Entry inference system

Abstract

An entry inference system comprising: a detection unit that detects a user in a prescribed first area and outputs a detection image that is a detection result; a control unit that infers whether or not the user will enter a prescribed second area from the first area on the basis of the detection image; and a temperature sensor that measures the temperature of the control unit or of the vicinity of the control unit, wherein the control unit includes an acquisition unit that acquires the detection image from the detection unit and acquires the temperature from the temperature sensor, an inference mode determination unit that determines an inference mode on the basis of the temperature, an inference unit that uses the inference mode determined by the inference mode determination unit to infer whether or not the user will enter the second area from the first area or to infer the probability of entry on the basis of the detection image, and an output unit that outputs the inference results inferred by the inference unit.

Description

Entry Estimation System

[0001] One aspect of the present disclosure relates to an entry estimation system. This application claims priority based on Japanese Application No. 2024-205947 filed on November 27, 2024, and incorporates all the contents described in the Japanese application.

[0002] Patent Document 1 discloses a technique for identifying a pedestrian waiting to cross a road and controlling a traffic signal related to the crosswalk by analyzing the line of sight of a person in a waiting area waiting for a signal at the crosswalk.

[0003] Japanese Unexamined Patent Application Publication No. 2017-208141

[0004] The entry estimation system according to one embodiment includes a detection unit that detects a user in a predetermined first area and outputs a detection image as a detection result, a control unit that estimates whether the user enters a predetermined second area from the first area based on the detection image, a temperature sensor that measures the temperature around the control unit or the control unit itself. The control unit has an acquisition unit that acquires the detection image from the detection unit and acquires the temperature from the temperature sensor, an estimation mode determination unit that determines an estimation mode based on the temperature, an estimation unit that estimates whether the user enters a predetermined second area from the first area or the probability of entry in the estimation mode determined by the estimation mode determination unit based on the detection image, and an output unit that outputs the estimation result by the estimation unit.

[0005] FIG. 1 is a diagram for explaining the outline of the entry estimation system according to this embodiment. FIG. 2 is a block diagram illustrating the hardware configuration of the control unit. FIG. 3 is a functional block diagram of the control unit. FIG. 4 is a diagram for explaining the change of the estimation range according to the temperature. FIG. 5 is a diagram for explaining the change of the estimation range according to the temperature. FIG. 6 is a diagram for explaining and comparing the normal mode and the low load mode. FIG. 7 is a flowchart for explaining the change process of the estimation range according to the temperature. FIG. 8 is a diagram for explaining the change of the estimation process according to the temperature. FIG. 9 is a diagram for explaining and comparing the normal mode and the low load mode. FIG. 10 is a flowchart for explaining the change process of the estimation process according to the temperature.

[0006] To accurately estimate pedestrians waiting to cross the street, it is necessary to perform complex processing, such as deep learning, using a large amount of information obtained from video footage, including pedestrian trajectories, face orientation, gaze, behavior, interactions between pedestrians, and the walking environment. This requires a high-speed computing device in the control unit, which leads to problems with high power consumption and heat generation. To address such heat issues, forced air cooling, such as fans, is usually used. However, in configurations where the control unit is located outdoors, for example, it is difficult to install fans due to product lifespan and weight considerations, making it difficult to adequately suppress heat generation.

[0007] This disclosure is made in view of the above circumstances and aims to appropriately suppress heat generation even in a configuration in which the control unit is located outdoors.

[0008] According to one aspect of this disclosure, in an entry estimation system, heat generation can be appropriately suppressed even in a configuration in which the control unit is located outdoors.

[0009] [Description of Embodiments of the Disclosure] First, the contents of the embodiments of the disclosure will be listed and described.

[0010] [1] An entry estimation system according to one embodiment includes: a detection unit that detects a user in a predetermined first area and outputs a detection image as a result of the detection; a control unit that estimates whether or not the user enters a predetermined second area from the first area based on the detection image; and a temperature sensor that measures the temperature of the area surrounding the control unit or the control unit itself. The control unit includes: an acquisition unit that acquires a detection image from the detection unit and the temperature from the temperature sensor; an estimation mode determination unit that determines an estimation mode based on the temperature; an estimation unit that estimates whether or not the user enters the second area from the first area or the probability of entry based on the detection image using the estimation mode determination unit; and an output unit that outputs the estimation result from the estimation unit.

[0011] In this type of intrusion estimation system, a temperature sensor measures the temperature around the control unit or the control unit itself. Based on this temperature, an estimation mode for intrusion is determined, and intrusion estimation is performed based on the user's detected image using this estimation mode. By determining the estimation mode according to the temperature around the control unit or the control unit itself, it becomes possible, for example, to perform intrusion estimation in a way that prevents the control unit from overheating if the temperature around the control unit is rising. This means that even if the control unit is installed outdoors and forced air cooling such as a fan cannot be used, the estimation mode can be changed according to the temperature around the control unit to appropriately suppress the heat generated by the control unit. This extends the product life of the control unit. Furthermore, even if a cooling fan is provided, the configuration is simple, allowing for a lighter weight for the equipment.

[0012] [2] In the entry estimation system described in [1] above, the estimation mode determination unit may determine the estimation mode such that, when the temperature is above a predetermined first threshold, the processing load on the estimation unit is reduced compared to normal conditions when the temperature is not above the first threshold. With such a configuration, the processing load on the estimation unit can be reduced when the temperature is above a predetermined first threshold, so that heat generation in the control unit can be appropriately suppressed even when forced air cooling such as a fan cannot be used.

[0013] [3] In the entry estimation system described in [2] above, the estimation mode determination unit may determine the estimation mode such that, at high temperatures, the estimation range, which is the range of the detected image used by the estimation unit for estimation, becomes narrower compared to the normal state. By narrowing the estimation range at high temperatures, the processing load on the estimation unit at high temperatures can be reliably reduced, and the heat generation of the control unit can be appropriately suppressed.

[0014] [4] In the entry estimation system described in [3] above, the estimation range under normal conditions and the estimation range under high temperatures may be set in advance so that they correspond to predetermined ranges in the detected image. By setting the estimation ranges under normal conditions and high temperatures in advance in this way, it is possible to reliably reduce the processing load under high temperatures and suppress variations in the estimation accuracy by the estimation unit.

[0015] [5] In the entry estimation system described in [4] above, the estimation range at high temperatures may be set in advance to include at least the area in the first area that is continuous with the second area. With such a configuration, even at high temperatures, the estimation range can include the area that is expected to be passed through when entering the second area from the first area (an area in which a user who intends to enter the second area can be detected), and the estimation accuracy of the estimation unit can be ensured.

[0016] [6] In the entry estimation system described in any one of [3] to [5] above, the estimation mode determination unit may determine the estimation mode such that the estimation range is wider compared to the high-temperature state when the temperature falls below a predetermined second threshold while the estimation mode is set to the high-temperature state. In this way, when the temperature around the control unit or the control unit itself becomes sufficiently low and heat generation from the control unit is no longer a problem, the estimation accuracy of the estimation unit can be further improved by widening the estimation range.

[0017] [7] In the entry estimation system described in any one of [1] to [6] above, the estimation unit may estimate whether one or more users will enter the second area or the probability of them entering by inputting multiple detection images in a time series into an AI model that outputs whether one or more users in the detection images intend to enter the second area or the probability of them entering. In this way, by using an AI model, entry estimation can be performed easily and with high accuracy.

[0018] [8] In the entry estimation system described in any one of [2] to [6] above, the estimation mode determination unit may determine the estimation mode such that, under normal circumstances, it is in a normal mode that estimates whether or not one or more users in the detected image intend to enter the second area, or the probability that they do, and under high temperature conditions, it is in a low-load mode that estimates whether or not there are users in the detected image who intend to enter the second area, or the probability that they do. In this way, under normal circumstances, the estimation is made more accurate by estimating the intention of each user to enter, and under high temperature conditions, the processing load is reduced by simply estimating whether or not there are users in the detected image who intend to enter, or the probability that they do, thereby making it possible to perform estimation that is less likely to cause overheating of the control unit.

[0019] [9] In the entry estimation system described in [8] above, the estimation mode determination unit may determine the estimation mode to be in normal mode when the temperature is below a predetermined second threshold while in low-load mode. In this way, when the temperature becomes sufficiently low and heat generation in the control unit is no longer a problem, the estimation accuracy of the estimation unit can be further improved by switching to normal mode, which allows for detailed estimation.

[0020]

[10] In the entry estimation system described in [8] or [9] above, the estimation unit may, in normal mode, input multiple detection images in a time series to an AI model that outputs whether or not one or more users in the detection images intend to enter the second area or the probability of them entering, and in low-load mode, input multiple detection images in a time series to an AI model that outputs whether or not there are users in the detection images who intend to enter the second area or the probability of them entering, thereby estimating whether or not a user will enter the second area from the first area or the probability of them entering.

[0021] [Details of Embodiments of the Disclosure] Specific examples of embodiments of the disclosure are described below with reference to the drawings. The present invention is not limited to these examples, but is indicated by the claims, and all modifications within the meaning and scope of the claims are intended to be included. In the description of the drawings, the same elements are denoted by the same reference numerals, and redundant descriptions are omitted.

[0022] Figure 1 is a diagram illustrating the overview of the entry estimation system 1 according to this embodiment. First, the overview of the entry estimation system 1 will be explained. Note that the temperature sensor 30 (see Figure 3), which will be described later, is not shown in Figure 1.

[0023] The entry estimation system 1 is a system that estimates whether a user will enter a predetermined second area from a first area based on a detection image, which is the result of a user detection in a predetermined first area by a camera 20 (detection unit). In this embodiment, an area including a pedestrian crossing E1 is given as an example of the second area. Also, a sidewalk E2 adjacent to the pedestrian crossing E1 is given as an example of the first area. That is, in this embodiment, the first area and the second area are adjacent to each other. Sidewalk E2 includes an area that a user Y entering the pedestrian crossing E1 passes through immediately before entering, and includes an area where a user Y who plans to enter the pedestrian crossing E1 waits when a vehicle is passing over the pedestrian crossing E1. For example, user Y waits at a traffic light on sidewalk E2, then enters the pedestrian crossing E1, crosses the pedestrian crossing E1, and reaches the sidewalk on the opposite side. Note that examples of user Y may include not only people walking, but also people moving on vehicles (bicycles, wheelchairs, kick scooters, strollers, etc.). Furthermore, examples of user Y may include people walking while using their smartphones (smartphone zombies), people using white canes, etc. User Y may also include moving objects other than people (animals, robots, etc.).

[0024] As shown in Figure 1, users Y detected on sidewalk E2 can be divided into users Y1 who have the intention to cross (enter) the pedestrian crossing E1 and users Y2 who do not have the intention to cross (enter). The entry estimation system 1 may, for example, estimate whether user Y will enter the pedestrian crossing E1 by estimating the intention to cross of the detected user Y.

[0025] As shown in Figure 1, the entry estimation system 1 includes a camera device 2, which includes a camera 20 (detection unit) and a control unit 10. The camera device 2 is installed on the sidewalk E2. More specifically, the camera device 2 is installed on a pole 50 installed on the sidewalk E2. The pole 50 is a traffic signal pole on which a traffic signal 60 is installed. The pole 50 may also be a signpost or a lighting pole, etc. The camera device 2 may be installed near the top of the pole 50.

[0026] Camera 20 is a detection unit that detects a user Y passing through the sidewalk E2 and outputs a detection image as the detection result. Camera 20 is installed directly above the sidewalk E2, and by setting the detection range downward from the perspective of camera 20, it detects user Y on the sidewalk E2. Directly above the sidewalk E2 means vertically above a part of the sidewalk E2. Note that camera 20 does not necessarily have to be installed directly above the sidewalk E2; it may be installed at an oblique angle to the sidewalk E2 as long as it can detect the entire area of ​​the sidewalk E2. The sidewalk E2 may be set up as a rectangular shape in plan view of several meters x several meters, for example, as an area of ​​4 meters x 4 meters.

[0027] Camera 20 only needs to have its horizontal and vertical angles set so that it can detect the entire sidewalk E2, and the horizontal and vertical angles may be set to, for example, ±20°. Note that sidewalk E2 may be the entire detection range of camera 20 or only a part of it.

[0028] Camera 20 periodically captures images of the sidewalk E2 at predetermined time intervals and continuously transmits the captured images, which are the imaging results (detection results), to the control unit 10. Camera 20 transmits the imaging results (detection results) to the control unit 10 via a base station (not shown), or via direct wireless communication or wired communication without going through a base station.

[0029] In this embodiment, a camera 20 is used as an example of the detection unit, but the detection unit may be configured in a way other than a camera 20, as long as it is capable of detecting a user. Specifically, the detection unit may be a radar (radio wave sensor) such as a millimeter-wave sensor or a detection mechanism such as a lidar.

[0030] The control unit 10 is composed of one or more control computers. Figure 2 is a block diagram illustrating the hardware configuration of the control unit 10. For example, the control unit 10 has the circuit 120 shown in Figure 2. The circuit 120 has one or more processors 121, a memory 122, a storage 123, and an input / output port 124. The storage 123 has a storage medium that can be read by a computer, such as a hard disk. The storage medium stores a program for executing a predetermined intrusion estimation processing procedure. The storage medium may be a removable medium such as a non-volatile semiconductor memory, a magnetic disk, or an optical disk. The memory 122 temporarily stores the program loaded from the storage medium of the storage 123 and the calculation results by the processor 121. The processor 121 works in cooperation with the memory 122 to execute the above program and constitutes each of the functional modules described later. The input / output port 124 inputs and outputs electrical signals according to commands from the processor 121.

[0031] Furthermore, the hardware configuration of the control unit 10 is not necessarily limited to a configuration in which each functional module is composed of a program. For example, each functional module of the control unit 10 may be composed of a dedicated logic circuit or an ASIC (Application Specific Integrated Circuit) that integrates such circuits.

[0032] The control unit 10 estimates whether user Y enters the crosswalk E1 from the sidewalk E2 based on the detection image output from the camera 20. Figure 3 is a functional block diagram of the control unit 10. As shown in Figure 3, the control unit 10 has an acquisition unit 11, an estimation mode determination unit 12, an estimation unit 13, and an output unit 14. As shown in Figure 3, the entry estimation system 1 further includes a temperature sensor 30. The temperature sensor 30 is a temperature sensor that measures the temperature around the control unit 10 and may be located inside the camera device 2 or outside the camera device 2. The temperature around the control unit 10 means the temperature of the region where the temperature changes due to the effect of heat generation according to the processing load of the control unit 10. In this embodiment, the temperature sensor 30 is described as measuring the temperature around the control unit 10, but the temperature sensor 30 may measure the temperature of the control unit 10 itself (for example, the junction temperature of the IC). The temperature sensor 30 periodically measures the temperature at predetermined time intervals and transmits the measured temperature to the control unit 10.

[0033] The acquisition unit 11 acquires a detected image from the camera 20 and the temperature around the control unit 10 from the temperature sensor 30. The acquisition unit 11 may acquire information each time information is transmitted from the camera 20 and the temperature sensor 30, or it may acquire information in batches at predetermined time intervals. The information acquired by the acquisition unit 11 (detected image and temperature) is associated with at least a time (detection time or measurement time).

[0034] The estimation mode determination unit 12 determines the estimation mode of the estimation unit 13 based on the ambient temperature around the control unit 10, which is acquired by the acquisition unit 11 (measured by the temperature sensor 30). The estimation mode is information that indicates the conditions under which the estimation unit 13 will perform the estimation process.

[0035] The estimation mode determination unit 12 may determine the estimation mode so as to reduce the processing load on the estimation unit 13 when the temperature around the control unit 10 exceeds a predetermined first threshold, i.e., at high temperatures, and when the temperature around the control unit 10 does not exceed the first threshold, i.e., under normal conditions. The first threshold here may be, for example, 80°C to 90°C.

[0036] The estimation mode determination unit 12 may determine the estimation mode at high temperatures such that the estimation range, which is the range of the detected image used for estimation by the estimation unit 13, becomes narrower compared to normal conditions. Figure 4 is a diagram illustrating the change in the estimation range according to temperature. At normal times (when the temperature around the control unit 10 does not exceed the first threshold), the estimation mode determination unit 12 determines that it is in normal mode and sets the estimation range to, for example, the normal range DE1 (for example, the maximum range that the camera 20 can detect on the sidewalk E2). Here, the normal range DE1 only needs to be larger than the reduced range DE2 described later. At high temperatures (when the temperature around the control unit 10 exceeds the first threshold), the estimation mode determination unit 12 determines that it is in low-load mode and sets the estimation range to, for example, the reduced range DE2, which is narrower than the normal range DE1.

[0037] The normal range DE1 (estimated range under normal conditions) and the reduced range DE2 (estimated range at high temperatures) may be pre-set to correspond to predetermined ranges in the detected image. For example, the reduced range DE2 may be pre-set to include at least an area continuous with the pedestrian crossing E1 on the sidewalk E2, as shown in Figure 4.

[0038] As shown in Figure 5, the normal range DE1 may be set to have a resolution of, for example, 1280 x 720 px. The reduced range DE2 may be set to have a resolution of, for example, 320 x 180 px.

[0039] The estimation unit 13 estimates, based on the detected image, whether user Y enters the crosswalk E1 from the sidewalk E2, using the estimation mode determined by the estimation mode determination unit 12. The estimation unit 13 may, for example, estimate whether one or more users Y enter the crosswalk E1 by inputting multiple detected images in a time series into an AI model that outputs whether one or more users Y in the detected images intend to enter the crosswalk E1. Such an AI model uses a large amount of information, such as the pedestrian's (user Y's) trajectory, face direction, gaze, behavior, interaction between pedestrians, and walking environment, to output whether user Y intends to enter the crosswalk E1, and is provided by well-known technology. As described above, here it is assumed that the estimation mode determination unit 12 determines either the normal mode, where the estimation process is performed with the normal range DE1 as the estimation range, or the low-load mode, where the estimation process is performed with the reduced range DE2 as the estimation range. In this embodiment, the estimation unit 13 is described as estimating whether or not user Y enters the pedestrian crossing E1 from the sidewalk E2 using the estimation mode determined by the estimation mode determination unit 12. However, the estimation unit 13 may also estimate the probability (likelihood) that user Y enters the pedestrian crossing E1 from the sidewalk E2 using the estimation mode determined by the estimation mode determination unit 12.

[0040] Figure 6 is a diagram illustrating the comparison between the normal mode and the low-load mode. As shown in Figure 6, in the normal mode, the estimation unit 13 inputs multiple images (past N frames (N is a natural number)) in a time series for the detection range DE1 (for example, with a resolution of 1280 × 720 px) into the AI ​​model and outputs whether or not each user Y in the detection image intends to enter the crosswalk E1. Also, as shown in Figure 6, in the low-load mode, the estimation unit 13 inputs multiple images (past N frames (N is a natural number)) in a time series for the detection range DE2 (for example, with a resolution of 320 × 180 px) into the AI ​​model and outputs whether or not each user Y in the detection image intends to enter the crosswalk E1. Thus, in the example shown in Figure 6, the processing and AI model are common to both the normal mode and the low-load mode, except for the estimation range of the detection image.

[0041] The output unit 14 is an output unit that outputs the estimation result from the estimation unit 13, and may control an external device 90 based on the estimation result. In this case, the external device 90 may be, for example, an on-board device of a vehicle traveling near the pedestrian crossing E1. Alternatively, the external device 90 may be a device that controls signals (a device that changes the color of the signal or controls the extension or shortening of the signal time).

[0042] Figure 7 is a flowchart illustrating the process of changing the estimation range according to temperature. As shown in Figure 7, first, the estimation mode determination unit 12 determines whether or not the temperature around the control unit 10 exceeds a first threshold (step S1).

[0043] In step S1, if it is determined that the first threshold is not exceeded, the input image is treated as a high-resolution image (an image whose estimation range is the normal range DE1) (step S2), and based on the input image, the estimation unit 13 performs AI processing in normal mode to estimate whether or not user Y enters the pedestrian crossing E1 (step S3). The estimation result is then output by the output unit 14, and the process of step S1 is executed again.

[0044] In step S1, if it is determined that the input image exceeds the first threshold, the input image is regarded as a low-resolution image (an image whose estimated range is the reduced range DE2) (step S5), and AI processing in the low-load mode by the estimation unit 13 is executed based on the input image, and it is estimated whether the user Y enters the crosswalk E1 (step S6). Then, the estimation result is output by the output unit 14 (step S7).

[0045] Then, the estimation mode determination unit 12 determines whether the temperature around the control unit 10 is lower than the second threshold (step S8). In this way, the estimation mode determination unit 12 determines whether the temperature around the control unit 10 is lower than the second threshold in the state where the estimation mode at high temperature is adopted. The second threshold here may be, for example, 50°C to 60°C. If it is determined in step S8 that the temperature is lower than the second threshold, the estimation mode determination unit 12 determines the estimation mode so that the estimated range becomes wider compared with the high-temperature case, specifically, so that the estimated range becomes the normal range DE1, and executes the processing after step S2.

[0046] Note that the processing of the estimation mode determination unit 12 and the estimation unit 13 is not limited to the above-described mode. Hereinafter, referring to FIGS. 8 to 10, another mode of the processing of the estimation mode determination unit 12 and the estimation unit 13 will be described.

[0047] Figure 8 illustrates the change in estimation processing according to temperature. In the example described above, the process of changing the estimation range according to temperature was explained, but here, the estimation processing by the AI ​​model is explained as being changed according to temperature. As shown in Figure 8, the estimation mode determination unit 12 may determine the estimation mode such that, under normal conditions, it is in a normal mode that estimates whether one or more users in the detected image intend to enter the pedestrian crossing E1, and under high temperature conditions, it is in a low-load mode that estimates whether or not there are users in the detected image who intend to enter the pedestrian crossing E1. In this way, in the normal mode, more accurate estimation is achieved by estimating the intention of each user to enter, and in the low-load mode, the processing load is reduced by simply estimating whether or not there are users in the detected image who intend to enter, making it possible to perform estimation that is less likely to cause overheating of the control unit. Note that the estimation mode determination unit 12 may be in a normal mode that estimates the probability of an intention to enter the pedestrian crossing E1 under normal conditions, or in a low-load mode that estimates the probability of a user intending to enter the pedestrian crossing E1 being present under high temperature conditions.

[0048] FIG. 9 is a diagram for explaining a comparison between the normal mode and the low load mode. As shown in FIG. 9, in the normal mode, the estimation unit 13 inputs a plurality of detection images along the time series to an AI model that outputs whether one or more users Y in the detection image have the intention (crossing intention) to enter the crosswalk E1. In this case, the AI model outputs whether each user Y has the crossing intention. Also, as shown in FIG. 9, in the low load mode, the estimation unit 13 inputs a plurality of detection images along the time series to an AI model that outputs whether there is a user Y in the detection image who has the intention to enter the crosswalk E1. In this case, the AI model outputs whether there is a user Y with the crossing intention. Based on the output from such an AI model, the estimation unit 13 estimates whether the user Y will enter the crosswalk E1. In the present embodiment, the estimation unit 13 is described as inputting a plurality of detection images along the time series to an AI model that outputs whether one or more users Y in the detection image have the intention (crossing intention) to enter the crosswalk E1. However, the estimation unit 13 may input a plurality of detection images along the time series to an AI model that outputs the probability that one or more users Y in the detection image will enter the crosswalk E1. Also, in the present embodiment, the estimation unit 13 is described as estimating whether one or more users will enter the crosswalk E1. However, the estimation unit 13 may estimate the probability that one or more users will enter the crosswalk E1.

[0049] FIG. 10 is a flowchart for explaining the estimation process change process according to the temperature. As shown in FIG. 10, first, it is determined by the estimation mode determination unit 12 whether the temperature around the control unit 10 exceeds the first threshold value (step S11).

[0050] In step S11, if it is determined that the first threshold is not exceeded, the input image is acquired (step S2), and based on the input image, the estimation unit 13 performs normal mode AI processing (step S13), and the AI ​​model outputs the estimated result of the intention to cross for each user Y (pedestrian) (step S14).

[0051] In step S11, if it is determined that the first threshold is exceeded, the input image is acquired (step S15), and based on the input image, the estimation unit 13 performs low-load AI processing (step S16), and the AI ​​model outputs whether or not there is a user Y (pedestrian) who intends to cross the road (step S17).

[0052] Then, the estimation mode determination unit 12 determines whether the temperature around the control unit 10 is below the second threshold (step S18). In this way, the estimation mode determination unit 12 determines whether the temperature around the control unit 10 is below the second threshold while in the high-temperature estimation mode (low-load mode). If it is determined in step S8 that the temperature is below the second threshold, the estimation mode determination unit 12 determines the estimation mode so that the estimation process becomes the normal mode estimation process, and executes the processes from step S2 onward.

[0053] Next, the effects and benefits of the entry estimation system 1 according to this embodiment will be described.

[0054] The entry estimation system 1 includes a camera 20 that detects user Y on sidewalk E2 and outputs a detection image as the detection result, a control unit 10 that estimates whether user Y enters the crosswalk E1 from sidewalk E2 based on the detection image, and a temperature sensor 30 that measures the temperature around the control unit 10. The control unit 10 includes an acquisition unit 11 that acquires the detection image from the camera 20 and the temperature around the control unit 10 from the temperature sensor 30, an estimation mode determination unit 12 that determines the estimation mode based on the temperature around the control unit 10, an estimation unit 13 that estimates whether user Y enters the crosswalk E1 from sidewalk E2 based on the detection image using the estimation mode determined by the estimation mode determination unit 12, and an output unit 14 that outputs the estimation result from the estimation unit 13.

[0055] According to this intrusion estimation system 1, the temperature sensor 30 measures the temperature around the control unit 10, and based on this temperature, an estimation mode for intrusion is determined. Based on this estimation mode, intrusion estimation is performed using the user's detected image. In this way, by determining the estimation mode according to the temperature around the control unit 10, it becomes possible, for example, to perform intrusion estimation in a mode that prevents the control unit 10 from generating excessive heat if the temperature around the control unit 10 is rising. This means that even if the control unit 10 is installed outdoors and forced air cooling such as a fan cannot be used, the estimation mode can be changed according to the temperature around the control unit 10 to appropriately suppress the heat generation of the control unit 10.

[0056] In the above-described entry estimation system 1, the estimation mode determination unit 12 may determine the estimation mode such that, when the temperature around the control unit 10 is above a predetermined first threshold, the processing load on the estimation unit 13 is reduced compared to normal conditions when the temperature around the control unit 10 is not above the first threshold. With such a configuration, the processing load on the estimation unit 13 can be reduced at high temperatures, so that even when forced air cooling such as a fan cannot be used, the heat generation of the control unit 10 can be appropriately suppressed.

[0057] In the above-described entry estimation system 1, the estimation mode determination unit 12 may determine the estimation mode such that, at high temperatures, the estimation range, which is the range of the detected image used for estimation by the estimation unit 13, becomes narrower compared to normal conditions. By narrowing the estimation range at high temperatures, the processing load on the estimation unit 13 at high temperatures can be reliably reduced, and the heat generation of the control unit 10 can be appropriately suppressed.

[0058] In the above-described entry estimation system 1, the estimation range under normal conditions and the estimation range under high-temperature conditions may be set in advance so that they correspond to predetermined ranges in the detected image. By setting the estimation ranges under normal conditions and high-temperature conditions in advance in this way, it is possible to reliably reduce the processing load under high-temperature conditions and suppress variations in the estimation accuracy by the estimation unit 13.

[0059] In the above entry estimation system 1, the estimation range at high temperatures may be pre-set to include at least the area on the sidewalk E2 that is continuous with the pedestrian crossing E1. With such a configuration, even at high temperatures, the estimation range can include the area that is expected to be passed through when entering the pedestrian crossing E1 from the sidewalk E2 (an area where a user Y with the intention of entering the pedestrian crossing E1 can be detected), thereby ensuring the estimation accuracy of the estimation unit 13.

[0060] In the above-described entry estimation system 1, the estimation mode determination unit 12 may, when the temperature around the control unit 10 is below a predetermined second threshold while the estimation mode is set to high temperature, determine the estimation mode so that the estimation range is wider compared to the high-temperature state. In this way, when the temperature around the control unit 10 becomes sufficiently low and heat generation from the control unit 10 is no longer a problem, the estimation accuracy of the estimation unit 13 can be further improved by widening the estimation range.

[0061] In the above entry estimation system 1, the estimation unit 13 may estimate whether one or more users Y will enter the crosswalk E1 by inputting multiple detection images in a time series into an AI model that outputs whether or not one or more users in the detection images intend to enter the crosswalk E1. In this way, by using an AI model, entry estimation can be performed easily and with high accuracy.

[0062] In the above-described entry estimation system 1, the estimation mode determination unit 12 may determine the estimation mode such that, under normal circumstances, it is set to a normal mode in which it estimates whether one or more users Y in the detected image intend to enter the pedestrian crossing E1, and under high temperature conditions, it is set to a low-load mode in which it estimates whether or not there are users in the detected image who intend to enter the pedestrian crossing E1. In this way, under normal circumstances, the estimation is made more accurate by estimating the intention of each user Y to enter, and under high temperature conditions, the processing load is reduced by simply estimating whether or not there are users Y who intend to enter in the detected image, making it possible to perform estimation that is less likely to cause overheating of the control unit 10.

[0063] In the above-described entry estimation system 1, the estimation mode determination unit 12 may determine the estimation mode to be in normal mode when the temperature around the control unit 10 is below a predetermined second threshold while in low-load mode. In this way, when the temperature around the control unit 10 becomes sufficiently low and heat generation from the control unit 10 is no longer a problem, the estimation accuracy of the estimation unit 13 can be further improved by switching to normal mode, which allows for detailed estimation.

[0064] In the above entry estimation system 1, the estimation unit 13 may, in normal mode, input multiple detection images in a time series to an AI model that outputs whether or not one or more users in the detection images intend to enter the pedestrian crossing E1, and in low-load mode, input multiple detection images in a time series to an AI model that outputs whether or not there is a user Y in the detection images who intends to enter the pedestrian crossing E1, thereby estimating whether or not user Y enters the pedestrian crossing E1 from the sidewalk E2. In this way, by inputting detection images to AI models with different processing contents for each mode, entry estimation in each mode can be performed easily and with high accuracy.

[0065] The various embodiments and modifications described above may be combined as appropriate without departing from the spirit of this disclosure.

[0066] 1...Entry estimation system 2...Camera device 10...Control unit 11...Acquisition unit 12...Estimation pattern determination unit 13...Estimation unit 14...Output unit 20...Camera (detection unit) 30...Temperature sensor 50...Pole 60...Traffic signal 90...External device 120...Circuit 121...Processor 122...Memory 123...Storage 124...Input / output port DE1...Normal range DE2...Reduced range E1...Crosswalk E2...Sidewalk Y...User Y1...User Y2...User

Claims

1. An entry estimation system comprising: a detection unit that detects a user in a predetermined first area and outputs a detection image as the detection result; a control unit that estimates whether or not the user enters a predetermined second area from the first area based on the detection image; and a temperature sensor that measures the temperature around the control unit or the control unit itself, wherein the control unit includes: an acquisition unit that acquires the detection image from the detection unit and the temperature from the temperature sensor; an estimation mode determination unit that determines an estimation mode based on the temperature; an estimation unit that estimates whether or not the user enters the second area from the first area or the probability of entry based on the detection image using the estimation mode determined by the estimation mode determination unit; and an output unit that outputs the estimation result from the estimation unit.

2. The entry estimation system according to claim 1, wherein the estimation mode determination unit determines the estimation mode when the temperature is high and exceeds a predetermined first threshold, compared to normal conditions when the temperature is not above the first threshold, so as to reduce the processing load on the estimation unit.

3. The entry estimation system according to claim 2, wherein the estimation mode determination unit determines the estimation mode at high temperatures, compared to normal conditions, such that the estimation range, which is the range of the detected image used by the estimation unit for estimation, becomes narrower.

4. The entry estimation system according to claim 3, wherein the estimated range during normal conditions and the estimated range during high temperatures are set in advance so that they correspond to predetermined ranges in the detection image.

5. The entry estimation system according to claim 4, wherein the estimation range at high temperature is pre-set to include at least a region in the first area that is continuous with the second area.

6. The entry estimation system according to any one of claims 3 to 5, wherein the estimation mode determination unit determines the estimation mode such that, when the temperature is below a predetermined second threshold in the state of the estimation mode at high temperature, the estimation range is wider compared to the state at high temperature.

7. The entry estimation system according to any one of claims 1 to 6, wherein the estimation unit estimates whether or not one or more users in the detection images intend to enter the second area or the probability of them entering, by inputting the multiple detection images in a time series into an AI model that outputs whether or not one or more users in the detection images intend to enter the second area or the probability of them entering.

8. The entry estimation system according to any one of claims 2 to 6, wherein the estimation mode determination unit determines the estimation mode such that, in normal times, it sets up a normal mode in which it estimates whether or not one or more users in the detected image have the intention to enter the second area, or the probability that they do; and in high-temperature times, it sets up a low-load mode in which it estimates whether or not there are users in the detected image who have the intention to enter the second area, or the probability that they do.

9. The entry estimation system according to claim 8, wherein the estimation mode determination unit determines the estimation mode such that the system becomes the normal mode when the temperature is below a predetermined second threshold while the system is in the low-load mode.

10. The entry estimation system according to claim 8 or 9, wherein the estimation unit inputs a plurality of detection images in a time series into an AI model that outputs whether or not one or more users in the detection images intend to enter the second area or the probability of them entering, in the normal mode, and inputs a plurality of detection images in a time series into an AI model that outputs whether or not there are users in the detection images who intend to enter the second area or the probability of them entering, in the low load mode, thereby estimating whether or not a user enters the second area from the first area or the probability of them entering.

Images