a reference plane from the planes detected by the plane detection unit 231 (see FIG. 2). The reference plane selection unit 232 selects a reference plane based on selection information from the input unit 272 of the user interface unit 270 input by the user's operation. The reference plane selection unit 232 can designate specific plane information that is not based on the plane detection unit 231 as the floor plane.

The reference plane storage unit 233 is a functional unit that stores information related to the reference plane selected by the reference plane selection unit 232 (see FIG. 2). Information related to the reference plane stored in the reference plane storage unit 233 can be used by the detection unit 240 .

The detection unit 240 selectively detects a detection target object based on the point group on the reference plane (the reference plane determined by the reference plane determination unit 230) among the point groups in the monitoring region 20 in the three-dimensional point cloud data 100. It is a functional unit for detection (see FIG. 2). The detection unit 240 may additionally use an RGB (Red-Green-Blue) image for detection processing. The detection unit 240 includes a detection condition specifying unit 241 , a detection processing unit 242 , an exclusion condition specifying unit 243 and an exclusion processing unit 244 .

The detection condition designation unit 241 is a functional unit that designates detection conditions for detection target object candidates (see FIG. 2). The detection condition designation unit 241 designates detection conditions for detection target object candidates based on condition information from the input unit 272 of the user interface unit 270 that is input by a user's operation. Conditions for detection of candidate objects to be detected include, for example, size ranges and volume ranges of objects, as well as conditions with a small amount of calculation (conditions with a small amount of calculation compared to exclusion conditions), such as discovering moving objects by checking them in chronological order. is mentioned. Note that it is not necessary to specify the detection conditions for the detection target object candidates. Designation of candidate detection conditions may be skipped.

The detection processing unit 242 detects detection target object candidates that satisfy the detection conditions for the detection target object candidates specified by the detection condition specifying unit 241 based on the point groups in the monitoring region 20 in the three-dimensional point cloud data 100 . (See FIG. 2). The detection processing unit 242 may use an image created by the plane detection unit 231 (an image in which the monitoring area 20 is set in the three-dimensional point cloud data 100). The detection processing unit 242 deletes the point groups of the areas other than the monitoring area 20 in the three-dimensional point cloud data 100, and extracts the reference plane (see figure In 1, the point cloud related to the ground 14) is deleted to isolate the point cloud related to each object on the reference plane. The detection processing unit 242 performs segmentation processing (grouping of the point cloud) based on the point cloud in the monitoring area 20 in which the point cloud of each object on the reference plane is isolated (the point cloud related to the reference plane is deleted). ) to detect an object (corresponding to a detection target object candidate) as one object in one segment. Candidates of detection target object candidates are the tracks 11, 12, and 13, the worker 21, the rail vehicle 22, and the trains 23 and 24 in the example of FIG. In the segmentation process, for example, a group of points in which the distance between adjacent points is within a predetermined distance can be processed as one segment. The detection processing unit 242 detects, as a detection target object candidate, an object that satisfies the detection conditions for the detection target object candidate specified by the detection condition specifying unit 241, based on the point cloud related to the object detected as one segment and one object. . The detection target object candidates are the worker 21, the rail vehicle 22, and the trains 23 and 24 in the example of FIG. AI (Artificial Intelligence), pattern matching, or the like may be used to detect detection target object candidates. The detection processing unit 242 stores point groups related to detected detection target object candidates.

The exclusion condition designation unit 243 is a functional unit that designates exclusion conditions (detection conditions for objects excluded from detection) (see FIG. 2). The exclusion condition specifying unit 243 is a functional unit that specifies an exclusion condition including a part of the peculiar form of the detection exclusion object based on the condition information from the input unit 272 of the user interface unit 270 input by the user's operation. (see FIG. 3). As exclusion conditions, some peculiar (concrete, unique) form of the detection exclusion object, for example, some peculiar shape or structure of the detection exclusion object, some plane of the detection exclusion object (e.g., top surface, side surface, driver side surface, conductor side surface, etc.), a predetermined plane (e.g., train characteristic position and orientation of the driver's side surface, conductor's side surface, if any), a characteristic dimension range of a predetermined plane of the part of the detection exclusion object (e.g., vertical length range, horizontal length range, etc.) that require a large amount of calculation (a condition that requires a large amount of calculation compared to the detection condition for the detection target object candidate). As for the method of specifying the exclusion condition, the configuration of the detection exclusion object may be detected from the three-dimensional point cloud data 100 photographed in advance, and may be selected or adjusted. An RGB image may also be used as a supplement when specifying exclusion conditions. The exclusion conditions are set and stored in a list or database. Even after specifying the exclusion condition, the determination result of the determination unit 252 may be accumulated and machine learning may be performed to update the exclusion condition.

The exclusion processing unit 244 detects objects (detection exclusion objects; in FIG. 1, trains 23, 24) from the detection target object candidates to determine the detection target objects (the worker 21 and the rail vehicle 22 in FIG. 1) (see FIGS. 2 and 3). The exclusion processing unit 244 excludes all of the detection target object candidates detected by the detection processing unit 242 one by one. Although the exclusion processing unit 244 performs the exclusion processing after the detection processing by the detection processing unit 242, the exclusion processing may be performed before the detection processing by the detection processing unit 242. FIG. The exclusion processing unit 244 includes a form detection unit 251 , a determination unit 252 and an exclusion unit 253 .

The form detection unit 251 detects the form of each detection target object candidate (for example, shape, plane, position and direction of the plane, size range of the plane, etc.) based on point groups related to all the detection target object candidates detected by the detection processing unit 242. etc.) (see FIG. 3). Detection of the form of the detection target object candidate is performed by detecting the form of each detection target object candidate as much as possible from the point group related to the detection target object candidate.

The determination unit 252 is a functional unit that determines whether or not there is a shape of each detection target object candidate detected by the shape detection unit 251 that corresponds to the exclusion condition specified by the exclusion condition specification unit 243. (See Figure 3). AI (Artificial Intelligence), pattern matching, or the like may be used to determine whether or not there is an exclusion condition. If there is no applicable object, the determination unit 252 determines all of the detection target object candidates as detection target objects.

The excluding unit 253 excludes objects (objects excluded from detection; trains 23 and 24 in FIG. 1) determined by the determining unit 252 to meet the exclusion conditions from the candidates for detection objects (workers in FIG. 1). 21, a functional unit that determines the track/road vehicle 22) (see FIG. 3). The exclusion unit 253 may label the point group of the determined object to be detected (the worker 21 and the road-rail vehicle 22 in FIG. 1) so that it can be tracked.

Taking the case of excluding trains as an example, since trains are composed of many planes and the standards are generally unified, objects including trains were detected from the 3D point cloud data 100 of the 3D sensor 300. After that, the plane specified by the exclusion condition is further detected from there, and it is determined that the detected plane corresponds to the specific form of the train (height, ratio to the whole train, arrangement, etc.) , trains (objects to be excluded from detection) can be excluded from detection target object candidates based not only on the dimensions of the entire train, but also on the shape of a part of the train.

Note that, if there is an object (detection excluded object) having a form determined by the determination unit 252 to satisfy the exclusion condition, the exclusion unit 253 excludes all objects in the monitored space in which the detection excluded object exists. You may do so. Also, instead of the exclusion unit 253, a functional unit that can perform a different action (for example, notify the user) from the normal detection result may be used.

The result generation unit 260 is a functional unit that generates (or calculates) a result to be notified to the administrator based on the information on the monitoring area (20 in FIG. 1), the information on the reference plane, and the detection information. (see Figure 2). The result generating unit 260 acquires information about the monitoring area 20 from the monitoring area designating unit 220, acquires information about the reference plane from the reference plane determining unit 230, and detects detection information (at least detection target) from the detecting unit 240. information about the object). The generated result can be the figure of the monitoring area 20, the figure of the reference plane, and the figure and specifications of the object of the detection information (for example, the size, volume, position of the center of gravity, moving speed, etc. of the object). The result generating section 260 outputs the generated result to the user interface section 270 . The result generation unit 260 determines whether or not there is a detection target object (for example, a labeled detection target object) in the detection information, and if there is a detection target object, displays warning information along with the generated result as a user interface. You may make it output toward the face part 270. FIG. The result generator 260 includes a specification calculator 261 and a figure generator 262 .

The specification calculator 261 is a functional unit that calculates specifications based on the detected information (see FIG. 2). The calculated specifications include, for example, the dimensions, volume, center-of-gravity position, moving speed, etc. of the object to be detected (the worker 21 and the road-rail vehicle 22 in FIG. 1). The specification calculation unit 261 outputs the calculated specifications to the user interface unit 270 as a result.

The graphic generation unit 262 is a functional unit that generates a graphic based on information related to the monitoring area (20 in FIG. 1), information related to the reference plane, and detection information (including detection conditions, exclusion conditions, and detection results). (see FIG. 2). The generated graphics include, for example, the graphics of the monitoring area 20, the graphics of the reference plane, and the graphics of the detection information (the detection frames 61 and 62 and the graphics of the object in FIG. 1). A figure is generated taking into account the position. Also, in the generation of graphics, the generation of some of the elements of the monitoring area 20, the reference plane, and the detection information can be omitted. Note that the graphic generation unit 262 generates detection target objects (worker 21 and rail vehicle 22 in FIG. 1), detection exclusion objects (trains 23 and 24 in FIG. 1), and non-detection target objects (worker 25 in FIG. 1). In order to clarify the difference between the two, only the graphic of the detection target object may be generated, and the graphics of the detection exclusion object and the non-detection target object may not be generated. The graphic generation unit 262 outputs the generated graphic as a result to the user interface unit 270 .

The user interface unit 270 is a functional unit that exchanges information between the user and the object detection device 200 (see FIG. 2). The user interface section 270 has an output section 271 and an input section 272 .

The output unit 271 is a functional unit that outputs various information (display output, audio output, transmission output, etc.) (see FIG. 2). The output unit 271 can output the preprocessed three-dimensional point cloud data 100 from the noise removal unit 212 of the preprocessing unit 210 . The output unit 271 may output the monitoring area 20 extracted by the monitoring area designating unit 220 or the plane position determined by the reference plane determining unit 230 . The output unit 271 can output the results from the result generation unit 260 (detection frames 61 and 62 in FIG. 1, specifications, etc.) and warning information (if acquired). As a method of outputting the warning information, for example, the background color may be changed or blinked for display output, a warning sound may be output by voice, or the warning information may be transmitted and output to the terminal of the user. The output three-dimensional point cloud data 100 and results may be saved as screen shots, numerical data, or the like. It should be noted that the results from the result generation unit 260 need not be the detection frames 61 and 62 as long as the position of the object to be detected can be specified, and at the same time arbitrary information calculated by the specification calculation unit 261 may be output at the same time. .

The input unit 272 is a functional unit that receives a user's operation (see FIG. 2). The user operates the input unit 272 (e.g., keyboard operation, mouse click operation, touch panel tap operation, etc.) to input data (e.g., characters, numbers, positions, areas, etc.) to the monitoring area designation unit 220 Alternatively, it outputs to the reference plane determining section 230 or the detecting section 240 . The input unit 272 outputs information related to the conditions (width, depth, height, angle) input by the user's operation to the monitoring area designating unit 220 . The input unit 272 selects one or more planes from among one or more planes detected by the reference plane determination unit 230 based on the point cloud in the monitoring area (20 in FIG. 1) in the three-dimensional point cloud data 100 by the user's operation. An instruction is given to select a reference plane (ground 14 in FIG. 1). The input unit 272 inputs detection conditions (for example, height range, size range, volume range, time series to find moving objects, etc.) and exclusion conditions (plane position of detection exclusion object). direction, dimension range of the detection exclusion object, etc.) to the detection unit 240 . The input unit 272 instructs the start/end of the overall processing in this configuration function by user's operation.

The three-dimensional sensor 300 is a device that three-dimensionally senses and captures the surface of a subject (workers 21 and 25, track and road vehicle 22, trains 23 and 24, tracks 11, 12 and 13, ground 14, etc. in FIG. 1). (see FIGS. 1 and 2). The three-dimensional sensor 300 is communicably connected to the object detection device 200 . The three-dimensional sensor 300 generates three-dimensional point cloud data 100 in a predetermined format by photographing an object, and outputs the generated three-dimensional point cloud data 100 to the object detection device 200 . Note that the three-dimensional point cloud data 100 may be generated by the object detection device 200 instead of by the three-dimensional sensor 300 . For the three-dimensional sensor 300, for example, a ToF (Time of Flight) camera, a stereo camera, a three-dimensional LIDAR (Laser Imaging Detection And Ranging), a depth sensor, a ranging sensor, a range camera, or the like can be used. Here, the three-dimensional point cloud data 100 is data generated in a predetermined format by the three-dimensional sensor 300, and is drawn as a point cloud (collection of many points having XYZ coordinate (three-dimensional coordinate) information). This is the point cloud data (see Fig. 2). The three-dimensional sensor 300 can be changed to a sensor device with various output formats according to environmental conditions such as shooting distance, angle of view, indoors/outdoors, presence/absence of sunshine, etc. required for a solution, and customer requests. There may be a plurality of three-dimensional sensors 300 (whether or not they are of the same type). Impossible wide range monitoring may be performed. Also, the three-dimensional sensor 300 may move within a range where the relative position with respect to the reference plane does not change.

Next, the operation of the object detection device in the object detection system according to Embodiment 1 will be described with reference to the drawings. 4 is a flowchart schematically showing the operation of the object detection device in the object detection system according to the first embodiment; FIG. For the detailed configuration of the object detection system, see FIG. 2 and its description.

First, the preprocessing unit 210 of the object detection device 200 acquires the three-dimensional point cloud data 100 (the point cloud data of the area including the monitoring area 20) of the subject photographed by the three-dimensional sensor 300 (step A1).

Next, as preprocessing, the format conversion unit 211 of the preprocessing unit 210 of the object detection device 200 converts the format of the acquired three-dimensional point cloud data 100 into a common format that can be used commonly in the object detection device 200 (step A2).

Next, as preprocessing, the noise removal unit 212 of the preprocessing unit 210 of the object detection device 200 removes noise from the point cloud in the three-dimensional point cloud data 100 format-converted by the format conversion unit 211 (step A3).

Next, the monitoring area designating unit 220 of the object detection device 200 monitors the entire captured view of the three-dimensional point cloud data 100 based on the designation information from the input unit 272 of the user interface unit 270 input by the user's operation. A region 20 (range, rotation angle) is specified (step A4). After specifying the monitoring area 20 first, step A4 can be skipped.

Next, the reference plane determination unit 230 of the object detection device 200 detects a plane from the point group in the monitoring area among the point groups of the three-dimensional point group data 100, and detects the plane from the point group input by the user's operation. Based on the selection information from the input unit 272 of the unit 270, a reference plane (for example, the ground 14) is determined (selected) from among the detected planes, and information related to the determined reference plane is stored (step A5). ). Note that step A5 can be skipped after the reference plane is first determined.

Next, the detection condition designation unit 241 of the detection unit 240 of the object detection device 200 sets the detection conditions for the detection target object candidate based on the designation information from the input unit 272 of the user interface unit 270 input by the user's operation. (For example, a range of dimensions, a range of volumes, a time-series confirmation to find a moving object, etc.) are specified (step A6). Note that step A6 can be skipped after specifying the detection conditions for the detection target object candidate first. Further, it is not necessary to specify the detection conditions for the detection target object candidates, and if no detection conditions are specified, the preset detection conditions for the detection target object candidates can be used, and step A6 can be skipped. .

Next, the detection processing unit 242 of the detection unit 240 of the object detection device 200 detects detection target object candidates that satisfy the specified detection target object candidate detection condition based on the point group within the monitoring area 20 (step). A7).

Next, the exclusion condition specifying unit 243 of the detection unit 240 of the object detection device 200 specifies an exclusion condition (detection exclusion object Some peculiar forms) are specified (step A8). Note that step A8 can be skipped after the exclusion condition is first specified. Moreover, the exclusion condition does not have to be specified, and if the detection condition is not specified, a preset exclusion condition can be used and step A8 can be skipped.

Next, the exclusion processing unit 244 of the detection unit 240 of the object detection device 200 analyzes the point group related to the detected detection target object candidate, and the object (detection exclusion object; trains 23 and 24) are excluded from the detection target object candidates, and the detection target objects (worker 21 and rail vehicle 22 in FIG. 1) are determined (step A9). Details of step A9 will be described later (see FIG. 5).

Next, the result generation unit 260 of the object detection device 200 generates (or calculates) a result to be notified to the administrator based on the latest detection result of the detection unit 240, and causes the output unit 271 to output the result ( Step A10). Thereafter, the process returns to step A1, and the flow from step A1 to step A10 is repeatedly executed as long as new three-dimensional point cloud data 100 is obtained.

Next, details of the object detection device exclusion process (step A9) in the object detection system according to the first embodiment will be described with reference to the drawings. FIG. 5 is a flowchart schematically showing details of exclusion processing of an object detection device of the object detection system according to the first embodiment. For the detailed configuration of the object detection system, see FIG. 2 and its description.

In the exclusion process (step A9) after step A8 in FIG. The form (eg, shape, plane, position and direction of the plane, size range of the plane, etc.) of each detection target object candidate is detected based on the point group of the candidate (step B1). Detection of the form of the detection target object candidate is performed by detecting the form of each detection target object candidate as much as possible from the point group related to the detection target object candidate.

Next, the determination unit 252 of the exclusion processing unit 244 of the detection unit 240 of the object detection device 200 selects, among the forms of detected detection target object candidates, those corresponding to the exclusion conditions specified by the exclusion condition specifying unit 243. (step B2). If there is no matching object, all the detection target object candidates are determined as detection target objects.

Next, the exclusion unit 253 of the exclusion processing unit 244 of the detection unit 240 of the object detection device 200 detects objects (detection exclusion objects; trains 23 and 24 in FIG. 1) determined by the determination unit 252 to meet the exclusion conditions. Objects to be detected (the worker 21 and the road-rail vehicle 22 in FIG. 1) are determined by excluding them from the target object candidates (step B3). After that, the process proceeds to step A10 in FIG.

Next, the operation of the object detection device in the object detection system according to Embodiment 1 will be described using the drawings while comparing with a comparative example. FIG. 6 is an image diagram showing an example of detecting an object using an object detection system according to a comparative example. Note that FIG. 1 should be referred to for an example of detecting an object using the object detection system according to the first embodiment.

The object detection system according to the comparative example is a system without the exclusion processing unit (244 in FIG. 2) of the object detection system according to the first embodiment. For the same 3D point cloud data as when the object was detected using the object detection system according to the first embodiment (see FIG. 1), when the object is detected using the object detection system according to the comparative example, the result is as shown in FIG. Become. In FIGS. 1 and 6, there are three tracks 11, 12, 13, and a monitoring area 20 is set for the two tracks 12, 13 on the far side. Therefore, the detection conditions are set large.

However, in the comparative example of FIG. 6, since there is no exclusion process, only the detection process based on the detection conditions detects the train 23 in a situation where only a part of the train 23 enters, such as the right end of the track 12 in the monitoring area 20. There is a possibility of erroneously detecting the target object. In the comparative example of FIG. 6, although most of the train 24 is reflected like the left end of the track 13, the point cloud of the middle part of the train 24 is affected by the distance to the sensor and the obstacles. is interrupted, the train 24 may be erroneously detected as an object to be detected.

Here, although the rail vehicle 22 is less than half the length of each vehicle of the trains 23 and 24, it is common to the trains 23 and 24 as a vehicle running on the tracks 11, 12 and 13. When viewed from (driver's seat side, conductor's seat side) (when projected on a plane), the vertical and horizontal sizes are very similar to the vehicle. Therefore, when the head or tail of the trains 23 and 24 is captured at the end of the monitoring area 20 (when the point group of the trains 23 and 24 is located at the end of the monitoring area 20), or when the trains 23 and 24 are three In at least two cases, when the three-dimensional point cloud data captured far from the dimensional sensor 300 has a large sparse part (when the train point cloud is interrupted due to the influence of a shield etc.) It becomes difficult to distinguish from the trains 23 and 24, and there is a possibility that the trains 23 and 24 are erroneously detected as the rail vehicle 22 (see FIG. 6). In addition, as a method of detecting the worker 21 and the rail vehicle 22 within the monitoring area 20, the monitoring area 20 is limited to a portion where the trains 23 and 24 do not exist while the train is passing, and the entire area is monitored again after the train is passing. Since we want to return to the state under the area 20, there is a possibility that the change of the monitoring area 20 cannot cope with the false detection.

On the other hand, in Embodiment 1 of FIG. 1, the detection process is performed based on the detection conditions to detect the worker 21, the rail vehicle 22, and the trains 23 and 24 as detection target object candidates. By performing exclusion processing using an exclusion condition that designates the form of the trains 23 and 24 (specific forms of the trains 23 and 24 not found in the rail vehicle 22), the trains 23 and 24 can be selectively excluded as detection exclusion objects, The worker 21 and the road-rail vehicle 22 can be correctly detected as objects to be detected. Further, in Embodiment 1 of FIG. 1, light-load detection processing is performed under conditions with a small amount of calculation, such as size comparison, and the number of detected objects is sieved to reduce the parameters used in subsequent exclusion processing. In addition, since exclusion processing is performed based on exclusion conditions that require a large amount of calculation based on complex forms of individual objects, it is possible to selectively detect detection target objects without impairing real-time performance of the entire processing.

According to the first embodiment, the exclusion process is performed based on the exclusion condition that designates the particular shape of a part of the detection exclusion object. can also contribute to identifying and detecting objects. In the example of FIG. 1, when detecting an intrusion into the monitoring area 20 (dangerous area) at the railway construction site using the three-dimensional sensor 300, only the worker 21 and the rail vehicle 22 are detected, and the trains 23 and 24 are detected. can be excluded from detection.

In addition, according to the first embodiment, it is possible to specify a detection condition for detecting an arbitrary object and specify an exclusion condition for not detecting it, so it is possible to improve the accuracy of selective object detection. can.

[Embodiment 2]
An object detection device according to Embodiment 2 will be described with reference to the drawings. FIG. 7 is a block diagram schematically showing the configuration of the object detection device according to the second embodiment.

The object detection device 200 is a device that selectively excludes detection-excluded objects in the monitoring area based on the three-dimensional point cloud data 100 of the site, and detects detection target objects other than the detection-excluded objects in the monitoring area 20. . The object detection device 200 includes an input section 272 , a monitoring area designation section 220 , a reference plane determination section 230 and a detection section 240 .

The input unit 272 is configured to receive a user's operation. The monitoring area designating section 220 is configured to designate a monitoring area for the three-dimensional point cloud data 100 relating to the site based on designation information from the input section 272 . The reference plane determination unit 230 is configured to determine a reference plane based on the point cloud within the monitoring region in the three-dimensional point cloud data 100 . The detection unit 240 selectively detects a detection target object that matches the condition information from the input unit 272, based on the point group on the reference plane among the point groups in the monitoring area in the three-dimensional point cloud data 100. is configured to

The detection unit 240 includes a detection condition specifying unit 241 , a detection processing unit 242 , an exclusion condition specifying unit 243 and an exclusion processing unit 244 . The detection condition designation unit 241 is configured to designate detection conditions for detection target object candidates based on condition information. The detection processing unit 242 is configured to detect a detection target object candidate that satisfies the detection condition based on the point group within the monitoring area in the three-dimensional point group data 100 . The exclusion condition specifying unit 243 is configured to specify an exclusion condition including a part of the detection exclusion object's specific form based on the condition information. The exclusion processing unit 244 determines the detection target object by excluding the detection exclusion object that meets the specified exclusion condition from the detection target object candidate based on the point group related to the detected detection target object candidate. It is configured.

According to the second embodiment, the exclusion process is performed based on the exclusion condition including the specific shape of a part of the detection excluded object. It can contribute to identifying and detecting objects.

Note that the object detection apparatuses according to the first and second embodiments can be configured by so-called hardware resources (information processing apparatus, computer), and those having the configuration illustrated in FIG. 8 can be used. For example, hardware resource 1000 includes processor 1001 , memory 1002 , network interface 1003 , etc., which are interconnected by internal bus 1004 .

Note that the configuration shown in FIG. 8 is not intended to limit the hardware configuration of the hardware resource 1000 . The hardware resource 1000 may include hardware not shown (for example, an input/output interface). Alternatively, the number of units such as the processors 1001 included in the device is not limited to the illustration in FIG. For the processor 1001, for example, a CPU (Central Processing Unit), MPU (Micro Processor Unit), GPU (Graphics Processing Unit), or the like can be used.

For the memory 1002, for example, RAM (Random Access Memory), ROM (Read Only Memory), HDD (Hard Disk Drive), SSD (Solid State Drive), etc. can be used.

For the network interface 1003, for example, a LAN (Local Area Network) card, network adapter, network interface card, or the like can be used.

The functions of the hardware resource 1000 are implemented by the processing modules described above. The processing module is implemented by the processor 1001 executing a program stored in the memory 1002, for example. Also, the program can be downloaded via a network or updated using a storage medium storing the program. Furthermore, the processing module may be realized by a semiconductor chip. In other words, the functions performed by the above processing modules may be realized by executing software in some kind of hardware.

Some or all of the above embodiments may also be described in the following appendices, but are not limited to the following.

[Appendix 1]
an input unit configured to receive a user's operation;
a monitoring area designation unit configured to designate a monitoring area for three-dimensional point cloud data relating to a site based on designation information from the input unit;
a reference plane determination unit configured to determine a reference plane based on the point cloud within the monitoring area in the three-dimensional point cloud data;
It is configured to selectively detect a detection target object that matches the condition information from the input unit based on the point group on the reference plane among the point groups in the monitoring area in the three-dimensional point group data. a detector,
with
The detection unit is
a detection condition specifying unit configured to specify a detection condition for a detection target object candidate based on the condition information;
a detection processing unit configured to detect a detection target object candidate that satisfies the detection condition based on the point cloud within the monitoring region in the three-dimensional point cloud data;
an exclusion condition specifying unit configured to specify an exclusion condition including a particular form of a detection exclusion object based on the condition information;
The detection target object is determined by excluding a detection exclusion object that meets the specified exclusion condition from the detection target object candidates, based on the point cloud related to the detected detection target object candidate. an exclusion processing unit;
comprising
Object detection device.
[Appendix 2]
The detection processing unit is
a process of deleting point groups in areas other than the monitoring area in the three-dimensional point cloud data, and deleting point groups related to the reference plane from the point clouds in the monitoring area to group the point clouds; ,
A process of detecting an object as one segment and one object based on the grouped point cloud;
A process of detecting an object that satisfies the detection condition as a detection target object candidate based on a point cloud related to the detected object;
is configured to do
The object detection device according to appendix 1.
[Appendix 3]
The exclusion processing unit is configured to exclude all of the detection target object candidates detected by the detection processing unit one by one.
3. The object detection device according to appendix 1 or 2.
[Appendix 4]
The exclusion processing unit
a form detection unit configured to detect the form of each of the detection target object candidates based on the detected point cloud associated with the detection target object candidates;
a determination unit configured to determine whether or not there is a configuration of each of the detected detection target object candidates that satisfies the exclusion condition;
an exclusion unit configured to determine the detection target object by excluding the detection exclusion object determined to meet the exclusion condition from detection target object candidates;
comprising
4. The object detection device according to any one of appendices 1 to 3.
[Appendix 5]
The reference plane determination unit
a plane detector configured to detect one or more planes based on the point cloud within the monitored area in the three-dimensional point cloud data;
a reference plane selection unit configured to select at least one of the reference planes from among the one or more planes based on selection information from the input unit;
comprising
5. The object detection device according to any one of appendices 1 to 4.
[Appendix 6]
a result generation unit configured to generate a result regarding detection of the detection target object using at least the monitoring area, the reference plane, the detection condition, and the exclusion condition;
an output unit configured to output at least the result;
further comprising
6. The object detection device according to any one of Appendices 1 to 5.
[Appendix 7]
The detection processing unit is configured to label a point cloud related to the detection target object,
The result generation unit is configured to determine the presence or absence of the detection target object based on the presence or absence of the label, and to generate warning information when the detection target object is present,
The output unit is configured to output the warning information,
The object detection device according to appendix 6.
[Appendix 8]
The object detection device according to any one of Appendices 1 to 7;
a three-dimensional sensor that is communicatively connected to the object detection device and that captures a scene;
An object detection system comprising:
[Appendix 9]
An object detection method for detecting an object using hardware resources,
Designating a monitoring area for 3D point cloud data related to the site based on the input designation information;
determining a reference plane based on a point cloud within the monitoring area in the three-dimensional point cloud data;
a step of selectively detecting a detection target object that matches input condition information based on the point group on the reference plane among the point groups in the monitoring area in the three-dimensional point group data;
including
In the detecting step,
designating a detection condition for a detection target object candidate based on the condition information;
a step of detecting a detection target object candidate that satisfies the detection condition based on the point cloud within the monitoring area in the three-dimensional point cloud data;
designating an exclusion condition including a particular shape of a part of an object excluded from detection based on the condition information;
determining the detection target object by excluding a detection exclusion object that meets the specified exclusion condition from the detection target object candidates, based on the point cloud related to the detected detection target object candidate;
including,
Object detection method.
[Appendix 10]
A program that causes a hardware resource to execute processing for detecting an object,
A process of designating a monitoring area for 3D point cloud data related to the site based on the input designation information;
a process of determining a reference plane based on the point cloud within the monitoring area in the three-dimensional point cloud data;
a process of selectively detecting a detection target object that matches input condition information based on a point group on the reference plane among the point groups in the monitoring area in the three-dimensional point group data;
causing the hardware resource to execute
In the detecting process,
a process of designating detection conditions for detection target object candidates based on the condition information;
a process of detecting a detection target object candidate that satisfies the detection condition based on the point cloud within the monitoring area in the three-dimensional point cloud data;
a process of designating an exclusion condition including a particular shape of a part of an object excluded from detection based on the condition information;
a process of excluding a detection-excluded object that meets the specified exclusion condition from the detection-target-object candidates based on the point cloud related to the detected detection-target-object candidate, and determining the detection-target object;
run the
program.

It should be noted that each disclosure of the above patent documents is incorporated herein by reference and can be used as the basis or part of the present invention as necessary. Within the framework of the entire disclosure of the present invention (including claims and drawings), modifications and adjustments of the embodiments and examples are possible based on the basic technical concept thereof. Also, various combinations or selections of various disclosure elements (including each element of each claim, each element of each embodiment or example, each element of each drawing, etc.) within the framework of the full disclosure of the present invention (if necessary not selected) is possible. That is, the present invention naturally includes various variations and modifications that can be made by those skilled in the art according to the entire disclosure including the claims and drawings and the technical idea. Also, with regard to numerical values and numerical ranges described in this application, it is assumed that any intermediate values, sub-numerical values and sub-ranges thereof are described even if not specified. Furthermore, each disclosure item of the above-cited document may be used in combination with the items described in this document as a part of the disclosure of the present invention in accordance with the spirit of the present invention, if necessary. are considered to be included in (belong to) the disclosure of the present application.

Reference Signs List 1 object detection system 10 shooting site 11 first track 12 second track 13 third track 14 ground 20 surveillance area 21 worker (object to be detected)
22 Rail vehicle (object to be detected)
23, 24 trains (objects excluded from detection)
25 worker (non-detection object)
31 First track point group 32 Second track point group 33 Third track point group 34 Ground point group 41 Worker point group (detection target object point group)
42 Rail-road vehicle point cloud (detection target object point cloud)
43, 44a, 44b Train point cloud (detection exclusion object point cloud)
45 Worker point cloud (non-detection target object point cloud)
53, 54a, 54b train point cloud (detection target object point cloud)
61 worker detection frame 62 rail/road vehicle detection frame 63, 64a, 64b train detection frame 100 three-dimensional point cloud data 200 object detection device 210 preprocessing unit 211 format conversion unit 212 noise removal unit 220 monitoring area designation unit 221 range designation unit 222 Angle designation unit 230 Reference plane determination unit 231 Plane detection unit 232 Reference plane selection unit 233 Reference plane storage unit 240 Detection unit 241 Detection condition designation unit 242 Detection processing unit 243 Exclusion condition designation unit 244 Exclusion processing unit 251 Form detection unit 252 Determination unit 253 exclusion unit 260 result generation unit 261 specification calculation unit 262 graphic generation unit 270 user interface unit 271 output unit 272 input unit 273 alarm unit 300 three-dimensional sensor 1000 hardware resource 1001 processor 1002 memory 1003 network interface 1004 internal bus

Claims (10)

an input unit configured to receive a user's operation;
a monitoring area designation unit configured to designate a monitoring area for three-dimensional point cloud data relating to a site based on designation information from the input unit;
a reference plane determination unit configured to determine a reference plane based on the point cloud within the monitoring area in the three-dimensional point cloud data;
It is configured to selectively detect a detection target object that matches the condition information from the input unit based on the point group on the reference plane among the point groups in the monitoring area in the three-dimensional point group data. a detector,
with
The detection unit is
a detection condition specifying unit configured to specify a detection condition for a detection target object candidate based on the condition information;
a detection processing unit configured to detect a detection target object candidate that satisfies the detection condition based on the point cloud within the monitoring region in the three-dimensional point cloud data;
an exclusion condition specifying unit configured to specify an exclusion condition including a particular form of a detection exclusion object based on the condition information;
The detection target object is determined by excluding a detection exclusion object that meets the specified exclusion condition from the detection target object candidates, based on the point cloud related to the detected detection target object candidate. an exclusion processing unit;
comprising
Object detection device. The detection processing unit is
a process of deleting point groups in areas other than the monitoring area in the three-dimensional point cloud data, and deleting point groups related to the reference plane from the point clouds in the monitoring area to group the point clouds; ,
A process of detecting an object as one segment and one object based on the grouped point cloud;
A process of detecting an object that satisfies the detection condition as a detection target object candidate based on a point cloud related to the detected object;
is configured to do
The object detection device according to claim 1. The exclusion processing unit is configured to exclude all of the detection target object candidates detected by the detection processing unit one by one.
3. The object detection device according to claim 1 or 2. The exclusion processing unit
a form detection unit configured to detect the form of each of the detection target object candidates based on the detected point cloud associated with the detection target object candidates;
a determination unit configured to determine whether or not there is a configuration of each of the detected detection target object candidates that satisfies the exclusion condition;
an exclusion unit configured to determine the detection target object by excluding the detection exclusion object determined to meet the exclusion condition from detection target object candidates;
comprising
The object detection device according to any one of claims 1 to 3. The reference plane determination unit
a plane detector configured to detect one or more planes based on the point cloud within the monitored area in the three-dimensional point cloud data;
a reference plane selection unit configured to select at least one of the reference planes from among the one or more planes based on selection information from the input unit;
comprising
The object detection device according to any one of claims 1 to 4. a result generation unit configured to generate a result regarding detection of the detection target object using at least the monitoring area, the reference plane, the detection condition, and the exclusion condition;
an output unit configured to output at least the result;
further comprising
The object detection device according to any one of claims 1 to 5. The detection processing unit is configured to label a point cloud related to the detection target object,
The result generation unit is configured to determine the presence or absence of the detection target object based on the presence or absence of the label, and to generate warning information when the detection target object is present,
The output unit is configured to output the warning information,
The object detection device according to claim 6. an object detection device according to any one of claims 1 to 7;
a three-dimensional sensor that is communicatively connected to the object detection device and that captures a scene;
An object detection system comprising: An object detection method for detecting an object using hardware resources,
Designating a monitoring area for 3D point cloud data related to the site based on the input designation information;
determining a reference plane based on a point cloud within the monitoring area in the three-dimensional point cloud data;
a step of selectively detecting a detection target object that matches input condition information based on the point group on the reference plane among the point groups in the monitoring area in the three-dimensional point group data;
including
In the detecting step,
designating a detection condition for a detection target object candidate based on the condition information;
a step of detecting a detection target object candidate that satisfies the detection condition based on the point cloud within the monitoring area in the three-dimensional point cloud data;
designating an exclusion condition including a particular shape of a part of an object excluded from detection based on the condition information;
determining the detection target object by excluding a detection exclusion object that meets the specified exclusion condition from the detection target object candidates, based on the point cloud related to the detected detection target object candidate;
including,
Object detection method. A program that causes a hardware resource to execute processing for detecting an object,
A process of designating a monitoring area for 3D point cloud data related to the site based on the input designation information;
a process of determining a reference plane based on the point cloud within the monitoring area in the three-dimensional point cloud data;
a process of selectively detecting a detection target object that matches input condition information based on a point group on the reference plane among the point groups in the monitoring area in the three-dimensional point group data;
causing the hardware resource to execute
In the detecting process,
a process of designating detection conditions for detection target object candidates based on the condition information;
a process of detecting a detection target object candidate that satisfies the detection condition based on the point cloud within the monitoring area in the three-dimensional point cloud data;
a process of designating an exclusion condition including a particular shape of a part of an object excluded from detection based on the condition information;
a process of excluding a detection-excluded object that meets the specified exclusion condition from the detection-target-object candidates based on the point cloud related to the detected detection-target-object candidate, and determining the detection-target object;
run the
program.

Images