WO2023139706A1 - Dispositif de traitement - Google Patents

Dispositif de traitement Download PDF

Info

Publication number
WO2023139706A1
WO2023139706A1 PCT/JP2022/001838 JP2022001838W WO2023139706A1 WO 2023139706 A1 WO2023139706 A1 WO 2023139706A1 JP 2022001838 W JP2022001838 W JP 2022001838W WO 2023139706 A1 WO2023139706 A1 WO 2023139706A1
Authority
WO
WIPO (PCT)
Prior art keywords
processing
trajectory
trajectories
searching
vectorization
Prior art date
Application number
PCT/JP2022/001838
Other languages
English (en)
Japanese (ja)
Inventor
裕太 井手口
貴志 竹内
Original Assignee
日本電気株式会社
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 日本電気株式会社 filed Critical 日本電気株式会社
Priority to PCT/JP2022/001838 priority Critical patent/WO2023139706A1/fr
Priority to JP2023574958A priority patent/JPWO2023139706A1/ja
Publication of WO2023139706A1 publication Critical patent/WO2023139706A1/fr

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
    • G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
    • G01C21/26—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 specially adapted for navigation in a road network
    • G01C21/34—Route searching; Route guidance
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles

Definitions

  • the present invention relates to processing apparatuses, processing methods, and recording media.
  • the trajectories of moving objects such as cars, ships, airplanes, and unmanned guided vehicles may be determined.
  • Non-Patent Document 1 and Non-Patent Document 2 describe techniques for determining the trajectory of a moving object such as a ship's trajectory.
  • the techniques described in Non-Patent Literature 1 and Non-Patent Literature 2 after classifying the track hypotheses, global hypotheses are generated, and track hypotheses are selected based on the generated global hypotheses.
  • Non-Patent Document 3 for example, describes a method for generating global hypotheses.
  • Patent Document 1 describes an example of technology used to detect and track an object based on sensor data.
  • Non-Patent Documents 1 to 3 When generating global hypotheses as described in Non-Patent Documents 1 to 3, trajectories and trajectories that can be common to the trajectory are sometimes searched.
  • trajectories and trajectories that can be common to the trajectory are sometimes searched.
  • the number of moving bodies to be tracked increases, the number of necessary processes increases by the number of combinations. As a result, the processing time has increased.
  • the tracking of moving objects as described above may require processing in real time, so it is desirable that the processing be as fast as possible.
  • combinatorial processing such as searching for common possible trajectories, is complicated. Therefore, the efficiency of parallelization was poor and speeding up was difficult.
  • an object of the present invention is to provide a processing device, a processing method, and a recording medium that can solve the problem that it is difficult to speed up combination processing such as searching for sharable trajectories.
  • an acquisition unit that acquires information indicating a trajectory, which is a route traveled by the mobile object, that is specified based on a sensing result of the mobile object;
  • a vectorization unit that vectorizes a process of searching for a combination of trajectories acquired by the acquisition unit and causes a vector processing unit to execute the process; It has a configuration of
  • a processing method that is another aspect of the present disclosure includes: The information processing device Acquiring information indicating a trajectory, which is a route traveled by the moving body, which is specified based on the sensing result of the moving body; A configuration is adopted in which the process of searching for combinations of acquired trajectories is vectorized and executed by the vector processing unit.
  • a recording medium that is another aspect of the present disclosure includes: information processing equipment, Acquiring information indicating a trajectory, which is a route traveled by the moving body, which is specified based on the sensing result of the moving body;
  • a computer-readable recording medium recording a program for realizing a process of vectorizing a process of searching for a combination of acquired trajectories and causing a vector processing unit to execute the process.
  • each configuration as described above it is possible to provide a processing device, a processing method, and a recording medium capable of speeding up combination processing such as searching for sharable trajectories.
  • FIG. 10 is a diagram showing an example of determining a trajectory and a common possible trajectory in the first stage; It is a figure which shows an example of vectorization.
  • FIG. 10 is a diagram showing another example of vectorization;
  • FIG. 10 is a diagram showing another example of vectorization;
  • It is a figure which shows an example of a determination condition.
  • FIG. 4 is a diagram showing a detailed example of vectorization in the first method;
  • FIG. 4 is a diagram showing a detailed example of vectorization in the first method
  • FIG. 10 is a diagram showing a detailed example of vectorization in the second method
  • FIG. 10 is a diagram showing a detailed example of vectorization in the third method
  • It is a flowchart which shows the operation example of a processing apparatus.
  • FIG. 11 is a block diagram showing a configuration example of a processing device according to a second embodiment of the present disclosure
  • FIG. 1 and 2 are diagrams for explaining the outline of the present invention.
  • FIG. 3 is a block diagram showing a configuration example of the processing device 100.
  • FIG. 4 is a diagram showing an example of determination of a trajectory and a common possible trajectory in the first stage.
  • 5 to 7 are diagrams showing an example of vectorization.
  • FIG. 8 is a diagram showing an example of determination conditions.
  • 9 and 10 are diagrams showing detailed examples of vectorization in the first method.
  • FIG. 11 is a diagram showing a detailed example of vectorization in the second method.
  • FIG. 12 is a diagram showing a detailed example of vectorization in the third method.
  • FIG. 13 is a flow chart showing an operation example of the processing device 100 .
  • a processing device 100 that is an information processing device that searches for a combination of simultaneously selectable trajectories will be described.
  • FIG. 1 it is assumed that three moving objects are observed in one frame (frame 1) and three moving objects are observed in the next frame (frame 2) using sensors such as radio waves and sound waves (the number of moving objects can be arbitrary).
  • mobile 1 in frame 1 is assumed to correspond to any of mobiles 1, 2, 3 in frame 2.
  • trajectory 11 formed by moving body 1 of frame 1 to moving body 1 of frame 2 trajectory 12 formed by moving body 1 of frame 1 to moving body 2 of frame 2 is assumed to be formed.
  • trajectory 13 formed by moving body 1 of frame 1 to moving body 3 of frame 2 is assumed to be formed. The same can be said for moving bodies 2 and 3 in frame 1 .
  • the processing device 100 described in this embodiment searches for combinations of trajectories as described above without omission by the method illustrated in FIG.
  • the processing device 100 focuses on a certain trajectory and extracts common possible trajectories that are trajectories that can be shared with the focused trajectory.
  • the processing device 100 performs the above-described extraction process so as not to extract the same combination that is only different in order.
  • the processing device 100 focuses on the trajectory 22 and extracts trajectories 11 , 33 , 13 , and 31 as common possible trajectories that can be shared with the trajectory 22 .
  • the processing device 100 extracts common possible trajectories by focusing on the trajectory 11 .
  • the processing device 100 extracts common possible trajectories so as not to extract the same combination as the combination extracted by focusing on the trajectory 22 .
  • the combination of “trajectory 22, trajectory 11” has already been extracted at the stage where attention is focused on trajectory 22 . Therefore, the processing device 100 extracts the trajectory 33, the trajectory 23, and the trajectory 32 as common possible trajectories that can be shared with the trajectory 11, but does not extract the trajectory 22 that is the same combination.
  • the processing device 100 extracts common possible trajectories so as not to extract the same combination by repeating the above-described processing focusing on each trajectory as the first-stage processing.
  • “Moving objects” refer to objects that move, such as cars, ships, airplanes, and unmanned guided vehicles.
  • the moving body may be other than those exemplified above.
  • a common possible trajectory refers to, for example, a trajectory that can physically coexist (exist).
  • trajectory 11 and trajectory 22 can coexist, but trajectory 11 and trajectory 13 cannot coexist because there are two moving bodies 1 in frame 1, and are not considered common possible trajectories.
  • Common possible trajectories may be partially limited in advance in preprocessing or the like, such as trajectories whose movement distance is within a threshold value among trajectories that can physically coexist, or trajectories formed between moving bodies having a predetermined feature amount.
  • the order of trajectories that the processing device 100 pays attention to may be determined arbitrarily.
  • the processing device 100 extracts common possible trajectories that are further commonable trajectories to each trajectory indicated by the common possible trajectories extracted in the first-stage processing, so as not to extract the same combination as in the first-stage processing. Further, after the processing of searching for combinations in the second stage is completed, the processing device 100 extracts, as the processing of the third stage, common possible trajectories that are further commonable trajectories indicated by the common possible trajectories extracted in the second stage processing. After that, the processing device 100 repeats the same processing until no more combinations are extracted.
  • the processing device 100 described in the present embodiment vectorizes the second and subsequent stages of extraction processing and causes the vector processing unit 130 such as a vector engine to perform the extraction processing. That is, the processing device 100 vectorizes the processing of searching for combinations in the second and subsequent stages, and causes the vector processing unit 130 to perform the processing. Further, as described above, the processing device 100 performs extraction processing so as not to extract the same combination. Therefore, the number of possible common trajectories decreases as the stage becomes later, or even in the same stage, as the attention is focused later.
  • the processing device 100 can change the vectorization method according to the processing status of searching for a combination of the order of attention, the number of stages of processing, the number of common possible trajectories in a given trajectory, and the like. It should be noted that specific examples of the judgment when changing the vectorization method and the vectorization method will be described later.
  • FIG. 3 shows a configuration example of the processing device 100.
  • the processing device 100 has, for example, a storage unit 110, an arithmetic processing unit 120, and a vector processing unit .
  • the processing device 100 is communicably connected to an external device such as a sensor that detects a moving object using sound waves or radio waves.
  • the processing device 100 can receive the result of sensing by the sensor or the like from an external device such as a sensor.
  • the storage unit 110 is a storage device such as a hard disk or memory.
  • the storage unit 110 stores processing information and programs 112 necessary for various processes in the arithmetic processing unit 120 and the like.
  • the program 112 realizes various processing units by being read and executed by the arithmetic processing unit 120 .
  • the program 112 is preloaded from an external device or recording medium via the data input/output function of the processing device 100 and stored in the storage unit 110 .
  • Main information stored in the storage unit 110 includes sensing information 111, for example.
  • the sensing information 111 includes information indicating the position of the moving object detected by an external device such as a sensor (for example, position information such as latitude and longitude information). For example, the sensing information 111 includes information indicating the positions of multiple moving objects in multiple frames. The sensing information 111 is updated each time information indicating a sensing result is received from an external device such as a sensor.
  • the arithmetic processing unit 120 has an arithmetic device such as a processor and its peripheral circuits.
  • the arithmetic processing unit 120 reads the program 112 from the storage unit 110 and executes it, so that the hardware and the program 112 cooperate to realize various processing units.
  • Main processing units realized by the arithmetic processing unit 120 include, for example, a trajectory determination unit 121 , a vectorization method determination unit 122 , and a vectorization execution unit 123 .
  • at least the vectorization method determination unit 122 and the vectorization execution unit 123 may include, for example, a compiler realized by the arithmetic processing unit 120 reading and executing the program 112 or the like.
  • the trajectory determination unit 121 identifies, for example, a plurality of trajectories such as paths traveled by the moving body based on the position information of the moving body in multiple frames indicated by the sensing information 111 . Then, the trajectory determination unit 121 performs the above-described first-stage processing on the plurality of identified trajectories. That is, the trajectory determination unit 121 performs processing for each specified trajectory to extract a common possible trajectory in the trajectory focused on an arbitrary trajectory. At this time, the trajectory determination unit 121 extracts common possible trajectories so as not to extract the same combination that differs only in order. Further, the trajectory determination unit 121 adds each trajectory indicated by the extracted common possible trajectories to the processing targets for the next common possible trajectory search.
  • FIG. 4 shows a processing example of the trajectory determination unit 121 when three moving bodies are detected in each of frames 1 and 2 as illustrated in FIG.
  • the trajectory determination unit 121 focuses on the trajectory 22, for example.
  • the trajectory determination unit 121 extracts a common possible trajectory that is a trajectory that can be shared with the trajectory 22 .
  • the trajectory determination unit 121 extracts trajectory 11, trajectory 33, trajectory 13, and trajectory 31 as common possible trajectories.
  • the trajectory determination unit 121 focuses on the trajectory 11 .
  • the trajectory determination unit 121 extracts a common possible trajectory, which is a trajectory that can be shared with the trajectory 11, so as not to extract the same combination as an already extracted combination.
  • the trajectory determination unit 121 extracts trajectories 33, 23, and 32 as common possible trajectories, but does not extract trajectory 22 that is the same combination as the already extracted combination.
  • the trajectory determination unit 121 extracts common possible trajectories so as not to extract the same combination by repeating the same process focusing on each specified trajectory, such as trajectory 33, trajectory 21, trajectory 23, . After that, the trajectory determination unit 121 adds each trajectory indicated by the extracted common possible trajectories to the processing targets for the next common possible trajectory search.
  • the trajectory determination unit 121 does not extract the same combination as the already extracted combination. Therefore, as shown in FIG. 4, there is a possibility that the same combination of sharable trajectories has been previously extracted for trajectories that are focused on later, and the number of sharable trajectories that are actually extracted decreases. For example, in the case illustrated in FIG. 4 , the trajectory determination unit 121 does not extract a sharable trajectory when focusing on the trajectory 31 . This is because all combinations of trajectories 31 and sharable trajectories have already been extracted before focusing on trajectory 31 .
  • the trajectory determination unit 121 may be configured to perform the above-described first-stage processing on the trajectory that has undergone predetermined preprocessing by the preprocessing unit realized by, for example, the arithmetic processing unit 120 reading and executing the program 112. For example, in the pre-processing, among trajectories that can physically coexist, a process of limiting common possible trajectories that can be shared with a certain trajectory, such as a trajectory whose moving distance is within a threshold or a trajectory formed between moving bodies having a predetermined feature amount, may be performed.
  • the trajectory determination unit 121 which functions as an acquisition unit that acquires information indicating a trajectory, may acquire information indicating a trajectory based on the position information of a moving object in multiple frames indicated by the sensing information 111, or may acquire information indicating a trajectory from another processing unit such as a preprocessing unit.
  • the vectorization method determination unit 122 is a determination unit that determines a method for vectorizing the process of searching for a combination for each one or more processing targets. For example, the vectorization method determination unit 122 determines which of the predetermined methods should be used to vectorize the process of searching for a combination based on the processing status of the process of searching for a combination of trajectories, such as the order of attention, the number of stages of processing, and the number of common possible trajectories in a given trajectory. For example, the vectorization method determination unit 122 determines whether to vectorize the combination search process using any one of the first method, the second method, and the third method, based on the processing status of the process of searching for a combination of trajectories.
  • the process of searching for a common possible trajectory and adding it to the next processing target is vectorized. That is, in the first method, the process of searching for combinations of trajectories indicated by common possible trajectories corresponding to one processing target is vectorized.
  • the first stage which is the preceding stage
  • trajectories 11, 33, 13, and 31 are extracted as common possible trajectories for trajectory 22.
  • FIG. Therefore, in the first method, common possible trajectories are further searched for each of trajectory 11, trajectory 33, trajectory 13, and trajectory 31, and each trajectory indicated by the searched common possible trajectory is vectorized as the next processing target.
  • FIG. 6 shows an overview of the second method.
  • the process of searching for combinations of trajectories indicated by common possible trajectories corresponding to a plurality of processing targets is vectorized.
  • trajectories 11, 33, 13, and 31 are extracted as common possible trajectories for trajectory 22.
  • FIG. Further, in the case of FIG.
  • the trajectory 33, the trajectory 23, and the trajectory 32 are extracted as common possible trajectories of the trajectory 11 in the first stage. Therefore, in the second method, trajectories 11, 33, 13, and 31 that can be common to trajectory 22, and trajectories 33, 23, and 32 that can be common to trajectory 11 are further searched for common trajectories, and the process of adding each trajectory indicated by the searched common trajectory to the next processing target is vectorized.
  • the number of trajectories to be collectively processed in the second method may be set arbitrarily. For example, the number of trajectories to be collectively processed in the second method may be set according to the total number of common possible trajectories corresponding to the respective trajectories to be collectively processed.
  • FIG. 7 shows an overview of the third method.
  • the third method combines some of the processes of the first method. For example, in the third method, for each trajectory indicated by the common possible trajectory corresponding to one certain trajectory in the previous stage, the processing for searching for common possible trajectories is performed multiple times, and then the processing for searching for combinations is vectorized so that the processing for adding to the next processing target is performed collectively. In this way, in the third method, the process of searching for a combination is vectorized so that after performing the process of searching for a combination in the first method a plurality of times, the process of adding each trajectory indicated by the searched common possible trajectory to the next processing target is collectively performed.
  • the vectorization method determination unit 122 determines which of the above-described first method, second method, and third method should be used to vectorize the combination search process based on the processing status of the combination search process.
  • FIG. 8 shows a diagram for explaining an example of determination based on the processing status.
  • the vectorization method determination unit 122 can determine which of the first method and the second method should be used for vectorization based on the number of stages of processing.
  • the vectorization method determination unit 122 determines to vectorize by the first method when the number of processing stages is less than a predetermined value, and determines to vectorize by the second method when the number of processing stages is equal to or greater than the predetermined value. In other words, the vectorization method determination unit 122 determines to vectorize the process of searching for a combination by the first method when it is assumed that the number of trajectories to be searched (that is, the number of sharable trajectories in the previous stage) is sufficiently large based on the processing status such as the number of processing stages.
  • the vectorization method determination unit 122 determines to vectorize the processing of searching for combinations by the second method. In this way, the vectorization method determination unit 122 can determine which of the first method and the second method should be used for vectorization based on the amount of processing determined based on the processing status such as the number of stages of processing.
  • the vectorization method determination unit 122 can determine which of the first method and the third method should be used for vectorization based on the order of attention. For example, the vectorization method determination unit 122 determines to vectorize by the first method when the order of attention is less than a predetermined value, and determines to vectorize by the third method when the order of attention is equal to or greater than the predetermined value. In other words, the vectorization method determination unit 122 determines to vectorize the process of searching for combinations by the first method when it is assumed that there are sufficiently many trajectories to be searched based on the order of attention.
  • the vectorization method determination unit 122 determines to vectorize the process of searching for combinations by the third method. In this way, the vectorization method determination unit 122 can determine which of the first method and the third method should be used for vectorization based on the amount of processing determined based on the processing status such as the order of attention.
  • the vectorization method determination unit 122 can determine whether to use any of the above determination methods or a combination of the above determination methods to vectorize the process of searching for a combination using the first method, the second method, or the third method.
  • the vectorization method determination unit 122 may determine, for example, based on the number of stages of processing, which of the first method and the third method should be used for vectorization, and whether to use the first method or the second method for vectorization based on the number of sharable trajectories in the previous stage.
  • the timing at which the vectorization method determination unit 122 makes the above determination may be set arbitrarily.
  • the vectorization method determination unit 122 can determine for each stage which method should be used for vectorization. That is, the vectorization method determination unit 122 can collectively determine which method is used for vectorization for a certain stage.
  • the vectorization method determination unit 122 may determine which method should be used to vectorize the process of searching for a combination of sharable trajectories corresponding to the next one or more trajectories each time a process corresponding to the result of determining that vectorization should be performed using a certain method is completed.
  • the vectorization execution unit 123 vectorizes the process of searching for combinations according to the determination of the vectorization method determination unit 122 . Then, the vectorization execution unit 123 causes the vector processing unit 130 to execute vector processing. It should be noted that the method described in this embodiment is a process of confirming common combinations, so simple vectorization is not possible. Therefore, the vectorization execution unit 123 performs vectorization using a buffer.
  • FIG. 9 is a diagram for explaining in more detail the process of vectorization by the first method.
  • the vectorization execution unit 123 first prepares a buffer corresponding to the trajectory number in the previous stage.
  • the first stage which is the preceding stage, includes trajectories 22, 11, 33, 21, 23, 12, 32, 13, and 31. Therefore, the vectorization execution unit 123 prepares buffers corresponding to the numbers of the trajectories in the storage unit 110 or the like.
  • the vectorization execution unit 123 also performs vectorization by generating an instruction for causing the vector processing unit 130 to perform the following processing, for example.
  • • 1 is stored in the buffer corresponding to the common possible trajectory in the previous stage among the prepared buffers. ⁇ Check if there is 1 in the buffer indicated by the preceding common possible trajectory corresponding to the trajectory to be judged, and if there is 1, the buffer adds the number of 1 to the common possible trajectory frame of the trajectory to be judged. ⁇ Perform the same judgment process for each trajectory.
  • the commands are as follows.
  • 1 is stored in buffers corresponding to trajectory 11, trajectory 33, trajectory 13, and trajectory 31, so there is 1 in buffer 33.
  • ⁇ indicating that there is no common possible trajectory is added to the common possible trajectory frame of trajectory 33, which is the determination target.
  • the buffer adds the number of 1 to the common possible trajectory frame of the trajectory to be determined.
  • 1 is stored in buffers corresponding to trajectory 11, trajectory 33, trajectory 13, and trajectory 31, so there is 1 in buffer 31. Therefore, 31 is added to the common possible trajectory frame of trajectory 13, which is the determination target.
  • - The preceding common possible trajectory corresponding to the trajectory 31 to be determined indicates that there is no combination of the trajectory 31 and the sharable trajectory. Therefore, no further confirmation is performed.
  • the trajectory 33 and the trajectory 31 indicated by the number added to the common possible trajectory frame are to be processed next.
  • the vectorization execution unit 123 When preparing the buffer, the vectorization execution unit 123 prepares the buffer corresponding to the trajectory number in the previous stage. At this time, the vectorization execution unit 123 can prepare a buffer corresponding to the trajectory number determined by the same processing (processing for each trajectory when multiple trajectories are collectively processed) in the previous stage.
  • FIG. 10 shows an example of a buffer prepared by the vectorization execution unit 123 in the third stage. Referring to FIG. 10 , when searching for common possible trajectories for trajectory 33 and trajectory 31 in the processing of the third stage, the trajectories determined by the same processing in the previous stage are trajectories 11, 33, 13, and 31.
  • the vectorization execution unit 123 can prepare buffers corresponding to the numbers of the trajectories in the storage unit 110 or the like. Further, when searching for a common possible trajectory for trajectory 32 in the processing of the third stage, trajectories 33, 23, and 32 are the trajectories determined by the same processing in the previous stage. Therefore, the vectorization execution unit 123 can prepare buffers corresponding to the numbers of the trajectories in the storage unit 110 or the like.
  • FIG. 11 is a diagram for explaining in more detail the process for vectorization by the second method.
  • the vectorization execution unit 123 first prepares a number of buffers corresponding to the number of the trajectory in the preceding stage and corresponding to the number of vector lengths corresponding to the number of trajectories to be processed simultaneously.
  • the first stage which is the preceding stage, includes trajectories 22, 11, 33, 21, 23, 12, 32, 13, and 31.
  • two trajectories are processed at the same time as an example. Therefore, as shown in FIG. 11, the vectorization execution unit 123 prepares a two-line buffer in the storage unit 110 or the like corresponding to each trajectory number.
  • the vectorization execution unit 123 also performs vectorization by generating an instruction for causing the vector processing unit 130 to perform the following processing, for example.
  • For each row of the buffer it is checked whether there is 1 in the buffer indicated by the preceding common possible trajectory corresponding to the trajectory to be determined, and if there is 1, the buffer adds the number of 1 to the common possible trajectory frame of the trajectory to be determined. ⁇ Similar determination processing is performed for each trajectory. ⁇ The trajectory indicated by the number added to the common possible trajectory frame is the next processing target.
  • the following instructions are used.
  • 1 is stored in the buffers corresponding to trajectories 11, 33, 13, and 31, which are sharable trajectories in the previous stage.
  • 1 is stored in the buffer corresponding to the trajectory 33, trajectory 23, and trajectory 32, which are the sharable trajectories in the previous stage.
  • ⁇ Check if there is a 1 in 33 of the first line of the buffer and 21 of the second line. (In this case, there is 1 in the buffer of 33 on the first line.
  • FIG. 12 is a diagram for explaining in more detail the process of vectorization by the third method.
  • vectorization is performed by generating an instruction indicating that the next processing target is to be added at the same time based on the results of the determination processing for the plurality of trajectories.
  • the vectorization execution unit 123 performs vectorization by generating instructions according to each method as described above according to the determination of the vectorization method determination unit 122 . Then, the vectorization execution unit 123 causes the vector processing unit 130 to execute vector processing using the generated instructions.
  • the vector processing unit 130 has a vector processor and the like, and executes vector processing according to instructions from the vectorization execution unit 123 .
  • the vector processing unit 130 may be a known vector engine or the like.
  • the above is a configuration example of the processing device 100 .
  • an operation example of the processing device 100 will be described with reference to FIG. 13 .
  • FIG. 13 is a flowchart showing an operation example of the processing device 100.
  • the trajectory determination unit 121 identifies, for example, a plurality of trajectories such as paths traveled by the moving object, based on the position information of the moving object in multiple frames indicated by the sensing information 111 . Then, the trajectory determination unit 121 performs the first-stage process on the plurality of specified trajectories. That is, the trajectory determination unit 121 extracts a common possible trajectory for the first stage (step S101).
  • the vectorization method determination unit 122 determines which of the predetermined methods should be used to vectorize the combination search process based on the processing status of the trajectory combination search process (step S102). For example, the vectorization method determination unit 122 determines whether to vectorize the combination search process using any one of the first method, the second method, and the third method, based on the processing status of the process of searching for a combination of trajectories.
  • the vectorization execution unit 123 vectorizes the process of searching for combinations according to the determination of the vectorization method determination unit 122 . Then, the vectorization execution unit 123 causes the vector processing unit 130 to execute vector processing (step S103).
  • step S104 After the vector processing in step S103, when the processing target is finished (step S104, Yes), the processing device 100 finishes the processing. On the other hand, if there is a process target (step S104, No), the vectorization method determination unit 122 returns to the determination of step S102.
  • the above is an operation example of the processing device 100 .
  • the processing device 100 has the vectorization execution unit 123 .
  • the vectorization execution unit 123 can vectorize the process of searching for a combination of trajectories and cause the vector processing unit 130 to perform the vectorization.
  • the process of searching for a combination of trajectories as described above is a memory bottleneck process. Therefore, the processing can be speeded up by causing the vector processing unit 130 with high memory access speed to perform the processing.
  • the processing device 100 also has a vectorization method determination unit 122 .
  • the vectorization execution unit 123 can execute vectorization by the method determined by the vectorization method determination unit 122 according to the processing status of the process of searching for a combination of trajectories.
  • the efficiency of vector processing can be improved by performing vectorization by a method determined according to the processing status of processing for searching for a combination of trajectories.
  • the processing device 100 described in the present embodiment searches for combinations of simultaneously selectable trajectories, as described above.
  • the result of the search performed by the processing device 100 can be used, for example, when tracking a moving object.
  • the result of the search performed by the processing device 100 may be used in situations other than tracking the moving object.
  • FIG. 14 is a diagram showing a hardware configuration example of the processing device 200.
  • FIG. 15 is a block diagram showing a configuration example of the processing device 200. As shown in FIG.
  • FIG. 14 shows a hardware configuration example of the processing device 200 .
  • the processing device 200 has the following hardware configuration as an example.
  • a processor CPU: Central Processing Unit
  • 201 that functions as a processing management unit that vectorizes processing and causes the vector processing unit to process it.
  • a vector processor 202 that functions as a vector processing unit that performs vector processing - Storage device 203 for storing program group 204
  • the processing device 200 may include a ROM (Read Only Memory), a RAM (Random Access Memory), a drive device for reading and writing to a recording medium outside the information processing device, a communication interface for connecting to a communication network outside the information processing device, an input/output interface for inputting and outputting data, a bus for connecting each component, and the like.
  • the processor 201 acquires the program group 204 and the processor 201 executes it, so that the processing device 200 can realize the functions of the acquisition unit 221 and the vectorization unit 222 shown in FIG.
  • the program group 204 is stored in advance in, for example, the storage device 203 or the ROM of the processing device 200, and is loaded into the RAM or the like by the processor 201 as necessary and executed.
  • the program group 204 may be supplied to the processor 201 via a communication network, or may be stored in a recording medium in advance, and the drive device may read the program and supply it to the processor 201 .
  • FIG. 14 shows a hardware configuration example of the processing device 200 .
  • the hardware configuration of the processing device 200 is not limited to that described above.
  • the acquisition unit 221 acquires information indicating a trajectory, which is a route traveled by the mobile body, specified based on the sensing result of the mobile body. For example, the acquisition unit 221 may acquire information indicating the trajectory based on the position information of the moving object indicated by the sensing result of the moving object.
  • the vectorization unit 222 vectorizes the process of searching for a combination of trajectories acquired by the acquisition unit 221 and causes the vector processing unit to execute it.
  • the vectorization unit 222 may vectorize the process of searching for a combination of trajectories by generating an instruction for extracting a common possible trajectory, which is a trajectory that can be shared with a trajectory, using a buffer.
  • the processing device 200 has the vectorization section 222 .
  • the vectorization unit 222 can vectorize the process of searching for a combination of trajectories and cause the vector processing unit to perform the vectorization.
  • the process of searching for a combination of trajectories as described above is a memory bottleneck process. Therefore, the processing can be speeded up by allowing the vector processing unit with high memory access speed to perform the processing.
  • a program that is another aspect of the present invention is a program that acquires information indicating a trajectory, which is a route traveled by a mobile body, specified based on the sensing result of the mobile body, vectorizes a process of searching for a combination of the acquired trajectories, and causes a vector processing unit to execute the process.
  • the processing method executed by an information processing device such as the processing device 200 described above is a method of acquiring information indicating a trajectory, which is a route traveled by a moving object, which is specified based on the sensing result of the moving object, vectorizing a process of searching for a combination of the acquired trajectories, and causing a vector processing unit to execute the processing.
  • Appendix 1 an acquisition unit that acquires information indicating a trajectory, which is a route traveled by the mobile object, that is specified based on a sensing result of the mobile object; a vectorization unit that vectorizes a process of searching for a combination of trajectories acquired by the acquisition unit and causes a vector processing unit to execute the process;
  • Appendix 3 The processing apparatus according to Appendix 1 or Appendix 2, a determination unit that determines whether to vectorize the process of searching for a combination of trajectories using a predetermined method based on the processing status of the process of searching for a combination of trajectories; The processing device, wherein the vectorization unit vectorizes a process of searching for a combination of trajectories according to the determination result of the determination unit.
  • the processing device determines which of a first method of vectorizing a process of searching for a combination of trajectories indicated by a common possible trajectory corresponding to one processing target and a second method of vectorizing a process of searching for a combination of each trajectory indicated by the common possible trajectories corresponding to a plurality of processing targets based on the processing status of the processing of searching for a combination of trajectories.
  • the processing device determines which of the first method and the second method is used for vectorization according to the amount of processing determined based on the processing status of processing for searching for a combination of trajectories. Processing device.
  • the determination unit determines which of a first method of vectorizing a process of searching for a combination of trajectories indicated by a common possible trajectory corresponding to one processing target, and a third method of vectorizing a process of performing a process of collectively adding processing targets according to the search result after performing multiple searches for combinations of each trajectory indicated by the common possible trajectory corresponding to one processing target.
  • the information processing device Acquiring information indicating a trajectory, which is a route traveled by the moving object, which is specified based on the sensing result of the moving object; A processing method in which the process of searching for combinations of acquired trajectories is vectorized and executed by the vector processing unit.
  • information processing equipment Acquiring information indicating a trajectory, which is a route traveled by the moving object, which is specified based on the sensing result of the moving object;
  • a computer-readable recording medium recording a program for vectorizing a process of searching for a combination of acquired trajectories and causing the vector processing unit to execute the process.
  • processing device 110 storage unit 111 sensing information 112 program 120 arithmetic processing unit 121 trajectory determination unit 122 vectorization method determination unit 123 vectorization execution unit 130 vector processing unit 200 processing device 201 processor 202 vector processor 203 storage device 204 program group 221 acquisition unit 222 vectorization unit

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Mathematical Physics (AREA)
  • Mathematical Analysis (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Computing Systems (AREA)
  • Automation & Control Theory (AREA)
  • Algebra (AREA)
  • Databases & Information Systems (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Navigation (AREA)

Abstract

La présente invention concerne un dispositif de traitement (200) comprenant : une unité d'acquisition (221) qui acquiert des informations indiquant des trajectoires qui ont été identifiées sur la base des résultats de détection de corps mobiles et qui sont les trajets sur lesquels les corps mobiles se sont déplacés ; et une unité de vectorisation (222) qui vectorise un traitement pour rechercher des combinaisons des trajectoires acquises par l'unité d'acquisition (221), et qui amène une unité de traitement de vectorisation à exécuter un tel traitement.
PCT/JP2022/001838 2022-01-19 2022-01-19 Dispositif de traitement WO2023139706A1 (fr)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/JP2022/001838 WO2023139706A1 (fr) 2022-01-19 2022-01-19 Dispositif de traitement
JP2023574958A JPWO2023139706A1 (fr) 2022-01-19 2022-01-19

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/JP2022/001838 WO2023139706A1 (fr) 2022-01-19 2022-01-19 Dispositif de traitement

Publications (1)

Publication Number Publication Date
WO2023139706A1 true WO2023139706A1 (fr) 2023-07-27

Family

ID=87348264

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2022/001838 WO2023139706A1 (fr) 2022-01-19 2022-01-19 Dispositif de traitement

Country Status (2)

Country Link
JP (1) JPWO2023139706A1 (fr)
WO (1) WO2023139706A1 (fr)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH11120166A (ja) * 1997-10-20 1999-04-30 Fujitsu Ltd ベクトル化装置および記録媒体
JP2019530608A (ja) * 2016-09-29 2019-10-24 ザ・チャールズ・スターク・ドレイパー・ラボラトリー・インコーポレイテッド 物体レベル融合の自律走行車両
JP2020015493A (ja) * 2018-07-17 2020-01-30 バイドゥ ユーエスエイ エルエルシーＢａｉｄｕ ＵＳＡ ＬＬＣ 自動運転車両のためのオブジェクト移動を予測するための方法およびシステム

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH11120166A (ja) * 1997-10-20 1999-04-30 Fujitsu Ltd ベクトル化装置および記録媒体
JP2019530608A (ja) * 2016-09-29 2019-10-24 ザ・チャールズ・スターク・ドレイパー・ラボラトリー・インコーポレイテッド 物体レベル融合の自律走行車両
JP2020015493A (ja) * 2018-07-17 2020-01-30 バイドゥ ユーエスエイ エルエルシーＢａｉｄｕ ＵＳＡ ＬＬＣ 自動運転車両のためのオブジェクト移動を予測するための方法およびシステム

Also Published As

Publication number Publication date
JPWO2023139706A1 (fr) 2023-07-27

Similar Documents

Publication Publication Date Title
Bruce et al. Deep parameter optimisation for face detection using the Viola-Jones algorithm in OpenCV
CN114450699A (zh) 由处理单元实现的方法、可读存储介质和处理单元
US11024022B2 (en) Data generation method and data generation device
JPWO2018047225A1 (ja) 学習装置、信号処理装置および学習方法
US11687824B2 (en) Intelligent data partitioning for distributed machine learning systems
Jeon et al. Target capacity filter pruning method for optimized inference time based on YOLOv5 in embedded systems
WO2023139706A1 (fr) Dispositif de traitement
CN116089895A (zh) 一种算子融合方法及装置
US10346716B2 (en) Fast joint template machining
Steinbring et al. GPU-accelerated progressive Gaussian filtering with applications to extended object tracking
CN117406725A (zh) 路径规划方法、装置、设备及存储介质
CN110705695B (zh) 搜索模型结构的方法、装置、设备和存储介质
CN110728359B (zh) 搜索模型结构的方法、装置、设备和存储介质
JP2007122207A (ja) プログラム分析プログラム、プログラム分析装置、プログラム分析方法
KR20190078692A (ko) 데이터 분포를 고려하여 데이터를 샘플링하는 장치 및 그 방법
US10191928B2 (en) Planar graph generation device and method
KR102168541B1 (ko) 제1 신경망을 이용한 제2 신경망 학습 방법 및 컴퓨터 프로그램
Jokić et al. Real-time mobile robot perception based on deep learning detection model
CN116679527A (zh) 版图拆分方法、版图拆分系统及存储介质
CN116468902A (zh) 图像的处理方法、装置和非易失性计算机可读存储介质
Wang et al. Speed-up Single Shot Detector on GPU with CUDA
US20220148298A1 (en) Neural network, computation method, and recording medium
KR20220068357A (ko) 딥러닝 객체 검출 처리 장치
US20240212390A1 (en) Electronic device for performing behavior recognition and operation method thereof
CN113805976B (zh) 数据处理方法及装置、电子设备及计算机可读存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22921865

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2023574958

Country of ref document: JP

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE