Information processing methods, programs, and information processing devices.

Abstract

The objective is to provide a technology that can detect a more accurate number of objects by correcting the inferred number of objects based on predetermined conditions. [Solution] The information processing device 3 includes a segmentation unit 312 that segments a plurality of image frames into predetermined number of frames, an inference unit 313 that infers the number of objects for each segmented image frame based on a predetermined algorithm, a determination unit 314 that determines whether or not to correct the inferred number of objects based on predetermined conditions, and an output unit 315 that corrects the number of objects based on the result of the determination and outputs the corrected number of objects.

Description

Technical Field

[0001] The present invention relates to an information processing method, a program, and an information processing apparatus.

Background Art

[0002] There is a disclosed technique for detecting pedestrians using video data captured by a photographing device mounted on a vehicle such as a drive recorder, and inferring the number of pedestrians.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] According to the above-described conventional technology, it is possible to infer pedestrians from image frames constituting video data using a person inference algorithm such as machine learning. However, there is a problem that there are many incorrect inferences, such as inferring the same pedestrian as a different pedestrian or inferring an object that is not a person, such as a utility pole, as a person.

[0005] Therefore, the disclosed technology has been made in view of such circumstances, and an object thereof is to provide a technology capable of detecting a more accurate number of objects from a plurality of image frames.

Means for Solving the Problems

[0006] An information processing device according to one aspect of the present invention comprises: a segmentation unit that segments a plurality of image frames into predetermined number of frames; an inference unit that infers the number of objects for each segmented image frame based on a predetermined algorithm; a determination unit that determines whether or not to correct the inferred number of objects based on predetermined conditions; and an output unit that corrects the number of objects based on the result of the determination and outputs the corrected number of objects. [Effects of the Invention]

[0007] According to the present invention, it is possible to detect a more accurate number of objects from multiple image frames. [Brief explanation of the drawing]

[0008] [Figure 1] This is a diagram illustrating the overview of a pedestrian detection system according to an embodiment of the present invention. [Figure 2] A diagram showing an example of the functional configuration of an information processing device according to an embodiment of the present invention. [Figure 3] A figure showing an example of inference processing according to an embodiment of the present invention. [Figure 4] A figure showing an example of a correction process according to an embodiment of the present invention. [Figure 5] A flowchart illustrating the processing procedure of an information processing apparatus according to an embodiment of the present invention. [Modes for carrying out the invention]

[0009] Embodiments of the present invention will be described in detail below with reference to the drawings. The same elements are denoted by the same reference numerals, and redundant descriptions are omitted.

[0010] <Examples of application> This invention enables the accurate detection of pedestrian numbers, for example, by using image frames acquired from video footage captured by a drive recorder mounted on a vehicle. By accurately detecting the number of pedestrians, it is possible to determine, for example, how many people are along a certain road, and to quantify the vibrancy and activity level along that road based on the number of people.

[0011] <System Overview> Figure 1 is a diagram illustrating the overview of the pedestrian detection system 1 according to this embodiment. In the example shown in Figure 1, the pedestrian detection system 1 is configured to include a drive recorder 2 mounted on a vehicle and an information processing device 3 that detects the number of pedestrians, etc.

[0012] The drive recorder 2 and the information processing device 3 can communicate with each other via the network N. While Figure 1 shows three drive recorders 2 and one information processing device 3, the number of drive recorders 2 and information processing devices 3 included in the pedestrian detection system 1 is not limited to these; any number of devices can be included. Furthermore, it is not mandatory for the drive recorder 2 and the information processing device 3 to communicate with each other to send and receive data. For example, the data captured by the drive recorder 2 could be stored in a storage device, and the data could be output from the storage device to the information processing device 3.

[0013] Network N is a communication network for communication between the drive recorder 2 and the information processing device 3. For example, Network N may be the Internet, LAN, dedicated line, packet communication network, part of the Public Switched Telephone Network (PSTN), mobile phone network, ISDNs (Integrated Service Digital Networks), wireless LANs, LTE (Long Term Evolution), CDMA (Code Division Multiple Access), other communication lines, or a combination thereof. Network N may be wired or wireless. Furthermore, the video data is not limited to drive recorders mounted on automobiles, but includes information captured by drive recorders mounted on motorcycles, autonomous vehicles, etc., and includes information captured by drive recorders mounted on any vehicle that can communicate directly with the information processing device 3 via Network N.

[0014] <Dashcam 2> Drive recorder 2 is, for example, a communication-type drive recorder and has at least the function of recording. Drive recorder 2 is, for example, mounted in a predetermined position on the front side of the passenger compartment of the vehicle and records the area in front of the vehicle through the windshield.

[0015] The drive recorder 2 includes a control unit including a CPU and memory, an imaging unit including a camera, a communication unit for connecting to the network N, an input unit such as a touch panel for receiving user operations, a storage unit for storing various programs necessary for executing processing in the control unit and recorded video data, and a display for displaying the screen. The drive recorder 2 transmits video data captured by the camera, etc., to the information processing device 3 via the network N. Alternatively, the drive recorder 2 may temporarily store the recorded video data in the storage unit and transmit it from the storage unit to the information processing device 3 at a predetermined timing.

[0016] <Information Processing Device 3> The information processing device 3 has functions for analyzing traffic information such as pedestrians and road conditions, and is composed of, for example, a dedicated or general-purpose server computer. The information processing device 3 provides functions such as detecting the number of pedestrians along a certain road based on video data provided by each drive recorder 2. Here, the computer that constitutes the information processing device 3 does not necessarily have to be a single unit, but may be composed of multiple computers distributed on a network N.

[0017] Figure 2 is a block diagram showing the functional configuration of the information processing device 3. The information processing device 3 comprises a control unit 310, a communication unit 320, and a storage unit 330. The control unit 310 of the information processing device 3 includes an image frame input unit 311, a segmentation unit 312, an inference unit 313, a determination unit 314, and an output unit 315.

[0018] The control unit 310 is composed of a microcomputer equipped with a CPU, ROM, RAM, I / O interface, etc., and centrally controls each part of the information processing device 3 by executing various programs stored in storage means such as ROM. Also, the control unit 310 has a function group and the like used when inferring and correcting the number of pedestrians, which will be described later.

[0019] The communication unit 320 has a communication interface compliant with various communication standards, and performs transmission and reception of various information with external devices including each drive recorder 2. The communication unit 320 receives video data uploaded from each drive recorder 2 via the network N. Also, the communication unit 320 can transmit the number of pedestrians detected in the information processing device 3 to the communication terminals (such as mobile phones and smartphones) of each user.

[0020] The storage unit 330 is constituted by, for example, a disk drive or a semiconductor memory, etc., and stores various programs and various data used during the execution of these programs. Specifically, the storage unit 330 stores video data acquired from each drive recorder 2. The video data acquired from each drive recorder 2 includes account information of the drive recorder using this system (such as drive recorder ID, ID of the vehicle equipped, driver ID, etc.) and content information of the drive recorder (such as driving area, driving date and time, driving situation information, etc.).

[0021] Also, the storage unit 330 stores computer programs and the like for realizing the above-described inference and correction. The fps (frames per second) of the video data stored in the storage unit 330 is 30fps, and it is stored in the storage unit 330 for each frame. Note that the setting of fps is arbitrary and the setting can also be changed. Hereinafter, a method for accurately detecting the number of pedestrians, etc. from the image frames constituting the video data using the information processing device 3 will be described in detail.

[0022] <Introduction to Inference> Conventional person detection algorithms often produce false positives, such as detecting the same pedestrian as different pedestrians or detecting non-person objects like utility poles as people. Therefore, in this embodiment, the number of pedestrians is first estimated using a predetermined algorithm, and then post-processing is applied to the estimated result. Based on predetermined conditions, the number of pedestrians that is likely to be an erroneous estimation is corrected, thereby enabling more accurate detection of the number of pedestrians.

[0023] Figure 3 is an explanatory diagram illustrating the preprocessing and estimation processes required for detecting the number of pedestrians.

[0024] <Pre-treatment> The image frame input unit 311 receives multiple image frames from the storage unit 330 to be used for detecting the number of pedestrians. These multiple image frames are acquired from video data captured by the drive recorder 2.

[0025] The segmentation unit 312 segments the multiple image frames acquired from the drive recorder 2 into predetermined number of frames, starting from the first frame. In the example shown in Figure 3, the segmentation unit 312 segments the multiple image frames acquired from the image frame input unit 311 into groups of 10 frames, assigns a unique segment ID to each segment, and assigns a unique frame ID to each frame that makes up the segment. Note that the number of frames to be segmented is not limited to 10 frames; it may be segmented into any number of frames.

[0026] <Inference Processing> The inference unit 313 infers the number of pedestrians for each segmented image frame 40 based on a predetermined algorithm. The inference algorithm can be arbitrarily selected, but for example, for each segmented image frame 40, the predetermined algorithm may be YOLO v7 (You Only Look Once) or StrongSORT (Simple Online and Realtime Tracking) to infer the number of pedestrians.

[0027] More specifically, the inference unit 313 infers a pedestrian unique to each frame that makes up the image frame 40, and assigns an ID (hereinafter also referred to as pedestrian ID) to each unique pedestrian inferred. Based on the assigned pedestrian ID, the inference unit 313 counts the number of pedestrians.

[0028] As a condition for assigning pedestrian IDs, the inference unit 313 assigns the same and unique pedestrian ID to image frames in which a unique pedestrian has been inferred two or more times consecutively. On the other hand, if a unique pedestrian has been inferred from only a single image frame, a different unique pedestrian ID is assigned to that single image frame.

[0029] In the example shown in Figure 3, the inference unit 313 will be considered correct if it can infer that there are 3 pedestrians, pedestrians 51-53, for the image frame 40.

[0030] <Pedestrian 51 Inference> The inference unit 313 infers pedestrians consecutively in frames ID 1 to 3, and therefore assigns pedestrian ID 71 to the corresponding image frames 40. Furthermore, the inference unit 313 infers pedestrians consecutively in frames ID 6 to 7, and therefore assigns another pedestrian ID 72 to the corresponding image frames 40.

[0031] The inference unit 313 counts the number of pedestrians based on the assigned pedestrian IDs. Since the pedestrian IDs assigned to pedestrian 51 are "71" and "72", the inference unit 313 counts the number of pedestrians as two.

[0032] In reality, only one pedestrian ID should be assigned to pedestrian 51, and the number of pedestrians should be counted as one. However, the inference unit 313 actually assigned two pedestrian IDs, inferring one person as two, resulting in an incorrect inference (inferring one extra person).

[0033] <Pedestrian 52 Inference> The inference unit 313 infers pedestrians consecutively in frames ID 2 to 4, and therefore assigns pedestrian ID 73 to the corresponding image frames 40.

[0034] The inference unit 313 counts the number of pedestrians based on the assigned pedestrian ID. Since the pedestrian ID assigned to pedestrian 52 is "73", the inference unit 313 counts the number of pedestrians as one person.

[0035] In reality, only one pedestrian ID should be assigned to pedestrian 52, and the number of pedestrians should be counted as one person. However, the inference unit 313 actually assigned only one pedestrian ID, and since it inferred one person as one, the inference was correct.

[0036] <Pedestrian 53 Inference> The inference unit 313 infers pedestrians consecutively in frames ID 4 to 8, and therefore assigns pedestrian ID 74 to the corresponding image frames 40. Furthermore, the inference unit 313 also infers a pedestrian in frame ID 10, and therefore assigns another pedestrian ID 75 to the corresponding image frame 40.

[0037] The inference unit 313 counts the number of pedestrians based on the assigned pedestrian IDs. Since the pedestrian IDs assigned to pedestrian 53 are "74" and "75", the inference unit 313 counts the number of pedestrians as two.

[0038] In reality, pedestrian 53 should be assigned one pedestrian ID, and the number of pedestrians should be counted as one. However, the inference unit 313 actually assigned two pedestrian IDs, inferring one person as two, resulting in an incorrect inference (inferring one extra person).

[0039] <Inference regarding utility pole 54> The inference unit 313 infers pedestrians consecutively in frames ID 3 to 4, and therefore assigns pedestrian ID 76 to the corresponding image frames 40.

[0040] The inference unit 313 counts the number of pedestrians based on the assigned pedestrian ID. Since the pedestrian ID assigned to the utility pole 54 is "76", it counts one pedestrian.

[0041] The correct approach is not to assign a pedestrian ID to utility pole 54, but rather to count the number of pedestrians as 0. However, the inference unit 313 actually assigned one pedestrian ID, and since it inferred 0 people as 1, an incorrect inference (inferring one more person than intended) occurred.

[0042] <Summary> Based on the above, by using the predetermined algorithm, the inference unit 313 assigned pedestrian IDs 71 to 76 to pedestrians 51 to 53. Therefore, when counting the number of pedestrians based on the pedestrian IDs, it inferred that there were 2 people for pedestrian 51, 1 person for pedestrian 52, 2 people for pedestrian 53, and 1 person for utility pole 54, for a total of 6 people.

[0043] The inference unit 313 assigns only one pedestrian ID to each of the pedestrians 51-53, and if it can infer a total of 3 people based on the pedestrian IDs, it will be considered correct. However, the inference unit 313 infers a total of 6 people, so while it can roughly count the number of pedestrians, it is difficult to count them accurately.

[0044] Here, for example, since pedestrians in the dashcam footage are observed continuously across multiple frames, the determination unit 314 can determine the need for correction to the pedestrian ID inferred by the inference unit 313 based on predetermined conditions using density parameters described later, and improve the accuracy of the inference result by correcting it if necessary.

[0045] <Correction process> In this embodiment, in order to correct the number of pedestrians to a more accurate value, post-processing is applied to the result inferred by the inference unit 313. The determination unit 314 determines whether or not to correct the pedestrian ID assigned by the inference unit 313 based on predetermined conditions.

[0046] In this embodiment, a density parameter is used as a predetermined condition. The "density parameter" represents the ratio of the number of objects to the number of image frames that make up a segment. For example, if a segment consists of 10 frames, and the same pedestrian ID is detected in three frames, the density parameter will be "0.3". In this embodiment, if the density parameter is less than "0.3", the inference is judged to be incorrect, while if it is "0.3" or greater, the inference is judged to be correct. Note that the threshold value for the density parameter (here, "0.3") can be set arbitrarily.

[0047] For example, if the number of identical pedestrian IDs assigned by the inference unit 313 within 10 frames is less than 3, the determination unit 314 determines that the density parameter is less than "0.3" and falls below the threshold for the density parameter. Therefore, it corrects the inference of pedestrians with that pedestrian ID, and does not count that pedestrian ID as a pedestrian.

[0048] <Decision: Introduction> Figure 4 is an explanatory diagram illustrating the correction process necessary to improve the accuracy of pedestrian count detection. Based on pedestrian IDs 71-76, the inference unit 313 inferred that the total number of pedestrians was 6 for the segmented image frame 40. The decision unit 314 then determines whether or not correction is necessary for the number of pedestrians inferred by the inference unit 313, and performs the correction as needed.

[0049] <Pedestrian 51's judgment> Pedestrian ID 71 is inferred a total of three times (frames ID 1 to ID 3) within 10 frames by the inference unit 313. In this case, the density parameter is (0.3), which is above the set threshold for the density parameter (0.3), so the judgment unit 314 determines that it is not subject to correction. On the other hand, pedestrian ID 72 is inferred a total of two times (frames ID 6 to ID 7) within 10 frames by the inference unit 313. In this case, the density parameter is (0.2), which is below the set threshold for the density parameter (0.3), so the judgment unit 314 determines that it is subject to correction, and performs a correction to cancel pedestrian ID 72, deeming it to be an error.

[0050] The determination unit 314 counts the number of pedestrians based on the assigned pedestrian ID. Since the corrected pedestrian ID for pedestrian 51 is only "71", the determination unit 314 counts the number of pedestrians as one person.

[0051] <Pedestrian 52 Judgment> Pedestrian ID 73 is inferred a total of three times (frames ID 2-4) within 10 frames by the inference unit 313. In this case, the density parameter is (0.3), and since this is above the set threshold for the density parameter (0.3), the judgment unit 314 determines that it is not subject to correction.

[0052] The determination unit 314 counts the number of pedestrians based on the assigned pedestrian ID. Since the corrected pedestrian ID for pedestrian 52 is only "73", the determination unit 314 counts the number of pedestrians as one person.

[0053] <Pedestrian 53 Judgment> Pedestrian ID 74 is inferred a total of 5 times (frames ID 4-8) within 10 frames by the inference unit 313. In this case, the density parameter is (0.5), which is above the set threshold for the density parameter (0.3), so the judgment unit 314 determines that it is not subject to correction. On the other hand, pedestrian ID 75 is inferred a total of 1 time (frame ID 10) within 10 frames by the inference unit 313. In this case, the density parameter is (0.1), which is below the set threshold for the density parameter (0.3), so the judgment unit 314 determines that it is subject to correction, and performs a correction to cancel pedestrian ID 75, deeming it to be an error.

[0054] The determination unit 314 counts the number of pedestrians based on the assigned pedestrian ID. Since the corrected pedestrian ID for pedestrian 51 is only "74", the determination unit 314 counts the number of pedestrians as one person.

[0055] <Utility pole 54 Judgment> Pedestrian ID 76 is inferred a total of two times (frames ID 3-4) within 10 frames by the inference unit 313. In this case, the density parameter is (0.2), which is below the set threshold for the density parameter (0.3). Therefore, the judgment unit 314 determines that it is subject to correction and, judging that pedestrian ID 76 is incorrect, performs a correction to cancel pedestrian ID 76.

[0056] The determination unit 314 counts the number of pedestrians based on the assigned pedestrian ID. Since no corrected pedestrian ID has been assigned to the utility pole 54, the determination unit 314 counts the number of pedestrians as 0.

[0057] <Summary> The correct approach would be to assign only one pedestrian ID to each of pedestrians 51-53, and then count them as a total of three based on their pedestrian IDs. However, the corrected pedestrian IDs assigned by the judgment unit 314 are one each for pedestrians 51-53, and the unit correctly identifies three people as three, thus correctly detecting the number of pedestrians. In this way, the judgment unit 314 can improve the accuracy of the inference results of the inference unit 313 regarding the number of pedestrians.

[0058] <Output section> The output unit 315 outputs the result of post-processing. That is, the output unit 315 outputs the corrected number of pedestrians calculated by the judgment unit 314. Here, the output unit 315 may store the corrected number of pedestrians in the storage unit 330.

[0059] <Example of operation> Next, the operation of the information processing device 3 according to this embodiment will be described. Figure 5 is a flowchart showing an example of the processing of the information processing device 3 according to this embodiment.

[0060] First, in step S101, the segmentation unit 312 segments the multiple image frames acquired from the image frame input unit 311 into groups of 10 frames each, starting from the first frame.

[0061] Next, in step S102, the inference unit 313 infers the number of pedestrians for each segment based on a predetermined algorithm. Specifically, the inference unit 313 infers a unique pedestrian for each frame, assigns a pedestrian ID to each unique pedestrian inferred, and counts the number of pedestrians based on the assigned pedestrian ID.

[0062] Next, in step S103, the determination unit 314 determines whether or not to correct the number of pedestrians inferred by the inference unit 313 based on predetermined conditions. Here, density parameters are used as predetermined conditions.

[0063] In step S103, if the determination unit 314 determines that the pedestrian ID assigned by the inference unit 313 is subject to correction, it proceeds to step S104 (step S103: YES). If it determines that the pedestrian ID is not subject to correction, it proceeds to step S105 (step S103: NO).

[0064] Next, in step S104, the determination unit 314 corrects the pedestrian ID assigned by the inference unit 313 if the density parameter is less than the threshold for the density parameter. This is because if the density parameter is less than the threshold, it is considered noise and not a pedestrian. Based on the corrected pedestrian ID, the determination unit 314 calculates the total number of pedestrians.

[0065] Next, in step S105, the output unit 315 outputs the total corrected pedestrian IDs. The output unit 315 may also transmit information such as the total corrected pedestrian IDs to any information processing device.

[0066] As described above, in the present invention, multiple image frames are segmented into predetermined numbers of frames, and the number of pedestrians is inferred for each segmented image frame based on a predetermined algorithm. Based on predetermined conditions, it is determined whether or not to correct the inferred number of pedestrians, and based on the result of the determination, the number of pedestrians is corrected and the corrected number of pedestrians is output.

[0067] This invention enables the detection of a more accurate number of pedestrians by correcting the estimated number of pedestrians based on predetermined conditions, thereby canceling out pedestrians for whom the estimate is likely to be incorrect.

[0068] <Variation> The above embodiments or examples are provided to facilitate understanding of the present invention and are not intended to limit it. The present invention can be modified or improved without departing from its spirit, and equivalents thereof are also included. Furthermore, the present invention can form various disclosures by appropriately combining the multiple components disclosed in the above embodiments or examples. For example, some components may be deleted from all the components shown in the embodiments. Furthermore, components may be appropriately combined in different embodiments. For example, the determination unit 314 may determine whether or not to correct the number of pedestrians inferred by appropriately combining the continuity parameter and density parameter described later for a given condition.

[0069] <Example 1> (Continuity) The specified condition may be, for example, a continuity parameter that represents the continuity of the number of objects inferred from the number of image frames constituting the segment. The "continuity parameter" represents the ratio of consecutive objects to the number of image frames constituting the segment. For example, if the number of frames constituting a segment is 10, and the same pedestrian ID is inferred for "three consecutive frames", the continuity parameter would be "0.3". In Modification 1, if the continuity parameter is less than "0.3", it is determined that correction is necessary (i.e., the inference is incorrect), while if the continuity parameter is "0.3" or greater, it is determined that no correction is necessary (i.e., the inference is correct). Note that the threshold for the continuity parameter (in this case, "0.3") can be set arbitrarily.

[0070] For example, if the number of consecutive identical pedestrian IDs assigned by the inference unit 313 within 10 frames is less than 3, the continuity parameter is less than "0.3," which is below the threshold for the continuity parameter. Therefore, the inference of a pedestrian with that pedestrian ID is incorrect, and the judgment unit 314 corrects this by not counting that pedestrian ID as a pedestrian.

[0071] When the continuity parameter is used as a predetermined condition, the judgment unit 314 adopts the inference result only if the degree of continuity of inferences with the same ID exceeds a threshold, thereby improving the accuracy of the correction. As a result, the number of pedestrians can be detected more accurately.

[0072] <Modification 2> (likelihood) The specified conditions may, for example, be a likelihood parameter that represents the ratio of the sum of the likelihoods associated with the number of objects to the number of objects being inferred.

[0073] The inference unit 313 infers a pedestrian unique to each frame that makes up the image frame 40, and assigns a unique pedestrian ID as well as a score representing the likelihood that the person is a pedestrian (hereinafter, this score representing likelihood is also called the likelihood). The likelihood takes a value between 0 and 1, and approaches 1 as the likelihood of the person being a pedestrian increases. The inference unit 313 counts the number of pedestrians based on the assigned pedestrian ID.

[0074] In Modification 2, a likelihood parameter is used as a predetermined condition. The "likelihood parameter" represents the ratio of the sum of the likelihoods associated with the number of objects to the number of objects assigned the same pedestrian ID. For example, if a segment consists of 10 frames, and the same pedestrian ID is assigned to 4 frames, and the sum of the likelihoods associated with each pedestrian ID is 3 (=0.8+0.9+0.7+0.6), then the likelihood parameter is "0.75". In Modification 2, if the likelihood parameter is less than "0.75", it is determined that a correction is necessary (i.e., the inference is incorrect), while if the likelihood parameter is "0.75" or greater, it is determined that no correction is necessary (i.e., the inference is correct). Note that the threshold for the likelihood parameter (in this case, "0.75") can be set arbitrarily.

[0075] For example, if the ratio of the total likelihood associated with the same pedestrian ID to the number of identical pedestrian IDs assigned by the inference unit 313 within 10 frames is less than "0.75", the decision unit 314 determines that the likelihood parameter is less than "0.75" and falls below the threshold for the likelihood parameter, and therefore the inference of the pedestrian with that pedestrian ID is incorrect, and corrects the calculation so that the pedestrian ID is not counted as a pedestrian.

[0076] When the likelihood parameter is used as a predetermined condition, the accuracy of the correction increases as the accuracy of the likelihood assigned by the predetermined algorithm increases. As a result, the number of pedestrians can be detected more accurately.

[0077] <Variation 3> (Skip process) The segmentation unit 312 may thin out a plurality of image frames based on predetermined conditions and segment the resulting plurality of image frames.

[0078] For example, a predetermined condition is set so that if the image frame ID is even, it will be subject to decimation. The segmentation unit 312 decimates the input multiple image frame IDs if they are even, and segments the frames into groups of 10 frames, using only those with odd image frame IDs.

[0079] The segmentation unit 312 can adjust the amount of change in image frames by downsampling image frames based on predetermined conditions and segmenting the resulting multiple image frames. As a result, it can generate segments that are optimal for the captured video data.

[0080] Furthermore, the segmentation unit 312 can reduce the amount of data in the acquired image frames by downsampling image frames based on predetermined conditions and segmenting the resulting multiple image frames. As a result, multiple segments can be generated efficiently. [Explanation of Symbols]

[0081] 1...Pedestrian detection system, 2...Drive recorder, 3...Information processing device, 21...Input unit, 23...Display, 24...Control unit, 25...Communication unit, 26...Storage unit, 51-53...Pedestrian, 54...Utility pole, 311...Image frame input unit, 312...Segmentation unit, 313...Inference unit, 314...Decision unit, 315...Output unit, 320...Communication unit, 330...Storage unit

Claims

1. A segmentation unit that segments multiple image frames into predetermined numbers of frames, For each of the segmented image frames, an inference unit infers the number of objects based on a predetermined algorithm, A determination unit that determines whether or not to correct the inferred number of objects based on predetermined conditions, Based on the results of the above determination, an output unit corrects the number of objects and outputs the corrected number of objects, An information processing device equipped with the following features.

2. The predetermined condition is a density parameter that represents the ratio of the number of objects inferred from the number of image frames that constitute the segment. The information processing apparatus according to claim 1.

3. The predetermined condition is a continuity parameter that represents the continuity of the number of objects inferred from the number of image frames constituting the segment. The information processing apparatus according to claim 1.

4. The predetermined condition is a likelihood parameter that represents the ratio of the sum of the likelihoods associated with the number of objects to the number of objects inferred from the number of image frames constituting the segment. The information processing apparatus according to claim 1.

5. The segmentation unit thins out the plurality of image frames based on predetermined conditions and segments the plurality of image frames after thinning. The information processing apparatus according to claim 1.

6. The number of objects mentioned above is the number of pedestrians. The information processing apparatus according to claim 1.

7. The steps include: segmenting multiple image frames into predetermined frame numbers, For each of the segmented image frames, the number of objects is inferred based on a predetermined algorithm. A step of determining whether or not to correct the inferred number of objects based on predetermined conditions, The steps include correcting the number of objects based on the result of the above determination and outputting the corrected number of objects, Information processing methods including

8. On the computer, The steps include: segmenting multiple image frames into predetermined frame numbers, For each of the segmented image frames, the number of objects is inferred based on a predetermined algorithm. A step of determining whether or not to correct the inferred number of objects based on predetermined conditions, The steps include correcting the number of objects based on the result of the above determination and outputting the corrected number of objects, A program to execute.

Images