Information system

Abstract

To provide an information system capable of searching for an article by performing highly accurate image analysis in an analysis server 1.SOLUTION: An information system X includes a terminal 2 for transmitting image data by network communication, and an analysis server 1 for performing image analysis of the image data transmitted from the terminal 2. The camera 21 captures high-quality image data. The transmission unit 20 transmits high-quality image data captured by the camera 21 in real time while maintaining the high quality. The receiving unit 10 receives high-quality image data transmitted from the terminal 2. The adjusting unit 11 detects that the high-quality image data received by the receiving unit 10 has a characteristic that can decrease the analysis accuracy, or detects a characteristic of any of blocked-up shadows, blown-out highlights, and out-of-focus blur, and adjusts the high-quality image data in accordance with the detected characteristic. The image analysis unit 12 searches for a target article by high-accuracy image analysis from the high-quality image data adjusted by the adjustment unit 11.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention particularly relates to information systems that perform image analysis. [Background technology]

[0002] Conventionally, there have been cases where it has been necessary to find an item such as a part or a tool inside a factory, a warehouse, etc. However, the factory, warehouse, etc. may not always be organized with shelves, and it may take time to find an item.

[0003] For this reason, it is conceivable to create an information system in which image data from a device such as a wearable camera attached to a helmet is sent to an analysis server, and the analysis server analyzes the images to search for items. In this case, when transferring image data from a terminal using Wi-Fi (wireless LAN), there is a problem that there is a limit to the transfer bit rate, making it difficult to transmit high-quality video.

[0004] Here, referring to Patent Document 1 as a conventional transmission method, a technology is disclosed in which, when the transmission delay of image data of high importance becomes large, the image encoding rate is changed and transmitted at a low encoding rate, and image quality improvement processing is performed on the display side, thereby mitigating the transmission delay of the image of high importance and providing a high-quality displayed image (hereinafter referred to as the "conventional technology"). [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Laid-Open No. 2012-89989 Summary of the Invention [Problem to be solved by the invention]

[0006] However, when image data is transmitted after being degraded by encoding, as in the conventional technology, the image can become blurred, making it difficult to perform high-precision image analysis on the analysis server and making it difficult to find the item.

[0007] The present invention has been made in consideration of the above circumstances, and aims to provide an information system that can perform high-precision image analysis on an analysis server and search for items, thereby solving the above-mentioned problems. [Means for solving the problem]

[0008] The information system of the present invention is an information system comprising a terminal that transmits image data via network communication and an analysis server that performs image analysis of the image data transmitted from the terminal, wherein the terminal comprises a camera that captures high-quality image data and a transmission unit that transmits the high-quality image data captured by the camera in real time while maintaining the high quality, and the analysis server comprises a receiving unit that receives the high-quality image data transmitted from the terminal, an adjustment unit that detects characteristics of the high-quality image data received by the receiving unit that may reduce analysis accuracy, or detects any of the characteristics of black crush, white blowout, and out-of-focus, and adjusts the high-quality image data according to the detected characteristics, and an image analysis unit that searches for a target item using high-precision image analysis from the high-quality image data adjusted by the adjustment unit. The information system of the present invention is characterized in that the high-quality image data includes any one or any combination of high-definition data, low-compression data, uncompressed data, 3D stereoscopic data, point cloud data, HDR data, high color gamut data, wide gradation data, and multi-focal data. The information system of the present invention is characterized in that the adjustment unit optimizes any one or any combination of the resolution or color gamut, brightness, contrast, focus, stereoscopic viewpoint, and analysis target range of the analysis image. The information system of the present invention is characterized in that the image analysis unit divides images of objects located at different three-dimensional positions based on the 3D stereoscopic data or the point cloud data, and then analyzes the images. [Effects of the Invention]

[0009] According to the present invention, an information system can be provided that transmits high-quality image data captured by a camera in real time while maintaining the high quality, detects characteristics of the received high-quality image data that may reduce analytical accuracy, or detects any of the characteristics of crushed blacks, blown-out highlights, and out-of-focus, adjusts the high-quality image data according to the detected characteristics, and reliably performs high-precision image analysis to enable the search for the desired item. [Brief explanation of the drawings]

[0010] [Figure 1] 1 is a system configuration diagram of an information system X according to an embodiment of the present invention. [Figure 2] 10 is a flowchart showing the flow of an item search process according to an embodiment of the present invention. [Figure 3] 3 is a photograph showing an example of the adjustment process shown in FIG. 2. [Figure 4] 3 is a photograph showing an example of a depth map of the adjustment process shown in FIG. 2. DETAILED DESCRIPTION OF THE INVENTION

[0011] <Embodiment> Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 shows a schematic configuration of an information system X according to this embodiment. In this embodiment, the information system X includes a terminal 2 that transmits image data via network communication, and an analysis server 1 that performs image analysis on the image data transmitted from the terminal 2.

[0012] Terminal 2 may be a wearable camera attached to a helmet, an action cam and communication device, a mixed reality (MR) or virtual reality (VR) headset including a camera, a smartphone, a tablet device, a portable personal computer (PC), a dedicated terminal, etc.

[0013] The analysis server 1 is a PC server, a dedicated server, or the like for image analysis. The analysis server 1 may not be a single server, but may be composed of multiple devices forming a cluster, etc. In addition, the analysis server 1 may include the functions of a management server that manages items such as parts and tools in factories, warehouses, etc., or may be linked to such a management server.

[0014] More specifically, the terminal 2 includes a camera 21 and a transmitter 20. The analysis server 1 includes a receiving unit 10, an adjusting unit 11, an image analyzing unit 12, an input unit 13, and a display unit 14.

[0015] The camera 21 is an imaging device including a charge-coupled device (CCD) image sensor or a complementary metal-oxide-semiconductor (CMOS) image sensor that captures high-quality image data, and an optical lens. In this embodiment, the camera 21 may be a wearable camera, an action camera, a camera in an MR or VR headset, or a built-in camera in a smartphone, a tablet terminal, or a mobile PC. Furthermore, in this embodiment, the camera 21 may include a stereo camera, a multi-focal camera, a light field camera, a hologram camera, an infrared camera, a light detection and ranging (LiDAR), a terahertz wave radar, a millimeter wave radar, or the like that can acquire three-dimensional images including a depth map and distance data.

[0016] In this embodiment, the high-quality image data captured by the camera 21 includes any one or any combination of high-definition data, low-compression data, uncompressed data, 3D stereoscopic data, point cloud data, HDR data, high color gamut data, wide gradation data, and multi-focal data.

[0017] Among these, the high-definition data may be, for example, image data with a resolution of 4K (2160p), 8K (4320p), or higher. The low-compression data and uncompressed data may be image data encoded at low compression or no compression (RAW). Here, the data may be so-called "RAW" data, in which low-compression data is encoded at low compression on a frame-by-frame basis. The low-compression encoding may include, for example, video compression formats such as H.264, H.265, and MJPEG, as well as run-length and Huffman coding compression for transmission. Alternatively, the low-compression data may be video image data with a bit rate higher than the bit rate of a compression format used in a typical wearable camera or the like. The 3D data or point cloud data is stereoscopic image data, and may be distance (depth) data for each pixel in the image, or data indicating the XYZ coordinates of each vertex of a polygon of each object in the image data. The HDR data, i.e., high color gamut data or wide gradation data, may be HDR (High Dynamic Range), data with a color gamut wider than the usual 8-bit R (Red), G (Green), and B (Blue) data, or data with a wide gradation. For example, in this embodiment, 12-32-bit, fixed or floating-point color gamut data for each RGB color may be used. The multifocal data is image data that can be adjusted for focus in post-processing. Specifically, the multifocal data may be data that can acquire images with multiple focuses, such as light field images and hologram images, that can be focused on multiple points using focus synthesis.

[0018] In this embodiment, the image data may be either still image data or moving image data. In the case of moving image data (video data), the FPS (Frames Per Second) may be any value, such as a few frames to several hundred frames per second.

[0019] Transmitter 20 is a wireless device capable of transmitting high-quality image data captured by camera 21 in real time while maintaining the high quality. For example, transmitter 20 may be capable of transferring image data via 5G (5th Generation Mobile Communication System), 6G (6th Generation Mobile Communication System), or a higher-speed mobile network, or other high-speed wireless communication. Alternatively, in the case of Wi-Fi, transmitter 20 may be capable of transmitting data via a high-speed wireless LAN standard such as Wi-Fi 6 / Wi-Fi 6E (IEEE 802.11ax) or W-Fi 7 (IEEE 802.11be). Hereinafter, these will be simply referred to as "high-speed wireless lines."

[0020] In addition to these, the terminal 2 may also be equipped with an intercom that receives audio from the analysis server 1, a display that indicates the location of an item, a laser pointer, etc. (hereinafter referred to as "intercom, etc."). Therefore, the transmission unit 20 may be capable of receiving information from the analysis server 1 and presenting it to the user as a transmission / reception means.

[0021] The receiving unit 10 is a wireless device capable of receiving high-quality image data transmitted from the terminal 2. In this case, the receiving unit 10 can acquire the high-quality image data in the state in which it was transmitted from the transmitting unit 20 of the terminal 2.

[0022] The adjustment unit 11 is a control and calculation unit that adjusts the high-quality image data received by the receiving unit 10. The adjustment unit 11 may be configured, for example, with a CPU (Central Processing Unit), an MPU (Micro Processing Unit), a DSP (Digital Signal Processor), a GPU (Graphics Processing Unit), an NPU (Neural Processing Unit), an ASIC (Application Specific Processor), or the like.

[0023] In this embodiment, the adjustment unit 11 detects characteristics that may reduce the analytical accuracy of the image data through element analysis of the image using various heuristics, or detects any of the characteristics of black crush, white blowout, and out-of-focus, and adjusts the high-quality image data according to the detected characteristics. Specifically, the adjustment unit 11 can optimize any one or any combination of the resolution or color gamut, brightness, contrast, focus, stereoscopic viewpoint, and analysis target range of the analysis image.

[0024] The image analysis unit 12 is a control and calculation unit that searches for a target item through high-precision image analysis from the high-quality image data adjusted by the adjustment unit 11. The image analysis unit 12 is configured with, for example, a CPU, an MPU, a DSP, a GPU, an NPU, an ASIC, etc. The image analysis unit 12 is capable of executing various image recognition models using, for example, deep learning (DL). These various models may be, for example, a convolutional neural network (NN), end-to-end object detection with transformers (DETR), etc. This enables the image analysis unit 12 to tag and recognize the coordinates of objects (objects) of items such as parts and tools in the image data (hereinafter simply referred to as "searching for items").

[0025] Here, in this embodiment, the image analysis unit 12 may be capable of dividing images of objects located at different three-dimensional positions based on the 3D stereoscopic data or point cloud data, and then analyzing the images. Furthermore, the image analysis unit 12 may be capable of writing out standard image data (described later) from the high-quality image data and performing initial analysis based on this.

[0026] The input unit 13 is a keyboard, a pointing device such as a mouse, a trackball, a touchpad, or a touch panel, or other devices for inputting information.

[0027] The display unit 14 is a display device such as an LCD (Liquid Crystal Display), an organic EL (Organic Electro Luminescence, Organic Light Emitting Diode) display, an LED (Light Emitting Diode) display, electronic paper, or a fluorescent display tube. The input unit 13 and the display unit 14 may be integrally formed as a touch panel display or the like.

[0028] [Item search processing by information system X] Next, with reference to FIGS. 2 to 4, the item search process using the information system X according to the embodiment of the present invention will be described in more detail. In the item search process according to this embodiment, first, high-quality image data is captured by the terminal 2. The high-quality image data captured by the camera 21 is transmitted in real time while retaining its high quality. The high-quality image data transmitted from the terminal 2 is received by the analysis server 1. The received high-quality image data is detected to have characteristics that may reduce analysis accuracy, or to have characteristics such as crushed shadows, blown-out highlights, or out-of-focus, and the high-quality image data is adjusted according to the detected characteristics. The target item is then searched for using high-precision image analysis from the adjusted high-quality image data. Below, the item search process by this information system X will be explained in detail for each step using the flowchart in Figure 2. These processes are calculated by the respective control and calculation units of the terminal 2 and the analysis server 1, and are executed using their respective hardware resources.

[0029] First, in step S200, the camera 21 of the terminal 2 performs an image capturing process. When an instruction to search for an item is received from a user such as a worker working in a factory or warehouse, the camera 21 captures an image of the surroundings. At this time, the camera 21 captures high-quality image data and transmits it to the transmission unit 20. For example, the camera 21 may be a wearable camera attached to a user's helmet or the like, and capture the following image data: High bit rate ~ Uncompressed (for high quality images) 4K / 8K 3D stereoscopic data (high definition, multi-information applications) HDR~14bit RAW (to avoid blown-out highlights and crushed shadows) Multi-focus image (to avoid blurry images)

[0030] Specifically, for example, it is possible to acquire image data of 8K / 30p (ProRes RAW HQ 12bit) or 14bit RAW video with a high bit rate of approximately 3Gbps.

[0031] Next, in step S201, the transmitting unit 20 performs a transmission process. The transmission unit 20 is connected to the analysis server 1 via a high-speed wireless line and transmits high-quality image data captured by the camera 21 in real time at high speed while maintaining the high quality and minimizing processing load. For example, the transmission unit 20 performs low-load encoding and transmits clear, high-definition, high-quality image data via a 5G line. This makes it possible to transmit high-quality image data from the user to the analysis server 1. Furthermore, the transmission unit 20 transmits the data while minimizing the load of the encoding process on the sending side, thereby reducing the processing load.

[0032] Here, the processing of the analysis server 1 will be explained. In step S100, the receiving unit 10 performs image data receiving processing. The receiving unit 10 is connected to the terminal 2 and receives high-definition image data transmitted from the transmitting unit 20 of the terminal 2 in real time while maintaining the high quality. The receiving unit 10 stores the received image data in a buffer of a temporary recording medium such as RAM (Random Access Memory).

[0033] Next, in step S101, the adjustment unit 11 performs a decoding process. After transmitting the video, the adjustment unit 11 performs a decoding process. Alternatively, the adjustment unit 11 may not need to perform decoding if the data does not require decoding, such as uncompressed data. In other words, by minimizing the decoding process on the receiving side of the transmission, it is possible to reduce the decoding load on the analysis server 1.

[0034] Next, in step S102, the adjustment unit 11 determines whether to perform initial analysis. In this embodiment, analyzing all high-quality image data without fail can unnecessarily increase the processing load and processing time. Therefore, the adjustment unit 11 first generates an image for initial analysis and determines whether to perform initial analysis. Specifically, for example, if the user has set initial analysis to be performed, the adjustment unit 11 may determine "Yes" to perform initial analysis using the image. Alternatively, the adjustment unit 11 may determine "Yes" to perform initial analysis by performing element analysis using heuristics on a portion of the high-quality image data when the quality is sufficient or the complexity is high enough to meet a certain threshold, and high-quality analysis will take time. In other cases, for example, if the adjustment unit 11 is set to calculate high-precision analysis results from the beginning, or if the element analysis results are below a certain threshold and it is determined that initial analysis is not necessary, the adjustment unit 11 may determine "No" to perform initial analysis. Alternatively, if the processing capacity of the analysis unit is such that high-quality image data can be processed as is without initial analysis, the adjustment unit 11 may determine "No." If the answer is Yes, the adjustment unit 11 advances the process to step S103. If the answer is No, the adjustment unit 11 advances the process to step S105.

[0035] If an initial analysis is to be performed, the analysis unit performs processing to generate an image for initial analysis in step S103. The analysis unit writes out a frame (2D image) of image data with a standard resolution and color gamut (hereinafter referred to as "standard image data"), which may have a lower resolution and a narrower color gamut than the high-quality image data, from the transmitted high-quality image data as image data for initial analysis. This standard image data may be, for example, image data of 1920 x 1080 pixels (1080p), sRGB 8-bit SDR, etc.

[0036] Next, in step S104, the analysis unit performs an initial analysis process. First, as an initial analysis, the analysis unit detects areas of blocked up shadows, blown out highlights, and other color saturation in the image data for initial analysis. The analysis unit executes various image recognition models on the remaining areas of the image data for initial analysis to search for the item. In this way, if the target item is recognized, the item search process may be ended. In other words, the high-precision analysis described later may be performed only in a necessary area when necessary.

[0037] Here, in step S105, the adjustment unit 11 determines whether high-precision analysis and adjustment are necessary. If the setting is to perform high-precision analysis, the adjustment unit 11 determines Yes. At this time, the adjustment unit 11 performs the above-mentioned element analysis and determines Yes as adjustment is necessary if the accuracy of the analysis results of high-quality image data may be reduced, specifically, if there is a possibility that the object detection accuracy may be reduced, if there is black crush, white blowout, or out-of-focus, or if the image is highly complex and it is necessary to grasp the three-dimensional position. In other cases, i.e., if the setting is not to perform high-precision analysis and the target item is recognized in the initial analysis as described above, the adjustment unit 11 determines No. If the answer is Yes, the adjustment unit 11 advances the process to step S106. If the answer is No, the adjustment unit 11 ends the item search process.

[0038] If high-precision analysis and adjustment are required, the adjustment unit 11 performs adjustment processing in step S106. The adjustment process will be described in detail with reference to FIGS. As described above, if the element analysis detects that the accuracy of the analysis result of high-quality image data may be low, and / or if black crush, white blowout, out-of-focus, etc. are detected, the adjustment unit 11 adjusts the image in accordance with the respective detected results. The adjustment unit 11 may perform any one or any combination of brightness adjustment, color gamut adjustment, and focus adjustment. When the setting is for performing high-precision analysis, the adjustment unit 11 adjusts the received high-quality image data. Alternatively, the adjustment unit 11 writes out image data of higher quality than that of the initial analysis, such as 4K BT.2020 10-bit HDR or 14-bit RAW, as the set image format, and adjusts this.

[0039] First, image adjustment when crushed blacks or blown out highlights are detected will be described. 3(a) shows an example of high-quality image data before adjustment. From this high-quality image data, the adjustment unit 11 detects areas detected in the initial analysis results, or areas that are almost flat black or 100% white directly from the high-quality image data, as areas of crushed blacks, blown-out highlights, or other color saturation. For the detected areas, the adjustment unit 11 acquires wide color gamut data, wide gradation data, or HDR data for the same areas, and adjusts brightness, hue, contrast, etc. as appropriate, and writes out one or several corrected image data using the adjustment value that achieves the greatest contrast, for example. FIG. 3(b) shows an example of corrected image data that has been written out after adjustment. The adjustment unit 11 performs a high-precision analysis, as will be described later, on the corrected image data or on the entire image obtained by combining the corrected images of the detected regions.

[0040] In addition, when the accuracy of the analysis results may be low, that is, for example, when there is a possibility that the object detection accuracy may be low or when the object is highly complex and it is necessary to grasp the three-dimensional position, the adjustment unit 11 optimizes the resolution, three-dimensional image, and focus of the analysis image. Specifically, the adjustment unit 11 refers to the 3D stereoscopic data or point cloud data to identify the three-dimensional position of the image. If an area that appears flat in the 2D image is located at different coordinates in the three-dimensional space, the adjustment unit 11 can recognize it as a different object, and transmits this three-dimensionally classified area to the analysis unit for direct recognition. Alternatively, the adjustment unit 11 may divide the area that can be recognized as a different object into separate image data and transmit each piece of divided image data to the analysis unit described below. In other words, it is possible to classify and divide images of objects located at different three-dimensional positions in advance based on the 3D stereoscopic video or point cloud data, and then analyze them. Furthermore, when multifocal data is available, for focus adjustment, for example, the image with the highest contrast and focus is written out from the multifocal data for areas with low detection accuracy and sent to the analysis unit. These adjustments make it possible to improve the accuracy of object detection.

[0041] FIG. 4 shows an example of depth map data in which three-dimensional positions are classified. In this example, an object is recognized and classified into several levels, with the depth varying depending on the density. From the foreground, the floor, the playground equipment with the cat inside, the cat, the visible backside of the playground equipment, and the background are identified and classified. Specifically, the adjustment unit 11 can classify by detecting a sudden change in depth. For example, the playground equipment can obtain depth data at a relatively close distance, but the background suddenly becomes depth data at a far distance, which the adjustment unit 11 can distinguish.

[0042] Next, in step S107, the image analysis unit 12 performs high-precision analysis processing. The image analysis unit 12 performs high-precision image analysis from the high-quality image data adjusted by the adjustment unit 11. The image analysis unit 12 searches for the target item through high-precision image analysis. The administrator of the analysis server 1 can set instructions such as parameters for this search using the input unit 13 and display unit 14. Alternatively, the image analysis unit 12 can analyze high-quality images with high resolution, wide color gamut, wide gradation, 3D images, etc. as they are, as long as the processing is possible under the calculation load. If the image analysis unit 12 is able to find the item, it may transmit the information to the terminal 2 and indicate the location of the item to the user via an intercom or the like. This response to the user may be made by an administrator or the like using the input unit 13 and the display unit 14. This completes the item search process according to this embodiment.

[0043] The above configuration can provide the following effects. Conventionally, solutions that rely on analyzing image data collected during work using wearable cameras attached to workers have had the following problems: First, it was practically difficult to configure a wearable camera to record image data on a recording medium such as an SD card built into the camera and then transfer the image data to an analysis server along with the recording medium. This was because the recording medium meant that video analysis could not be performed while the worker was working, and real-time analysis was not possible. Furthermore, this configuration required the work of transferring the data from the recording medium to the analysis server for analysis, which wasted work time. In other words, this configuration made it impossible to provide real-time feedback to the worker on the analysis server's analysis results.

[0044] However, due to the capabilities of the devices, it was difficult for the workers to perform the analysis themselves. In other words, since workers move around while working, if they tried to perform the analysis process on their own devices such as smartphones or PCs, the power would be supplied from the device's battery. Image analysis consumes a lot of power, which shortens the processing time, making it impractical.

[0045] On the other hand, when transmitting video from a wearable camera via Wi-Fi or other communication standards, there are limitations on the bit rate of the video that can be transmitted, as well as the type and volume of information that can be transmitted. For example, when attempting to transmit image data via standard 2.4 GHz Wi-Fi, the image data must be encoded at a low bit rate in a format such as 2K H.264 8-bit RGB and sent to the analysis server. Image data transmitted in this manner is blurry and unable to transmit sufficient additional information. As a result, the received image data is not of sufficient image quality for image analysis, hindering analysis accuracy. The same problem exists when using conventional variable bit rate technology.

[0046] However, considering the intended use, connecting to the analysis server via a high-speed wired network was also difficult. This was because the imaging equipment, especially the cables, would impede work. In addition, the cables could lead to incorrect recognition during image data analysis, reducing the accuracy of the object recognition itself. For these reasons, there was a need for an information system that could perform high-precision image analysis on an analysis server and search for items.

[0047] In contrast, (1) an information system X according to an embodiment of the present invention is an information system comprising a terminal 2 that transmits image data via network communication and an analysis server 1 that performs image analysis of the image data transmitted from the terminal 2, wherein the terminal 2 comprises a camera 21 that captures high-quality image data and a transmission unit 20 that transmits the high-quality image data captured by the camera 21 in real time while maintaining the high quality, and the analysis server 1 comprises a receiving unit 10 that receives the high-quality image data transmitted from the terminal 2, an adjustment unit 11 that detects in the high-quality image data received by the receiving unit 10 characteristics that may reduce the analytical accuracy, or detects any of the characteristics of black crush, white blowout, and out-of-focus, and adjusts the high-quality image data according to the detected characteristics, and an image analysis unit 12 that searches for the target item from the high-quality image data adjusted by the adjustment unit 11 through high-precision image analysis.

[0048] With this configuration, the user's camera 21 captures image data with higher bit rate and resolution than conventional systems, and transmits the image data via network communication. High-resolution, high-bit rate, and information-rich image data is transmitted in real time at high speed using a high-speed wireless connection such as 5G, 6G, or high-speed Wi-Fi. The transmitted high-quality, information-rich image data is then analyzed with high precision while adjusting the image appropriately in real time by the analysis server 1. This improves the accuracy of image analysis and makes it possible to search for the target item. As a result, the item can be searched reliably, improving work efficiency. By processing high-quality image data, even cables and other items that are difficult to recognize using low-quality image data can be reliably recognized.

[0049] In addition, in the past, when a user captured image data using a camera, the user did not have specialized camera photography skills, or the image was captured using the camera's built-in auto-exposure and auto-focus functions, which often resulted in overexposure, underexposure, and out-of-focus images, making the image data unclear and hindering analysis.

[0050] In contrast, the information system X according to this embodiment detects any one of the characteristics of black crush, white blown out, or out of focus in the analysis server 1, and adjusts the high-quality image data according to the detected characteristic. This enables highly accurate recognition to be performed reliably, and the target item can be found more reliably.

[0051] Furthermore, when attempting to encode image data using a conventional device 2 such as a wearable camera or smartphone, not only does the image quality deteriorate, i.e., it is no longer of high quality, but the image encoding load is large and consumes a lot of battery power on device 2. In other words, the image encoding process load is large and the battery quickly runs out.

[0052] In contrast, the information system X according to this embodiment minimizes the encoding process on the terminal 2 side for high-quality image data, thereby reducing the processing load and reducing and improving the power consumption of the user terminal 2. This reduces the battery consumption of the terminal 2, making it possible to practically use it for searching for items in factories, warehouses, etc.

[0053] Furthermore, (2) the high-quality image data according to an embodiment of the present invention is characterized in that the information system described in (1) includes any one or any combination of high-definition data, low-compression data, uncompressed data, 3D stereoscopic data, point cloud data, HDR data, high color gamut data, wide gradation data, and multi-focal data, and may be a still image or a moving image.

[0054] With this configuration, the amount of video encoding performed on terminal 2 is reduced, and low-compression or uncompressed "RAW" data, data with a large color band such as high-resolution (HDR) or wide color gamut data, 3D stereoscopic data, point cloud data, and other stereoscopic image data are transmitted, and analysis server 1 performs image analysis processing as well as optimizing the resolution, color gamut, brightness, contrast, focus, stereoscopic viewpoint, analysis target range, etc. of the analyzed image. This allows for more reliable and accurate image recognition, making it possible to more reliably find the desired item. In addition, especially when transmitting "RAW" data, the encoding processing load on terminal 2 is reduced, which also reduces battery consumption on terminal 2.

[0055] Furthermore, (3) in the information system X according to the embodiment of the present invention, the adjustment unit 11 is characterized in that it is an information system described in (1) or (2) that optimizes any one or any combination of the resolution or color gamut, brightness, contrast, focus, stereoscopic viewpoint, and analysis target range of the analysis image.

[0056] This configuration allows for the capture and transmission of image data rich in color and brightness information, such as HDR (high dynamic range) and 14-bit RAW. This allows for the acquisition of information on ultra-high brightness, ultra-low brightness, and rich gradation in HDR and RAW, and the prevention and correction of blown-out highlights, crushed shadows, and other color saturations. Furthermore, 3D data such as 3D stereoscopic data, point cloud data, and stereoscopic images can be transmitted and analyzed, and image adjustments can be made using multi-focal image data. As a result, the adjusted images enable highly accurate image recognition. This allows for reliable search for the target item.

[0057] Furthermore, (4) the information system X according to an embodiment of the present invention is characterized in that the image analysis unit 12 is an information system described in any one of (1) to (3), in which the image analysis unit 12 divides images of objects located at different three-dimensional positions based on 3D stereoscopic video data or point cloud data, and then analyzes the images.

[0058] This configuration makes it possible to analyze images of objects in different three-dimensional positions after dividing them into separate images based on 3D images and point clouds. In other words, by dividing the image data into rough depth classifications, different objects can be analyzed as separate images in advance. This eliminates the need to confuse objects that are similar in color and difficult to distinguish in 2D images, such as a black cat and the darkened backside of playground equipment, and improves the accuracy of object analysis, as described below. In other words, it becomes less likely to misidentify objects, making it possible to more reliably search for the desired object.

[0059] Other Embodiments In the above embodiment, an example in which standard image data is written from high-quality image data has been described. However, standard image data may be captured almost simultaneously with the high-quality image data by the camera 21. In this case, the transmission unit 20 may transmit the standard image data almost simultaneously with the high-quality image data in real time.

[0060] This configuration eliminates the need to write standard image data from high-quality image data for initial analysis, thereby reducing the load on the analysis server 1. In addition, the analysis server 1 does not need to worry about image quality degradation when generating standard image data from high-quality image data.

[0061] In the above embodiment, an example has been described in which the initial analysis itself is performed using standard image data. However, as in the case where step S102 in Fig. 2 above is No, this initial analysis itself does not have to be performed. Specifically, if the analysis unit has the computing power to directly analyze high-quality image data, or if element analysis of image data is not performed, the adjustment unit 11 does not have to perform the initial analysis. In this case, the adjustment unit 11 may send the high-quality image data directly to the analysis unit.

[0062] This configuration allows high-quality image data adjustment and analysis to be performed without performing initial analysis, thereby reducing the effort required for initial analysis and, depending on the configuration, allowing image data analysis to be performed at higher speed and with lower load.

[0063] In the above embodiment, the adjustment unit 11 and the image analysis unit 12 are described as separate functional configurations. However, these may be configured as an integrated unit, or may be configured to execute DL NN as an integrated model.Furthermore, a configuration may be adopted in which filters other than those described in the above embodiments are used, or image adjustment is performed.

[0064] This configuration makes it possible to perform highly accurate recognition of objects that are more suited to the environment, and depending on the configuration, it may also be possible to reduce the computational load.

[0065] In the above embodiment, an example has been described in which the information system X is used to search for items such as parts and tools inside a factory or warehouse. However, the object to be searched does not have to be a part or a tool. For example, in information system X, image analysis may be used as a monitoring system to monitor suspicious objects, animals, etc. Alternatively, the information system X can be used to search for products, shipping containers, etc. in a factory or warehouse. Even in such a case, the transmitted high-quality, multi-information image data can be analyzed by the analysis server 1 while adjusting the image as needed.

[0066] This configuration makes it possible to apply the system to solutions that involve analyzing image data during various tasks.

[0067] In the above embodiment, an example has been described in which the input unit 13 and the display unit 14 are connected to the analysis server 1, and instructions and settings for analysis are given. However, the analysis server 1 does not have to be equipped with the input unit 13 and the display unit 14. In this case, it may be possible to give instructions for setting up the analysis server 1 and performing analysis using a touch panel and display of a smartphone, MR or VR headset, or the like of the terminal 2. In this case, it may be possible to give instructions to the analysis server 1 using dedicated application software (app). Furthermore, although the above-described embodiment describes the user being notified of the searched-for item via an intercom or the like, in the case of an MR or VR headset, it may also be possible to indicate the location of the desired item by overlaying it on the actual scenery.

[0068] This configuration improves user convenience and allows for a wide range of configurations to be adapted to suit a variety of applications.

[0069] It goes without saying that the configurations and operations of the above-described embodiments are merely examples, and can be modified as appropriate within the scope of the present invention. [Explanation of symbols]

[0070] 1. Analysis Server 2. Terminal 10 Receiving unit 11 Adjustment part 12 Image analysis section 13 Input section 14 Display section 20 Transmitter 21 Camera X Information Systems

Claims

1. An information system comprising a terminal that transmits image data via network communication and an analysis server that performs image analysis on the image data transmitted from the terminal, The terminal a camera that captures high-quality image data; a transmitting unit that transmits the high-quality image data captured by the camera in real time while maintaining the high quality, The analysis server a receiving unit that receives the high-quality image data transmitted from the terminal; an adjustment unit that detects that the high-quality image data received by the receiving unit has a characteristic that may reduce analysis accuracy, or detects any of the characteristics of black crush, white blowout, and out-of-focus, and adjusts the high-quality image data according to the detected characteristic; and an image analysis unit that searches for a target item by performing high-precision image analysis on the high-quality image data adjusted by the adjustment unit. An information system characterized by:

2. The high-quality image data includes: Includes any one or any combination of high definition data, low compression data, uncompressed data, 3D stereoscopic data, point cloud data, HDR data, high color gamut data, wide gradation data, and multi-focal data.

2. The information system according to claim 1.

3. The adjustment unit Optimize the resolution or color gamut, brightness, contrast, focus, stereoscopic viewpoint, and analysis target range of the analyzed image, or any combination thereof.

2. The information system according to claim 1.

4. The image analysis unit Based on the 3D stereoscopic data or the point cloud data, images of objects located at different positions in three dimensions are divided and then analyzed.

3. The information system according to claim 2.

Images