Analysis program, analysis system, and analysis method

Abstract

To provide an analysis program, an analysis system, and an analysis method configured to generate point cloud data of target objects from a video of a mobile body, using image recognition AI, for visualizing the density and group shape and forming thunderclouds by heat map analysis, so as to be grasped at a glance.SOLUTION: An analysis program causes a computer to execute: Step S2 of identifying a mobile body, which is a detection target, from video data, and detecting a position of the detection target; Step of generating, from the video data, a first layer, which is a heat map with a range coefficient for the detection target, and a second layer with a range coefficient different from that of the first layer; Step S5 of generating cloud heat map images by superimposing at least the detection target, the first layer, and the second layer; and Step S7 of synthesizing the cloud heat map images generated at predetermined time intervals from the video of the video data to form a video of the cloud heat map image, and generating a thundercloud heat map video.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present invention relates to an analysis program, an analysis system, and an analysis method, and more particularly to an analysis program, an analysis system, and an analysis method using heat map analysis. [Background technology]

[0002] People flow analysis is a known method of analyzing flow using heat map analysis. People flow analysis is used for analyzing customer behavior and behavior patterns in commercial facilities and retail businesses, predicting traffic congestion and planning traffic by understanding people's movement patterns, designing safe passageways and avoiding excessive congestion in places where large numbers of people gather, such as events, and infectious disease countermeasures by analyzing people's density and contact movement patterns. In these places, people flow analysis is carried out using sensors, mobile cameras, cloud-based platforms, and AI technology.

[0003] The people flow analysis method described in Patent Document 1 performs clustering of analysis areas from OD data information, and derives movement data within clusters and movement data between clusters. Specifically, the OD data is converted into graph data in graph theory, with the points of the analysis area as nodes and people movements as edges, and clustering is performed using network analysis. A centering unit calculates the center point for each cluster, and based on these results, people movement data between clusters and people movement data within a cluster are derived. This makes it possible to grasp the macroscopic flow of people in the area being analyzed as well as the microscopic flow in that area.

[0004] The people flow analysis program described in Patent Document 2 reads the time from the location information of a mobile device, generates a travel route from the cleansed location information, determines the means of transportation, and performs an expanded estimation of the population in a specified area. Specifically, the location information of a wearable device such as a mobile device is cleansed to remove noise, and the means of transportation is estimated from the transportation hub corresponding to the location information. Finally, an expanded estimation is performed based on the penetration rate of mobile devices using statistical methods, and the number of samples is expanded to the population of the estimated area to estimate the flow of people in the population of the target area. This makes it possible to analyze people flow data over a wide range while preventing the processing load for analyzing people flow data from becoming too high. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Patent Publication No. 2021-47763 [Patent Document 2] JP 2023-26294 A Summary of the Invention [Problem to be solved by the invention]

[0006] However, the people flow analysis method described in Patent Document 1 and the people flow analysis program described in Patent Document 2 were not suitable for analyzing people flow in people flow footage of a busy intersection, for example, to determine the routes people take and in which areas people are densely packed.

[0007] In addition, at outdoor events and stadiums, there was a demand to understand the changes in people flow due to the placement of handicapped persons, the width and shape of the passageway, or the presence or absence of protective fences, but there was no established method that could visualize the density and flow of people over time at a glance. In addition to people flow, there was also a demand for technology that could detect the movement of individual objects from video footage, create a heat map of the entire video, and analyze how the density of each object changes depending on the location. Objects here include animals, fish, microorganisms such as plankton, hemoglobin flowing in blood, and other cells.

[0008] Therefore, the present invention aims to provide an analysis program, analysis system, and analysis method that can grasp the density and group shape of an aggregate at a glance by visualizing and turning images of moving objects into thunderclouds through heat map analysis. [Means for solving the problem]

[0009] In order to solve the above problem, the present invention provides an analysis program for analyzing video data of the flow of moving objects, which causes a computer to execute the following steps: identifying the type of moving object that is the detection target from the video data and detecting the position of the detection target; generating a first layer, which is a heat map in which a range coefficient of the detection target is set from the video data, and a second layer, which is a heat map in which the range coefficient different from that of the first layer is set; generating a cloud heat map image by overlaying at least the detection target, the first layer, and the second layer; and visualizing the cloud heat map image by synthesizing the cloud heat map image generated at a predetermined time interval from the video data, thereby generating a thundercloud heat map image.

[0010] The term "mobile object" used here refers to vehicles, animals, fish, birds, bacteria, microorganisms, cells, etc. that are placed in a specific environment.

[0011] According to another aspect of the present invention, there is provided an analysis program for analyzing video data of a flow of objects or people, the analysis program causing a computer to execute the steps of identifying detection targets including at least people from the video data and detecting the positions of the detection targets, generating a first layer which is a heat map in which a detection range of the detection targets is set from the video data and a second layer which is a heat map in which the detection range different from that of the first layer is set, generating a cloud heat map image by superimposing at least the first layer and the second layer, and visualizing the cloud heat map image by synthesizing the cloud heat map images generated at predetermined time intervals from the video data to generate a thundercloud heat map image, wherein the detection targets are included in at least one of the first layer or the second layer.

[0012] It is also preferable that the method further includes the steps of generating a canvas background in which a predetermined shaped background is drawn around the detected target based on past position information of the detected target, and generating the cloud heat map image by overlaying the canvas background, the first layer, and the second layer.

[0013] It is also preferable that a plurality of the canvas backgrounds are generated, and at least one of the background shape, the background color, and the time back in time to which the position of the detection target goes back of the canvas backgrounds are different from each other, and that the cloud heat map image is generated by overlaying the canvas background, the first layer, and the second layer with a preset transparency.

[0014] According to another aspect of the present invention, there is provided an analysis system for analyzing video data of a moving body, the analysis system comprising: an object recognition unit that identifies the type of the moving body that is the detection target from the video data and detects the position of the detection target; and a heat map generation unit that generates a first layer, which is a heat map in which a range coefficient of the detection target is set from the video data, and a second layer, which is a heat map in which the range coefficient different from that of the first layer is set, wherein the heat map generation unit generates a cloud heat map image by overlaying the detection target, the first layer, and the second layer, and visualizes the cloud heat map image by synthesizing the cloud heat map image generated at a predetermined time interval from the image of the video data, thereby generating a thundercloud heat map image.

[0015] According to another aspect of the present invention, there is provided an analysis method for analyzing video data of a moving body, comprising the steps of: identifying the type of the moving body that is the detection target from the video data and detecting the position of the detection target; generating a first layer which is a heat map in which a range coefficient of the detection target is set from the video data, and a second layer which is a heat map in which the range coefficient different from that of the first layer is set; generating a cloud heat map image by overlaying the detection target, the first layer, and the second layer; and visualizing the cloud heat map image by synthesizing the cloud heat map images generated at predetermined time intervals from the video data, thereby generating a thundercloud heat map image. Effect of the Invention

[0016] According to this configuration, the heat map image is generated by superimposing the first and second layers, which are heat map images with different range coefficients, so that the distribution of the heat map is clearer in a single image, and the density and group shape of people or objects can be grasped at a glance compared to a single-layer heat map image. In addition, the thundercloud heat map image is generated by synthesizing heat map images at a predetermined frame rate, so that the image data of moving objects is clouded, and the area where the density is locally high can be displayed as a thundercloud by instantaneously changing color. In other words, the density and group shape of people, which are difficult to grasp at a glance in an image of people flow, can be visualized by heat map analysis, and the image can be thundercloudized by superimposing multiple heat map images. Note that the more heat map images are increased, the clearer the thundercloud can be expressed. Furthermore, since the heat map image is generated by superimposing the detection target, the first layer, and the second layer, the detection target is displayed in the thundercloud heat map image, and the overall thundercloud situation can be confirmed by the density of the detection target. In addition, since the type of moving object being detected is identified and its location is detected, even if the video data contains multiple moving objects of different types, the density of each moving object can be visualized and a thundercloud can be created.

[0017] With this configuration, the first and second layers are overlaid on a canvas base on which the background of the detection target is drawn in a specified shape, making the position of the detection target clearer, resulting in a clearer heat map image and enabling highly accurate thundercloud generation of thundercloud heat map images.

[0018] With this configuration, the multiple canvas backgrounds differ in at least one of the background shape, color, and time history, so by appropriately setting these conditions, the way the thundercloud heat map image is displayed can be set as desired. Also, since the canvas background and the first and second layers are superimposed with a preset transparency, the way the thundercloud heat map image is displayed can be set as desired by setting the transparency.

[0019] According to the present invention, it is possible to provide an analysis program, an analysis system, and an analysis method that can grasp the density and the group shape of a group of moving objects at a glance by visualizing and transforming the objects into thunderclouds through heat map analysis of moving objects in an image. [Brief description of the drawings]

[0020] [Figure 1] FIG. 1 is a block diagram of an analysis system according to a first embodiment of the present invention. [Diagram 2] 3 is a flowchart of the analysis system according to the first embodiment of the present invention. [Diagram 3] 3 shows video data of the analysis system according to the first embodiment of the present invention. [Figure 4] 3 shows icon data of the analysis system according to the first embodiment of the present invention. [Diagram 5] 1 shows the canvas of the analysis system according to the first embodiment of the present invention. [Figure 6] FIG. 2 is a diagram showing the movement history of people on a campus in the analysis system of the first embodiment of the present invention. [Figure 7] FIG. 2 is a diagram showing icons of target position data of the analysis system according to the first embodiment of the present invention. [Figure 8] FIG. 2 is an image diagram of nine layers of canvas bases superimposed on one another in the analysis system according to the first embodiment of the present invention. [Figure 9] 2 shows a superimposed canvas base formed by superimposing nine layers of canvas bases of the analysis system of the first embodiment of the present invention. [Figure 10] 13 is a heat map image when the range coefficient in the heat map of the analysis system according to the first embodiment of the present invention is set to a minimum. [Figure 11] 13 is a heat map image when a range coefficient in the heat map of the analysis system according to the first embodiment of the present invention is enlarged. [Figure 12] 12 is a superimposed heat map image obtained by superimposing the heat map image of FIG. 10 and the heat map image of FIG. 11 of the analysis system according to the first embodiment of the present invention. [Figure 13]FIG. 13 is an image diagram showing how a clouded heat map image is generated by overlaying a canvas background and a heat map image in the analysis system of the first embodiment of the present invention. [Figure 14] 1 is a clouding heat map image in one frame of the analysis system of the first embodiment of the present invention. [Figure 15] 1 is a thundercloud heat map image of the analysis system according to the first embodiment of the present invention. [Figure 16] FIG. 13 is an overall view of an analysis system according to a third embodiment of the present invention. [Figure 17] FIG. 13 is a block diagram of an analysis system according to a third embodiment of the present invention. [Figure 18] FIG. 13 is a block diagram of an analysis system according to a modified example of the present invention. [Figure 19] 13 shows one frame of video that is the analysis target data of an analysis system according to another modified example of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0021] An analysis system 1 according to a first embodiment of the present invention will be described with reference to Fig. 1 to Fig. 15. Analysis system 1 of this embodiment visualizes the flow of people in event venues, stadiums, train stations, stores, tourist sites, commercial facilities, traffic roads in urban areas, etc., and makes it possible to grasp at a glance the changes in density and crowd shapes over time. Specifically, analysis system 1 analyzes video of an intersection in the city as shown in Fig. 3.

[0022] The information processing terminal 10, which is the analysis system 1, is a computer and has a control unit 2, a storage unit 3, a display unit 4, an operation unit 5, and a communication unit 6. The control unit 2 has an object recognition unit 21, a background generation unit 22, and a heat map generation unit 23. The storage unit 3 mainly stores analysis object data 31, object position data 32, a canvas background 33, a heat map image 34, and a thundercloud heat map image 35. The control unit 2 is a CPU (Central Processing Unit) and develops various data and applications stored in the storage unit 3.

[0023] The target recognition unit 21 automatically recognizes the type and position of the detection target from image data 31A obtained by extracting frames of analysis target data 31, which is video data shown in FIG. 3, by using artificial intelligence based on deep learning, and generates target position data 32 shown in FIG. 4. Specifically, preprocessing such as noise removal and color normalization of the image data 31A is performed, and feature extraction is performed hierarchically using a convolutional neural network or the like to convert the image data into higher-order features. Clustering of the detection target is performed by a classifier, and a target of a predetermined image in the image data 31A is detected. At this time, the model makes a prediction from the input image based on the labeled image data 31A, and the algorithm is reset so as to calculate the error between the prediction result and the correct answer and minimize the error. The model that has completed learning performs image recognition processing on the next image data 31A. For example, the targets recognized by the target recognition unit 21 are people, vehicles, motorcycles, strollers, and pets in city images, one team, the other team, a ball, spectators, reserve players, and referees in stadium images, and performers, spectators, and security guards in event venues. The identification target of the target recognition unit 21 is not limited to this, and can be set arbitrarily based on various analysis target data 31.

[0024] As shown in Fig. 4 and Fig. 7, the target recognition unit 21 plots a person 21A and a vehicle 21B, which are detection targets, on a two-axis graph and stores the coordinates of each. As shown in Fig. 7, the target recognition unit 21 also identifies the type of vehicle 21B, specifically, a private car, a bus, a truck, a taxi, a motorcycle, an electric scooter, a bicycle, etc., and displays it with a different icon. The person 21A and the vehicle 21B are examples of the moving body of the present invention.

[0025] As shown in Figs. 5 to 9, the background generation unit 22 generates a canvas background 33 that displays a circle having a predetermined diameter centered on the current or past position of the person 21A based on the target position data 32 generated by the target recognition unit 21. The canvas background 33 shown in Fig. 5 is centered on the position of the person 21A going back a predetermined time, as shown by the circle indicated by the dotted line in Fig. 6. The predetermined time, the diameter of the circle, the color of the circle, and the transparency of the circle of the canvas background 33 are elements related to the appearance of the thundercloud heat map image 35, and are therefore set to predetermined values ​​in advance. Here, it is desirable to change the color of the circle for detection targets of different types within the same canvas background 33. When changing the appearance of the thundercloud heat map image 35, a desired image can be obtained by arbitrarily changing various conditions of the canvas background 33.

[0026] The canvas background 33 shown in FIG. 5 represents a circle of a person 21A represented by a solid line from the current position in FIG. 4, represented by a solid line, a predetermined time in the past, represented by a dotted line. The person 21A moves from the position represented by the dotted line to the position represented by the dashed line, and reaches the position represented by the solid line. The background generation unit 22 divides the time from the dotted line to the actual movement into nine parts, and generates nine canvas backgrounds 33 depicting the circle of the person 21A at each position. Specifically, the background generation unit 22 generates nine canvas backgrounds 33 that differ in at least one condition of a predetermined time, a circle diameter, a circle color, and a circle transparency. The person 21A and the vehicle 21B are displayed on each canvas background 33 based on the target position data 32. FIG. 9 shows a superimposed canvas background 33A obtained by superimposing nine canvas backgrounds 33 with a preset transparency as shown in FIG. 8. By superimposing a plurality of canvas backgrounds 33 like the superimposed canvas background 33A, a background can be generated that expresses the density of individual people 21A and vehicles 21B as if they were clouds.

[0027] 10 to 12, the heat map generation unit 23 generates a heat map image 34 centered on the current positions of the person 21A and the vehicle 21B based on the target position data 32 generated by the target recognition unit 21. The heat map generation unit 23 calculates the density of the detected targets from the position of the person 21A by Paze estimation and probability statistics based on a preset range coefficient, and generates the heat map image 34.

[0028] The range coefficient indicates a range in which the coordinate point cloud of the detection target such as the person 21A and the vehicle 21B is captured in the heat map. When the range coefficient is set to a large value, the heat map generating unit 23 captures the coordinate point cloud of the detection target as a group, calculates the density in a wide range by Paise estimation, and colors the detection target as high density and low density. In other words, when the range coefficient is set to a large value, the detection target tends to be captured as a group, and the rough density in the heat map image 34 can be grasped. When the range coefficient is set to a small value, the heat map generating unit 23 captures the coordinate point cloud of the detection target as a group in a small range, calculates the density in a narrow range by Paise estimation, and colors the detection target as high density and low density. In other words, when the range coefficient is set to a small value, the detection target tends to be captured as a point, and the density of each detection target in the heat map image 34 can be grasped.

[0029] The heat map generating unit 23 performs pre-processing to remove noise, missing values, etc., and constructs a probability distribution model that expresses a heat map based on the target position data 32 that indicates the position of the person 21A using the probability distribution of Pays estimation. A prior distribution based on a probability distribution is set using Pays estimation, and a posterior distribution is calculated using Pays' theorem to obtain an estimate of the density of the people flow. The positions of the people flow are probabilistically estimated by sampling from the posterior distribution, and these results are aggregated to generate a heat map image 34 of the people flow at each point as shown in Figures 10 and 11. Various parameters used in generating the heat map image 34 can be set arbitrarily depending on the state of the obtained image.

[0030] In this embodiment, the heat map generating unit 23 generates a heat map image 34 using a graph of a two-dimensional Student's t distribution. In this embodiment, areas where detection targets such as people 21A and vehicles 21B are densely concentrated are displayed in a darker color because they have a higher probability of appearing, and areas where detection targets such as people 21A and vehicles 21B are sparsely concentrated are displayed in a darker color because they have a lower probability of appearing. The heat map image 34 shown in FIG. 10 is generated by individually grasping the coordinate points of the detection targets with the range coefficient set to a minimum. The heat map image 34 shown in FIG. 11 is generated by expanding the range coefficient and grasping the coordinate points of the detection targets as a group.

[0031] Fig. 12 shows a superimposed heat map image 34A obtained by superimposing the heat map image 34 with the range coefficient shown in Fig. 10 set to the minimum and the heat map image 34 with the range coefficient shown in Fig. 11 set to the expanded range coefficient. Either the heat map image 34 with the range coefficient shown in Fig. 10 set to the minimum or the heat map image 34 with the range coefficient shown in Fig. 11 set to the expanded range coefficient is an example of the first layer of the present invention, and the other is an example of the second layer of the present invention.

[0032] The storage unit 3 is composed of a hard disk drive (HDD), a read only memory (ROM), and a random access memory (RAM). The HDD stores various applications, various data required for analyzing the flow of people or goods, and programs such as the OS. The RAM is a volatile memory and is used as a working area for programs, and the ROM is a non-volatile memory and is used as a storage area for programs.

[0033] The analysis target data 31 in the storage unit 3 is data that is the subject of analysis of the flow of moving objects, and is an image of an intersection in a city as shown in Fig. 3. The target position data 32 is data shown in Fig. 4 that is generated by the target recognition unit 21, and includes the type of detection target and the coordinates that are the position of the detection target. The analysis target data 31 is an example of the image data of the present invention.

[0034] The canvas background 33 is the data shown in Fig. 5, Fig. 8 and Fig. 9 generated by the background generation unit 22, and nine images are generated in this embodiment. The heat map image 34 is the data shown in Fig. 10 to Fig. 12 generated by the heat map generation unit 23, and two images are generated in this embodiment. The canvas background 33 and the heat map image 34 include icons and position information of the person 21A and the vehicle 21B of the target position data 32.

[0035] The display unit 4 is a display such as an LCD (Liquid Crystal Display) or an organic EL display, and displays information processed by the control unit 2. The operation unit 5 is an input interface for a user who uses the information processing terminal 10 to input operations, and is composed of a pointing device such as a mouse, a keyboard, a touch panel, a sensor, etc. The communication unit 6 communicates with external terminals via a network.

[0036] Next, the flow of analysis by the analysis system 1 will be described with reference to Fig. 2. The control unit 2 extracts a predetermined frame image from the analysis target data 31, which is an image stored in the storage unit 3 of the information processing terminal 10 (S1). In this embodiment, the frame rate is set to 30 FPS, and all images are subjected to heat mapping processing and visualized. The frame rate is not limited to this, and can be set arbitrarily depending on the resolution of the generated thundercloud heat map image 35 of people flow, the performance of the information processing terminal 10, etc.

[0037] The target recognition unit 21 recognizes the type and position of the person 21A and vehicle 21B that are the detection targets from the extracted image data 31A using AI, and generates target position data 32 shown in Fig. 4 (S2). At this time, the target recognition unit 21 recognizes the coordinates of the positions of the person 21A and vehicle 21B on the graph, plots icons on the graph, and stores them in the storage unit 3 together with the coordinate data (S3).

[0038] The background generation unit 22 generates a canvas background 33 based on the past positions of the person 21A and the vehicle 21B as shown in Fig. 5 based on the person 21A and the vehicle 21B of the target position data 32, and generates nine canvas backgrounds 33 with different times, circle diameters, circle colors, transparency, etc. as shown in Fig. 7 (S4). The heat map generation unit 23 generates a heat map image 34 shown in Fig. 10 in which the range coefficient, which is a heat map parameter, is set to a minimum, and a heat map image 34 shown in Fig. 11 in which the range coefficient is enlarged, based on the person 21A and the vehicle 21B of the target position data 32 (S5).

[0039] As shown in FIG. 13, the heat map generating unit 23 superimposes nine layers of canvas background 33 and two layers of heat map image 34 to obtain a cloud heat map image 34B shown in FIG. 14. At this time, the heat map generating unit 23 sets the transparency of each layer to a preset value. When the cloud heat map image 34B is generated for all frames, the control unit 2 connects all frames to obtain a thundercloud heat map image 35 (S7). The obtained thundercloud heat map image 35 is stored in the storage unit 3 and visualized as shown in FIG. 15 and displayed on the display unit 4 as a moving image.

[0040] According to this configuration, the clouded heat map image 34B is generated by superimposing the heat map image 34 shown in FIG. 10 with the minimum range coefficient and the heat map image 34 shown in FIG. 11 with the enlarged range coefficient, so that the distribution of the heat map is clearer in a single image, and the density and group shape of people or objects can be grasped at a glance compared to a single-layer heat map image. In addition, since the heat map image 34 is visualized at a predetermined frame rate, the image data of the moving objects, the person 21A and the vehicle 21B, are clouded, and the area where the density is locally high can be expressed as a thundercloud by instantaneously changing color. In other words, the density and group shape of people, which are difficult to grasp at a glance in the image of the flow of people, can be visualized by heat map analysis, and the image can be made into a thundercloud by superimposing multiple heat map images 34. Note that the more heat map images 34 are used, the clearer the thundercloud can be expressed. Furthermore, a clouded heat map image 34B is generated by superimposing the people 21A and vehicles 21B as the detection targets, a heat map image 34 shown in FIG. 10 with a minimum range coefficient, and a heat map image 34 shown in FIG. 11 with an expanded range coefficient, so that the detection targets are displayed in a visualized thundercloud heat map image 35, and the overall thundercloud situation can be confirmed based on the density of the detection targets. In addition, because the types of the people 21A and vehicles 21B as the detection targets are identified and their positions are detected, even in the analysis target data 31 in which multiple detection targets of different types exist, the density of each detection target can be visualized and thunderclouded.

[0041] With this configuration, a canvas background 33 in which the background of the detection target is drawn with a circle is overlaid on a heat map image 34 shown in Figure 10 with a very small range coefficient, and a heat map image 34 shown in Figure 11 with an expanded range coefficient, so that the position of the detection target becomes clearer, resulting in a beautiful cloud heat map image 34B, and a high-precision thundercloud heat map image 35 can be obtained.

[0042] With this configuration, the nine canvas backgrounds 33 differ in at least one of the background shape, color, and time going back, so that by appropriately setting these conditions, it is possible to arbitrarily set the way the thundercloud heat map image 35 is displayed. Also, the canvas background 33, the heat map image 34 shown in Fig. 10 with a very small range coefficient, and the heat map image 34 shown in Fig. 11 with an expanded range coefficient are superimposed with a preset transparency, so that it is possible to arbitrarily set the way the thundercloud heat map image 35 is displayed by setting the transparency.

[0043] Next, a second embodiment of the present invention will be described. The same components as those in the first embodiment are given the same reference numerals and the description thereof will be omitted.

[0044] In the analysis system 1 of the first embodiment, the canvas background 33 and the heat map image 34 are generated based on the position data of the target position data 32, and therefore icons of the person 21A and the vehicle 21B are displayed on the canvas background 33 and the heat map image 34. In contrast, in the second embodiment, icons of the person 21A and the vehicle 21B are not displayed on the canvas background 33 and the heat map image 34. In the process of S8, the control unit 2 generates a cloud heat map image 34B by superimposing the target position data 32, the nine-layer canvas background 33, and the two-layer heat map image 34. The control unit 2 generates the cloud heat map image 34B for all frames of the analysis target data 31 and joins them together to generate a thundercloud heat map video 35.

[0045] Next, a third embodiment of the present invention will be described with reference to Figures 16 and 17. The same components as those in the above-mentioned embodiment will be given the same reference numerals and the description thereof will be omitted.

[0046] The analysis system 201 is composed of an information processing terminal 210, an external server 211, an imaging unit 212, and a communication network 213. The information processing terminal 110, the external server 211, and the imaging unit 212 are capable of bidirectional communication via the communication network 213. In the analysis system 1 of the first embodiment, the flow of people in the analysis target data 31 in the storage unit 3 is analyzed, but in the analysis system 201 of the third embodiment, the image captured by the imaging unit 212 is temporarily stored in the external server 211, and the control unit 2 of the external server 211 generates a thundercloud heat map image 35 in real time.

[0047] 17, the external server 211 has a control unit 2, a storage unit 3, an operation unit 5, and a communication unit 6. The information processing terminal 210 is a computer, a smartphone, a tablet, a dedicated terminal, or the like, and has a control unit 202, a storage unit 203, a display unit 204, an operation unit 205, and a communication unit 206.

[0048] The image captured by the imaging unit 212 is stored in the memory unit 3 of the external server 211 via the communication network 213, and the control unit 2 generates a thundercloud heat map image 35. The thundercloud heat map image 35 can be viewed on the display unit 204 of the information processing terminal 210 via the communication network 213.

[0049] According to such a configuration, the image of the imaging unit 212 can be converted into a heat map image in real time, and by increasing the processing power of the external server 211, even an information processing terminal 210 with low power can view the thundercloud heat map image 35. Furthermore, since the information processing terminal 210 views the thundercloud heat map image 35 stored in the external server 211 via the communication network 213, the information processing terminal 210 can be carried around and viewed anywhere. Furthermore, the thundercloud heat map image 35 can be viewed simultaneously on multiple information processing terminals 210.

[0050] The analysis program, analysis system, and analysis method according to the present invention are not limited to the above-described embodiment, and various modifications are possible within the scope of the gist of the invention described in the claims.

[0051] In the above embodiment, the human circle is generated based on the past positions of the person 21A and the vehicle 21B as the canvas base 33, but is not limited to this. For example, the direction in which the person 21A and the vehicle 21B may move in the future may be predicted by a statistical method, and the human circle may be generated based on the future positions of the person 21A and the vehicle 21B.

[0052] In the above embodiment, nine canvas backgrounds 33 were generated, two heat map images 34 were generated, and 11 pieces of data were superimposed to obtain the cloud heat map image 34B, but the number of images can be set arbitrarily. For example, as shown in FIG. 18, the analysis system 301 does not have a background generation unit and has only the heat map generation unit 23. When the control unit 2 generates the superimposed heat map image 34, at least two heat map images 34 are superimposed. This makes it possible to omit the processing process by the background generation unit, so that even an information processing terminal with low performance can quickly generate the thundercloud heat map image 35. In addition, by creating three or more heat map images 34 and synthesizing them in the heat map generation unit 23, a clearer thundercloud heat map image 35 can be generated.

[0053] In the above embodiment, the background generating unit 22 generates nine canvas backgrounds 33 that differ in at least one condition, namely, a predetermined time, a diameter of a circle, a color of a circle, and a transparency of a circle, but this is not limited to this. The background generating unit 22 may create at least one or more canvas backgrounds 33 by setting various conditions to desired values.

[0054] In the above embodiment, the cloud heat map image 34B is generated for one frame, and then the cloud heat map image 34B is connected and visualized when the cloud heat map image 34B is generated for all frames, but the present invention is not limited to this. The analysis system only needs to be able to output a heat map image, and the object recognition unit 21 may recognize the detection object for all frames, the background generation unit 22 may generate the canvas background 33 for all frames, the heat map generation unit 23 may generate the heat map image 34 for all frames, the cloud heat map image 34B for all frames, and finally the thundercloud heat map image 35 may be generated.

[0055] In the above embodiment, the video of people flow shown in FIG. 2 is used as the analysis target data 31, but this is not limited thereto. The artificial intelligence of the target recognition unit 21 can recognize and identify various people and objects through image recognition, and can be applied to fields other than people flow analysis. For example, a video of a school of fish as shown in FIG. 19 can be analyzed by the analysis system 1, and a video of a thundercloud can be obtained by heat map analysis. By analyzing a video of a school of animals with the analysis system 1, the target recognition unit 21 can recognize the animals and a video of a thundercloud can be obtained by heat map analysis. In addition, by applying it to the medical field, the target recognition unit 21 can recognize hemoglobin flowing through the blood from a blood vessel image, and the flow of hemoglobin in the blood vessel can be visualized by a thundercloud. In addition, by analyzing a video of the disintegration and decomposition of a cell group or the process of the formation of an aggregate with the analysis system 1, the target recognition unit 21 can learn and detect specific cells, bacteria, etc., and visualize the process and express it as a thundercloud.

[0056] In the above embodiment, a circle with the detection target at its center is used as the canvas background 33, but this is not limited to this. Any shape, such as a polygon, cloud, or teardrop shape, can be selected with the detection target at its center, depending on the performance of the information processing terminal 10 and the appearance of the thundercloud heat map image 35. [Explanation of symbols]

[0057] 1, 201, 301 Analysis System 2. Control section 3 Storage section 4 Display section 21 Target Recognition Unit 22 Background Generation Section 23 Heat map generation section 31 Data to be analyzed 32 Target location data 33 Canvas Base 34 Heatmap images 34A Overlaid Heat Map Data 34B Cloud heat map image 35 Thunderstorm heat map video

Claims

1. An analysis program for analyzing video data of a moving object, On the computer, identifying a type of the moving object that is a detection target from the video data and detecting a position of the detection target; generating a first layer, which is a heat map in which a range coefficient of the detection object is set, from the video data, and a second layer, which is a heat map in which the range coefficient different from that of the first layer is set; A step of generating a cloud heat map image by superimposing at least the detection object, the first layer, and the second layer; An analysis program characterized by executing a step of visualizing the cloud heat map image by synthesizing the cloud heat map images generated at predetermined time intervals from the images of the video data, thereby generating a thundercloud heat map image.

2. An analysis program for analyzing video data of a moving object, On the computer, identifying a type of the moving object that is a detection target from the video data and detecting a position of the detection target; generating a first layer, which is a heat map in which a detection range of the detection object is set, from the video data, and a second layer, which is a heat map in which the detection range different from that of the first layer is set; generating a clouding heatmap image by overlaying at least the first layer and the second layer; and generating a thundercloud heat map image by synthesizing the cloud heat map image generated at a predetermined time interval from the image of the image data, The analysis program, characterized in that the detection object is included in at least one of the first layer or the second layer.

3. generating a canvas background in which a background of a predetermined shape is drawn on the detected object based on past position information of the detected object; The analysis program according to claim 1 or 2, further comprising a step of generating the cloud heat map image by overlaying the canvas background, the first layer, and the second layer.

4. A plurality of the canvas backgrounds are generated, and at least one of the shape of the background of the canvas background, the color of the background, and the time that the position of the detection target goes back is different from each other; The analysis program according to claim 3, characterized in that the cloud heat map image is generated by overlaying the canvas background, the first layer, and the second layer with a predetermined transparency.

5. An analysis system for analyzing video data of a moving object, comprising: an object recognition unit that identifies a type of the moving object that is a detection object from the video data and detects a position of the detection object; a heat map generating unit that generates a first layer, which is a heat map in which a range coefficient of the detection object is set, from the video data, and a second layer, which is a heat map in which the range coefficient different from the first layer is set; The heat map generation unit generates a cloud heat map image by overlaying the detection object, the first layer, and the second layer, and visualizes the cloud heat map image by synthesizing the cloud heat map image generated at a predetermined time interval from the image of the video data, thereby generating a thundercloud heat map image.

6. An analysis method for analyzing video data of a moving object, comprising: identifying a type of the moving object that is a detection target from the video data and detecting a position of the detection target; generating a first layer, which is a heat map in which a range coefficient of the detection object is set, from the video data, and a second layer, which is a heat map in which the range coefficient different from that of the first layer is set; generating a cloud heat map image by superimposing the detection object, the first layer, and the second layer; and a step of visualizing the cloud heat map image by synthesizing the cloud heat map image generated at a predetermined time interval from the image of the image data to generate a thundercloud heat map image.

Images