Artificial intelligence processor and artificial intelligence analysis device

Abstract

The application discloses an artificial intelligence processor and an artificial intelligence analysis device. The artificial intelligence processor 10 comprises a control module 100, a USB receiving module 200 and an artificial intelligence processing module 300. The control module 100 is connected with the USB receiving module 200 and the artificial intelligence processing module 300 respectively and is used for controlling the USB receiving module 200 and the artificial intelligence processing module 300. The USB receiving module 200 is used for receiving image information to be processed. The artificial intelligence processing module 300 is used for intelligently analyzing a target object in the image information to be processed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to an artificial intelligence processor and an artificial intelligence analysis device. BACKGROUND

[0002] With the development of science and technology, in order to meet people's living needs, more and more monitoring devices are used, and a large number of videos are also generated, so the analysis of the videos is particularly important. Therefore, artificial intelligence analysis technology based on artificial intelligence such as artificial intelligence analyzers has also developed a lot. Among them Figure 4 An existing intelligent analyzer is shown. Referring to Figure 4 The artificial intelligence analyzer includes an image acquisition module, a preprocessing module, and an artificial intelligence processing module. The image acquisition module is used to acquire images in real time and transmit the images to the preprocessing module for preprocessing (such as filtering, denoising, etc.), and then the preprocessed images are input to the artificial intelligence processing module for artificial intelligence analysis.

[0003] However, the traditional artificial intelligence analyzer has the following disadvantages due to the use of an embedded image acquisition module: 1) the image clarity is fixed and cannot be upgraded as needed, and the clarity inevitably leads to high prices, and the price-friendly clarity is low; 2) the image acquisition module is not easy to replace, and the image acquisition device is a consumable device that needs to be replaced by professionals when it is broken; 3) it can only be used in fixed scenes, for example, it can be used in scenes with good light conditions, but it cannot be used in low-light environments; it can be used in indoor scenes, but it cannot be used in outdoor harsh environments; 4) the number of cameras is fixed and cannot be increased, the present application provides an analysis device with a USB interface, which can increase the number of cameras according to user needs; 5) most of the algorithm modules of the existing USB artificial intelligence analysis device are implemented by software, and the present application provides an artificial intelligence processor based on a chip for low-power high-speed processing.

[0004] At present, there is no effective solution to the technical problems of the traditional artificial intelligence analyzer, such as fixed image acquisition module clarity, not easy to replace, low software implementation operation speed, high power consumption, fixed number of cameras that cannot be increased, and only being able to be used in fixed scenes. SUMMARY

[0005] The present application provides an artificial intelligence processor and an artificial intelligence analysis device to at least solve the technical problems of the traditional artificial intelligence analyzer, such as fixed image acquisition module clarity, not easy to replace, low software implementation operation speed, high power consumption, fixed number of cameras that cannot be increased, and only being able to be used in fixed scenes.

[0006] According to an aspect of the present application, there is provided an artificial intelligence processor, comprising: a control module, a USB receiving module, and an artificial intelligence processing module, wherein the control module is connected with the USB receiving module and the artificial intelligence processing module respectively, and is configured to control the USB receiving module and the artificial intelligence processing module; the USB receiving module is configured to receive image information to be processed; and the artificial intelligence processing module is configured to intelligently analyze a target object in the image information to be processed.

[0007] Optionally, the USB receiving module comprises a USB controller unit, and the USB controller unit is connected with the USB interface respectively, and is configured to receive the image information to be processed from the USB interface.

[0008] Optionally, the artificial intelligence processing module comprises a deep learning processor unit, and the deep learning processor unit is configured to intelligently analyze the target object in the plurality of image information to be processed, and generate an intelligent analysis result of the target object.

[0009] Optionally, the control module is configured to determine a network model used for the intelligent analysis of the target object in the image information to be processed and a number of the deep learning processor units according to a user configuration requirement.

[0010] Optionally, the artificial intelligence processing module comprises a plurality of deep learning processor units, and the plurality of deep learning processor units are configured to intelligently analyze the target object in the plurality of image information to be processed, and generate an intelligent analysis result of the target object; and the control module is further configured to determine the number of the deep learning processor units used for the intelligent analysis of the target object in the image information to be processed according to a number of the USB controller units receiving the image information to be processed.

[0011] Optionally, the artificial intelligence processor further comprises a preprocessing module, wherein the preprocessing module is connected with the control module, and is configured to pre-process a first image in the image information to be processed, and generate a second image suitable for the intelligent analysis of the artificial intelligence processing module.

[0012] Optionally, the preprocessing module comprises: a format conversion unit configured to convert a first image format of the first image into a second image format suitable for the intelligent analysis of the artificial intelligence processing module; a resolution conversion unit configured to convert a first image resolution of the first image into a second image resolution suitable for the intelligent analysis of the artificial intelligence processing module; and an image enhancement unit configured to suppress noise in the first image, and enhance detailed information in the first image.

[0013] Optionally, in the case that the first image resolution is higher than the second image resolution suitable for the intelligent analysis of the artificial intelligence processing module, the resolution conversion unit is configured to convert the first image resolution of the first image into the second image resolution suitable for the intelligent analysis of the artificial intelligence processing module by any one of the following operations: downsampling the first image to convert the first image resolution of the first image into the second image resolution; downsampling the first image and deforming the first image to convert the first image resolution of the first image into the second image resolution; cropping the first image and downsampling the first image to convert the first image resolution of the first image into the second image resolution; and segmenting the first image to convert the first image resolution of the first image into a plurality of second image resolutions.

[0014] Optionally, in the case that the first image resolution is lower than the second image resolution suitable for the intelligent analysis of the artificial intelligence processing module, the resolution conversion unit is configured to convert the first image resolution of the first image into the second image resolution suitable for the intelligent analysis of the artificial intelligence processing module by any one of the following operations: upsampling the first image to convert the first image resolution of the first image into the second image resolution; upsampling the first image and deforming the first image to convert the first image resolution of the first image into the second image resolution; and upsampling the first image and cropping the first image to convert the first image resolution of the first image into the second image resolution.

[0015] Optionally, the artificial intelligence processor further comprises an output module, wherein the output module is connected with the control module and is configured to output the image information data processed by the artificial intelligence processing module.

[0016] Optionally, the output module comprises an HDMI / DP output unit configured to be connected with an HDMI / DP output interface of the external device and transmit the processed image information data through the HDMI / DP output interface, and / or a network access unit configured to be connected with a network access interface of the external device and transmit the processed image information data through the network.

[0017] Optionally, the artificial intelligence processor further comprises a preprocessing module, an output module and a memory module, wherein the preprocessing module is connected with the control module and is configured to preprocess the first image in the to-be-processed video to generate the second image suitable for the intelligent analysis of the artificial intelligence processing module; the output module is connected with the control module and is configured to output the image information data processed by the artificial intelligence processing module; and the memory module is connected with the USB receiving module, the artificial intelligence processing module, the preprocessing module and the output module, and is configured to be connected with a memory of the external device.

[0018] According to another aspect of the present application, there is provided an artificial intelligence analysis device, comprising: a USB interface configured to receive image information to be processed; and an artificial intelligence processor, wherein the artificial intelligence processor is connected with the USB interface, and the artificial intelligence processor comprises: a control module, a USB receiving module and an artificial intelligence processing module, wherein the control module is connected with the USB receiving module and the artificial intelligence processing module respectively, and is configured to control the USB receiving module and the artificial intelligence processing module; the USB receiving module is configured to receive the image information to be processed; and the artificial intelligence processing module is configured to intelligently analyze a target object in the image information to be processed.

[0019] Optionally, the artificial intelligence analysis device further comprises: an HDMI / DP output interface, the HDMI / DP output interface being configured to transmit the processed image information data, and the artificial intelligence processor further comprises: an output module, wherein the output module is connected with the control module and the HDMI / DP output interface, and is configured to output the image information data processed by the artificial intelligence processing module.

[0020] Optionally, the artificial intelligence analysis device further comprises: a network access interface, the network access interface being configured to transmit the processed image information data, and the artificial intelligence processor further comprises: an output module, wherein the output module is connected with the control module and the network access interface, and is configured to output the image information data processed by the artificial intelligence processing module.

[0021] According to the technical solution of the present embodiment, the image acquisition module can be separated from the artificial intelligence analysis device, so that image acquisition modules with different resolutions can be changed as needed, and different image acquisition modules can also be replaced under different lighting conditions, temperature conditions and resolution conditions. Thus, the technical problems of the conventional artificial intelligence analysis device, such as fixed resolution of the image acquisition module, difficulty in replacement, low software implementation operation speed, high power consumption, fixed number of cameras, inability to increase the number of cameras and being only used in fixed scenes, are solved. BRIEF DESCRIPTION OF DRAWINGS

[0022] Some specific embodiments of the present application will be described in detail below with reference to the accompanying drawings, which are exemplary and non-limiting. The same reference numerals in the drawings denote the same or similar components or parts. Those skilled in the art should understand that the drawings are not necessarily drawn to scale. In the drawings:

[0023] Figure 1 is a schematic diagram of an artificial intelligence processor according to a first aspect of an embodiment of the present application;

[0024] Figure 2 is a schematic diagram of an artificial intelligence analysis device according to a second aspect of an embodiment of the present application;

[0025] Figure 3 is a schematic diagram of an artificial intelligence analysis connected with a USB camera according to a first aspect of the embodiments of the present application; and

[0026] Figure 4 is a schematic diagram of an intelligent analyzer existing in the prior art. DETAILED DESCRIPTION

[0027] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0028] In order to enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should fall within the scope of protection of the present application.

[0029] It should be noted that the terms "first", "second", and the like in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or a chronological sequence. It should be understood that the terms thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but can include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0030] It should be noted that the terms used herein are only for the purpose of describing specific embodiments, and are not intended to limit the exemplary embodiments according to the present application. As used herein, the singular form is intended to include the plural form unless the context clearly indicates otherwise, and it should also be understood that when the terms "comprise" and / or "include" are used in the specification, they indicate the presence of a feature, step, operation, device, component and / or combination thereof.

[0031] Reference Figure 1 and Figure 2As shown, according to the first aspect of the present application, an artificial intelligence analysis device 30 is provided, comprising: a USB interface 31 configured to receive image information to be processed; and an artificial intelligence processor 10, wherein the artificial intelligence processor 10 is connected with the USB interface 31, and the artificial intelligence processor 10 comprises: a control module 100, a USB receiving module 200 and an artificial intelligence processing module 300, wherein the control module 100 is connected with the USB receiving module 200 and the artificial intelligence processing module 300 respectively, and is used to control the USB receiving module 200 and the artificial intelligence processing module 300; the USB receiving module 200 is used to receive the image information to be processed; and the artificial intelligence processing module 300 is used to intelligently analyze a target object in the image information to be processed.

[0032] The artificial intelligence processor 10 may, for example, be a processor implemented based on FPGA. Specifically, it may, for example, be a processor implemented by using Zynq UltraScale+MPSoC of XILINX or Stratix 10 of Intel.

[0033] As described in the background, the conventional artificial intelligence analysis device has the following disadvantages due to the use of an embedded image acquisition module: 1) the image definition is fixed, and the high definition inevitably leads to high price, and the price-friendly definition is low; 2) the image acquisition module is not easy to replace, and the image acquisition device is a consumable device, which needs to be replaced by professionals after being damaged; 3) it can only be used in fixed scenes, for example, it can be used in scenes with good light conditions, but it cannot be used in low-light environments; it can be used in indoor scenes, but it cannot be used in outdoor harsh environments; 4) the number of cameras is fixed and cannot be increased, the present application provides an analysis device with a USB interface, which can increase the number of cameras according to user needs; 5) most of the algorithm modules of the existing USB artificial intelligence analysis device are implemented by software, and the present application provides a low-power high-speed processing artificial intelligence processor based on a chip, and the like.

[0034] Therefore, according to the technical solution of the present embodiment, an artificial intelligence analysis device 30 is provided, comprising: a USB interface 31 and an artificial intelligence processor 10 connected with the USB interface 31. Thus, the image acquisition module is no longer embedded in the artificial intelligence analysis device, but can be connected with the artificial intelligence analysis device 30 through the USB interface 31. Referring to Figure 3 As shown, the USB camera 40 can be detachably connected with the artificial intelligence analysis device 30 through the USB interface 31.

[0035] The improved AI analyzer uses a USB interface and can receive uncompressed and compressed video streams of various formats in real time. Most devices currently use a USB 3.0 interface, which has a maximum transmission bandwidth of 5Gbps and can support video streams of 3840*2160@25fps. If sending 1920*1080 resolution video, it can support 4 channels of 1920*1080@25fps real-time streaming video. If sending 1280*720 resolution video, it can support 8 channels of 1280*720@25fps real-time streaming video.

[0036] In this way, the image acquisition module is separated from the artificial intelligence analysis device 30, so that the image acquisition module with different resolution can be changed as needed. In addition, different image acquisition modules can be replaced or added under different lighting conditions, temperature conditions, and resolution conditions.

[0037] This solves the technical problems of traditional artificial intelligence analyzers, such as fixed image acquisition module clarity, difficulty in replacement, slow software processing speed, high power consumption, fixed number of cameras that cannot be increased, and the inability to use them in fixed scenarios.

[0038] In addition, refer to Figure 2 As shown, the present invention also integrates the control module 100, USB receiving module 200, artificial intelligence processing module 300, preprocessing module 400, and output module 500 into a single artificial intelligence processor 10, thereby simplifying the structure of the artificial intelligence analysis device 30. See also Figure 2 As shown, the artificial intelligence analysis device 30 only needs to be configured with an input interface (e.g., USB interface 31), an artificial intelligence processor 10, and an output interface (e.g., HDMI / DP output interface 32 and network input interface 33), thereby reducing the cost and complexity of hardware design, thereby reducing power consumption and greatly improving operating speed.

[0039] Furthermore, the artificial intelligence processing module 300 is used to intelligently analyze target objects in the image information to be processed. For example, the artificial intelligence processing module 300 can intelligently analyze human bodies, faces, vehicles, and animals in the image information.

[0040] Thus, according to the technical solution of the present embodiment, the USB receiving module 200 of the artificial intelligence processor 10 can receive image data transmitted by the USB camera 40 in real time, and then intelligently analyze the target object in the video data or image data through the artificial intelligence processing module 300, and mark the analysis result of the target object in the image. Thus, the artificial intelligence processor 10 of the present application can directly perform artificial intelligence analysis on the video data or image data in the host through the USB interface 31. Moreover, the artificial intelligence device 10 can analyze the image data to be processed in real time, and through the connection with the display device of the external device, the analysis process can be viewed in real time, and the intelligent analysis result can be sent to the cloud server in real time through the Internet.

[0041] Optionally, the USB receiving module 200 includes USB controller units 201-20n, which are connected with the USB interface 31 and configured to receive the image information to be processed from the USB interface 31.

[0042] Specifically, as shown in Figure 1 and Figure 2 , the USB receiving module 200 includes USB controller units 201-20n, which are connected with the USB interface 31 (such as the USB interface 31 in the artificial intelligence analysis device 30), so that the USB receiving module 200 built-in the artificial intelligence processor 10 can be directly connected with the USB interface 31, without the need to additionally set a USB driving circuit. Thus, the hardware cost of the artificial intelligence analysis device is reduced.

[0043] In addition, as shown in Figure 2 , the USB receiving module 200 may, for example, include a plurality of USB controller units 201-20n, so that it can receive image information to be processed from different USB input interfaces 31. For example, if the USB receiving module 200 has four USB controller units 201-204, the artificial intelligence processor 10 can receive four different image information to be processed. The four different image information to be processed may, for example, contain different target objects, such as image information containing vehicles, image information containing human bodies, and image information containing animals, etc. Thus, through the plurality of USB controller units 201-20n of the USB receiving module 200, different image information to be processed can be received, so that the artificial intelligence processor 10 can simultaneously intelligently analyze a plurality of target objects in a plurality of image information to be processed.

[0044] Optionally, the artificial intelligence processing module 300 includes multiple deep learning processor units 301 to 30n, wherein the multiple deep learning processor units 301 to 30n are configured to perform intelligent analysis on target objects in multiple image information to be processed, and generate intelligent analysis results of the target objects.

[0045] Specifically, refer to Figure 2 As shown, the artificial intelligence processing module 300 may include, for example, deep learning processor units 301-30n. These deep learning processor units 301-30n are configured to perform intelligent analysis on target objects in the image information to be processed and generate intelligent analysis results for the target objects. For example, the position information of the target objects in the image can be marked using rectangular boxes. Thus, intelligent detection of target objects in the image can be achieved through the deep learning processor units 301-30n.

[0046] In addition, further reference Figure 2 As shown, the artificial intelligence processing module 300 may include multiple deep learning processing units 301 to 30n, thereby enabling intelligent analysis of multiple target objects in parallel.

[0047] Furthermore, the AI ​​processor 10 can process multiple model applications through deep learning processor units. One network model used for intelligent target object analysis can intelligently analyze multiple target objects (e.g., simultaneously detecting humans and vehicles). Moreover, when the target object network model is complex and requires significant computing power, the model can be computed using the combined computing power of multiple deep learning processors.

[0048] Optionally, the control module 100 is configured to determine the number of network models and deep learning processor units used for intelligent analysis of target objects in the image information to be processed, according to user configuration requirements.

[0049] Specifically, refer to Figure 2 As shown, the control module 100 can, for example, determine the number of network models and deep learning processor units used for intelligent analysis of target objects in the image information to be processed, based on the user's configuration requirements. For instance, if the user needs to intelligently analyze a person and a dog in the image information to be processed, the control module 100 can determine the network model used for intelligent analysis of the person and the dog. Thus, the control module 100 can determine the number of deep learning processing units used according to the user's needs, thereby completing the intelligent analysis of the target object.

[0050] Optionally, the artificial intelligence processing module 300 includes multiple deep learning processor units 301-30n, wherein the multiple deep learning processor units 301-30n are configured to perform intelligent analysis on target objects in multiple image information to be processed, and generate intelligent analysis results of the target objects in the images; and the control module 100 is further configured to determine the number of deep learning processor units used for intelligent analysis of target objects in the image information to be processed based on the number of USB controller units receiving the image information to be processed.

[0051] Specifically, refer to Figure 2 As shown, the control module 100 can also be used to determine the number of deep learning processor units used for intelligent target object analysis of the image information to be processed based on the number of USB controller units receiving the image information to be processed. For example, if four USB controller units receive and process four image information to be processed, the control module 100 can determine the number of deep learning processor units to be used based on the total computational cost required for intelligent target object analysis of the four image information received by the four USB controller units. Thus, by determining the number of deep learning processor units to be used based on the USB controller units, the intelligent analysis of the target object can be completed.

[0052] Optionally, the artificial intelligence processor 10 further includes a preprocessing module 400, wherein the preprocessing module 400 is connected to the control module 100 and configured to preprocess the first image in the image information to be processed to generate a second image suitable for intelligent analysis by the artificial intelligence processing module 300.

[0053] Specifically, refer to Figure 2 As shown, the artificial intelligence processor 10 also includes a preprocessing module 400, which is connected to the control module 100. The control module 100 selectively controls the preprocessing module 400, for example, determining the image format conversion unit 410 and resolution conversion unit 420 based on the format and resolution of the image information to be processed received from the USB interface unit. The preprocessing module 400 preprocesses the first image in the image information to be processed, generating a second image suitable for intelligent analysis by the artificial intelligence processing module 300. Thus, the preprocessing module 400 can convert the first image of the received image information to be processed into a second image suitable for intelligent analysis by the artificial intelligence processing module 300.

[0054] Optionally, the pre-processing module 400 comprises: an image format conversion unit 410 configured to convert a first image format of the first image into a second image format suitable for the intelligent analysis of the artificial intelligence processing module 300; a resolution conversion unit 420 configured to convert a first image resolution of the first image into a second image resolution suitable for the intelligent analysis of the artificial intelligence processing module 300; and an image enhancement unit 430 configured to suppress noise in the first image and enhance detailed information in the first image.

[0055] In particular, referring to FIG. 4, the pre-processing module 400 comprises: an image format conversion unit 410 configured to convert a first image format of the first image into a second image format suitable for the intelligent analysis of the artificial intelligence processing module 300. For example, the USB interface can receive data in YUV format (i.e. the first image format), while the artificial intelligence processing module 300 can only receive data in RGB format (i.e. the second image format), so the data in YUV format can be converted into data in RGB format suitable for the artificial intelligence processing module 300 to receive by the image format conversion unit 410. Figure 2 Further, the pre-processing module 400 further comprises: a resolution conversion unit 420 configured to convert a first image resolution of the first image into a second image resolution suitable for the intelligent analysis of the artificial intelligence processing module 300. For example, the first image resolution of the first image is 1920*1080, while the second image resolution suitable for the artificial intelligence processing module 300 is 1024*768, so the first image resolution 1920*1080 of the first image can be converted into the second image resolution 1024*768 suitable for the intelligent analysis of the artificial intelligence processing module 300 by the resolution conversion unit 420.

[0056] Further, the pre-processing module 400 further comprises: an image enhancement unit 430 configured to suppress noise in the first image and enhance detailed information in the first image. For example, in the case that the target object and other information in the first image are unclear, the details in the first image need to be enhanced (for example, in the case that the target object is a human body, the edge information of the human body is enhanced), so that the target object in the first image becomes clear.

[0057] Therefore, the first image received by the USB receiving module 200 is pre-processed by the image format conversion unit 410, the resolution conversion unit 420 and the image enhancement unit 430 in the pre-processing module 400, and a second image suitable for the intelligent analysis of the artificial intelligence processing module 300 is generated.

[0058]

[0059] ​Optionally, when the resolution of the first image is higher than the resolution of the second image suitable for intelligent analysis by the artificial intelligence processing module 300, the resolution conversion unit 420 is configured to convert the first image resolution of the first image into the resolution of the second image suitable for intelligent analysis by the artificial intelligence processing module 300 through any of the following operations: downsampling the first image to convert the first image resolution of the first image into the second image resolution; performing downsampling and deformation operations on the first image to convert the first image resolution of the first image into the second image resolution; performing cropping and downsampling operations on the first image to convert the first image resolution of the first image into the second image resolution; and segmenting the first image to convert the first image resolution of the first image into multiple second image resolutions.

[0060] Specifically, refer to Figure 2 As shown, the resolution conversion unit 420 is configured to convert the first image resolution (e.g., 1920*1080) of the first image into a second image resolution (e.g., 512*512, 640*360, or 640*480) suitable for intelligent analysis by the artificial intelligence processing module 300 through any of the following operations:

[0061] 1. Downsample the first image. For example, if the resolution of the received first image is 1920*1080, while the resolution of the image suitable for intelligent analysis by the artificial intelligence processing module 300 is 640*360, the first image can be converted into an image with a resolution of 640*360 through a downsampling operation.

[0062] 2. The first image is downsampled and deformed according to the aspect ratio suitable for intelligent analysis and recognition by the artificial intelligence processing module 300 (the deformation operation can be, for example, stretching or compressing the image). For example, if the resolution of the received first image is 1920*1080, and the resolution of the image suitable for intelligent analysis by the image artificial intelligence processing module 300 is 640*480, the first image can be converted into an image with a resolution of 640*480 through downsampling and stretching / compression operations.

[0063] 3. Based on the aspect ratio of the image suitable for intelligent analysis by the AI ​​processing module 300, the first image is cropped, and then the cropped image is downsampled to obtain an image suitable for intelligent analysis by the AI ​​processing module 300. For example, if the received first image is 1920*1080, and the resolution of the image suitable for intelligent analysis by the AI ​​processing module 300 is 512*512, the first image can be cropped to a resolution of 1080*1080, and then the cropped image can be downsampled to obtain an image with a resolution of 512*512.

[0064] 4. The received first image is segmented into multiple sub-images suitable for intelligent analysis by the artificial intelligence processing module 300. For example, if the received first image is 1920*1080, and the resolution of the image suitable for intelligent analysis by the artificial intelligence processing module 300 is 512*512, the resolution conversion unit 420 can segment the 1920*1080 image into multiple sub-images with a resolution of 512*512. This allows for intelligent image analysis operations to be performed on each of the segmented sub-images.

[0065] Thus, through the above methods, the artificial intelligence processor 10 can convert the first image resolution of the first image into a second image resolution suitable for processing by the artificial intelligence processing module 300 by using the resolution conversion unit 420 in the preprocessing module 400.

[0066] Optionally, if the resolution of the first image is lower than the resolution of the second image suitable for intelligent analysis by the artificial intelligence processing module 300, the resolution conversion unit 420 is configured to convert the first image resolution of the first image into the second image resolution suitable for intelligent analysis by the artificial intelligence processing module 300 through any of the following operations: upsampling the first image to convert the first image resolution of the first image into the second image resolution; performing an upsampling operation and a deformation operation on the first image to convert the first image resolution of the first image into the second image resolution; and performing an upsampling operation and a cropping operation on the first image to convert the first image resolution of the first image into the second image resolution.

[0067] Specifically, refer to Figure 1 As shown, the resolution conversion unit 420 is configured to convert the first image resolution (e.g., 160*120 or 384*288) of the first image into a second image resolution (e.g., 512*512, 640*360 or 640*480) suitable for intelligent analysis by the artificial intelligence processing module 300 through any of the following operations:

[0068] 1. upsampling the first image. For example, when the resolution of the received first image is 160*120 and the resolution of the image suitable for intelligent analysis of the artificial intelligence processing module 300 is 640*480, the first image can be converted into an image with a resolution of 640*480 through upsampling operation.

[0069] 2. upsampling the first image and deforming the image according to the aspect ratio of the image suitable for intelligent analysis of the artificial intelligence processing module 300 (the deformation operation may, for example, be a stretching or compression operation on the image). For example, when the resolution of the received first image is 160*120 and the resolution of the image suitable for intelligent analysis of the artificial intelligence processing module 300 is 640*360, the first image can be converted into an image with a resolution of 640*360 through upsampling and stretching / compression operation.

[0070] 3. cropping the first image according to the aspect ratio of the image suitable for intelligent analysis of the artificial intelligence processing module 300, and then upsampling the cropped image to obtain an image suitable for intelligent analysis of the artificial intelligence processing module 300. For example, when the received first image is 384*288 and the resolution of the image suitable for intelligent analysis of the artificial intelligence processing module 300 is 512*512, the first image can be first cropped to a size of 288*288, and then the cropped image is upsampling processed to obtain an image with a resolution of 512*512.

[0071] Thus, through the above manner, the artificial intelligence processor 10 can convert the first image resolution of the first image into the second image resolution suitable for processing of the artificial intelligence processing module 300 through the resolution conversion unit 420 in the preprocessing module 400.

[0072] Optionally, the artificial intelligence analysis device 30 further comprises an HDMI / DP output interface 32, the HDMI / DP output interface 32 is configured to transmit the processed image information data, and the artificial intelligence processor 10 further comprises an output module 500, wherein the output module 500 is connected with the control module 100 and the HDMI / DP output interface 32, and is configured to output the image information data processed by the artificial intelligence processing module 300.

[0073] Optionally, the artificial intelligence analysis device 30 further comprises a network access interface 33, the network access interface 33 is configured to transmit the processed image information data through a network, and the artificial intelligence processor 10 further comprises an output module 500, wherein the output module 500 is connected with the control module 100 and the network access interface 33, and is configured to output the image information data processed by the artificial intelligence processing module 300.

[0074] Specifically, referring to Figure 2 and Figure 1 As shown, the artificial intelligence analysis device 30 includes an HDMI / DP output interface 32 and a network access interface 33. The HDMI / DP output interface 32 is connected to the HDMI / DP output unit 510 of the artificial intelligence processor 10 at one end, and the network access interface 33 is connected to the network access unit 520 of the artificial intelligence processor 10 at one end. The artificial intelligence processor 10 further includes an output module 500, wherein the output module 500 is connected to the control module 100, so as to complete the operation instruction of the output module 500 through the control module 100. And the output module 500 is used to output the result of intelligent analysis of the target object after the intelligent analysis of the artificial intelligence processing module 300, wherein the result data of intelligent analysis includes the position information of the labeled target object. For example, the output module 500 can output the processed image information data to a display device, so that the video stream image of artificial intelligence analysis can be viewed in real time. Further, the technical problem that the process of video analysis cannot be viewed in real time in the prior art is solved. In addition, the output module 500 can also send the processed image information data to a remote device through a network, and further analyze the processed image information data.

[0075] Optionally, the output module 500 includes: an HDMI / DP output unit 510, the HDMI / DP output unit 510 is configured to be connected with the HDMI / DP output interface 32 of the external device, and transmit the processed image information data through the HDMI / DP output interface 32; and / or a network access unit 520, the network access unit 520 is configured to be connected with the network access interface 33, and transmit the processed image information data through a network.

[0076] Specifically, referring to Figure 2 and Figure 2 As shown, the output module 500 includes an HDMI / DP output unit 510, which is used to be connected with the HDMI / DP output interface 32 of the external device, and transmit the processed image information data through the HDMI / DP output interface 32. For example, the HDMI / DP output unit 510 can transmit the processed image information data to an external display device through the HDMI / DP output interface 32, so that the video stream image of artificial intelligence analysis can be viewed in real time on the display device. Further, the technical problem that the process of video analysis cannot be viewed in real time in the prior art is solved.

[0077] Further, the output module 500 further comprises a network access unit 520 configured to be connected with the network access interface 33, for sending the processed image information data to a remote device. For example, the network access unit 520 can send the processed image information data to a remote server through the Internet, and then further analyze the image information data analyzed by the artificial intelligence.

[0078] In addition, the user can also remotely configure and manage the artificial intelligence analysis device 30 through the network access unit 520 by using the remote server.

[0079] Optionally, the artificial intelligence processor 10 further comprises a preprocessing module 400, an output module 500, and a memory module 600, wherein the preprocessing module 400 is connected with the control module 100 and is configured to preprocess the first image in the to-be-processed video to generate a second image suitable for intelligent analysis by the artificial intelligence processing module 300; the output module 500 is connected with the control module 100 and is configured to output the image information data processed by the artificial intelligence processing module 300; and the memory module 600 is connected with the USB receiving module 200, the artificial intelligence processing module 300, the preprocessing module 400, and the output module 500, and is configured to be connected with the memory 20 of the artificial intelligence analysis device 30.

[0080] Specifically, referring to FIG. 1, Figure 1 Optionally, the artificial intelligence processor 10 further comprises a preprocessing module 400 and an output module 500, wherein the preprocessing module 400 is connected with the control module 100 and is configured to preprocess the first image in the to-be-processed video to generate a second image suitable for intelligent analysis by the artificial intelligence processing module 300; and the output module 500 is connected with the control module 100 and is configured to output the image information data processed by the artificial intelligence processing module 300.

[0081] Further, the memory module 600 is connected with the USB receiving module 200, the artificial intelligence processing module 300, the preprocessing module 400, and the output module 500, and the memory module 600 is used to dynamically store the image data generated after the operation instruction of the control module 100 is completed by the USB receiving module 200, the artificial intelligence processing module 300, the preprocessing module 400, and the output module 500. The memory module 600 is connected with the memory 20 arranged in the artificial intelligence analysis device 30, wherein the memory module 600 can be a DRAM controller for example, and the memory of the peripheral device can be a DDR memory for example, so that the dynamically stored data is sent to the memory 20 of the peripheral device through the memory module 600, and then the storage of the data is completed.

[0082] Thus, through the embodiment, the image acquisition module is separated from the artificial intelligence analysis device 30, and thus can be replaced. Therefore, image acquisition modules with different resolutions can be changed as needed, and in addition, different image acquisition modules can be replaced under different lighting conditions, temperature conditions, and resolution conditions. Thus, the technical problems of the conventional artificial intelligence analysis instrument, such as fixed resolution of the image acquisition module, difficulty in replacement, and being only applicable to a specified scene, are solved.

[0083] In addition, referring to Figure 1 According to a second aspect of the embodiment, an artificial intelligence processor 10 is also provided. The artificial intelligence processor 10 can be a processor implemented based on an FPGA, for example. Specifically, it can be a processor implemented by using a Zynq UltraScale+MPSoC of XILINX or a Stratix 10 of Intel, for example.

[0084] The artificial intelligence processor 10 includes a control module 100, a USB receiving module 200, and an artificial intelligence processing module 300. The control module 100 can be a CPU processor module, for example. The control module 100 is connected to the USB receiving module 200 and the artificial intelligence processing module 300, respectively, and is configured to control the USB receiving module 200 and the artificial intelligence processing module 300. For example, the control module 100 can send a receiving instruction. After receiving the instruction, the USB receiving module 200 converts the information received by the USB data interface 31 into image information to be processed, and saves the image information to be processed in the memory 20 through the memory module 600 for subsequent use. For example, the control module 100 sends an intelligent analysis and detection operation instruction to the artificial intelligence processing module 300. The artificial intelligence processing module 300 reads the image information used for intelligent analysis from the memory 20 through the memory module 600, and then processes the image information through the artificial intelligence processing module 300 to complete intelligent analysis of the target image. Further, referring to Figure 2 The USB receiving module 200 is configured to receive image information to be processed, and convert the image information to be processed into image information to be processed. The image information to be processed can be information received through the USB interface 31. Then, the USB receiving module 200 can convert the received image information to be processed into image information to be processed.

[0085] Optionally, the USB receiving module 200 includes a USB controller unit 201 connected to the USB interface 31, and configured to receive the image information to be processed from the USB interface 31.

[0086] In addition, referring to Figure 1As shown, the USB receiving module 200 can include a plurality of USB controller units 201-20n, for example, so as to receive the image information to be processed from different USB input interfaces 31.

[0087] Optionally, the artificial intelligence processing module 300 includes a plurality of deep learning processor units 301-30n, wherein the deep learning processor units 301-30n are configured to intelligently analyze the target objects in the image information to be processed, and generate the intelligent analysis results of the target objects.

[0088] Optionally, the control module 100 is configured to determine the network model and the number of deep learning processor units used for the intelligent analysis of the target objects in the image information to be processed according to the user configuration requirements.

[0089] Optionally, the artificial intelligence processing module 300 includes a plurality of deep learning processor units 301-30n, wherein the deep learning processor units 301-30n are configured to intelligently analyze the target objects in the image information to be processed, and generate the intelligent analysis results of the target objects; and the control module 100 is further configured to determine the deep learning processor units used for the intelligent analysis of the target objects in the image information to be processed according to the USB controller units receiving the image information to be processed.

[0090] Optionally, the artificial intelligence processor 10 further includes a preprocessing module 400, wherein the preprocessing module 400 is connected with the control module 100, and is configured to pre-process a first image in the image information to be processed, and generate a second image suitable for the intelligent analysis of the artificial intelligence processing module 300.

[0091] Optionally, the preprocessing module 400 includes an image format conversion unit 410 configured to convert a first image format of the first image into a second image format suitable for the intelligent analysis of the artificial intelligence processing module 300; a resolution conversion unit 420 configured to convert a first image resolution of the first image into a second image resolution suitable for the intelligent analysis of the artificial intelligence processing module 300; and an image enhancement unit 430 configured to suppress noise in the first image, and enhance detailed information in the first image.

[0092] Optionally, in case that the first image resolution is higher than the second image resolution suitable for the intelligent analysis of the artificial intelligence processing module 300, the resolution conversion unit 420 is configured to convert the first image resolution of the first image into the second image resolution suitable for the intelligent analysis of the artificial intelligence processing module 300 by any one of the following operations: downsampling the first image so as to convert the first image resolution of the first image into the second image resolution; downsampling the first image and deforming the first image so as to convert the first image resolution of the first image into the second image resolution; cropping the first image and downsampling the first image so as to convert the first image resolution of the first image into the second image resolution; and segmenting the first image so as to convert the first image resolution of the first image into a plurality of second image resolutions.

[0093] Optionally, in case that the first image resolution is lower than the second image resolution suitable for the intelligent analysis of the artificial intelligence processing module 300, the resolution conversion unit 420 is configured to convert the first image resolution of the first image into the second image resolution suitable for the intelligent analysis of the artificial intelligence processing module 300 by any one of the following operations: upsampling the first image so as to convert the first image resolution of the first image into the second image resolution; upsampling the first image and deforming the first image so as to convert the first image resolution of the first image into the second image resolution; and upsampling the first image and cropping the first image so as to convert the first image resolution of the first image into the second image resolution.

[0094] Optionally, the artificial intelligence processor 10 further comprises an output module 500, wherein the output module 500 is connected with the control module 100 and is configured to output the image information data processed by the artificial intelligence processing module 300.

[0095] Optionally, the output module 500 comprises an HDMI / DP output unit 510 configured to be connected with the HDMI / DP output interface 32 of the external device and transmit the processed image information data through the HDMI / DP output interface 32, and / or a network access unit 520 configured to be connected with the network access interface 33 of the external device and transmit the processed image information data through the network.

[0096] Optionally, the artificial intelligence processor 10 further comprises a preprocessing module 400, an output module 500 and a memory module 600, wherein the preprocessing module 400 is connected with the control module 100 and configured to pre-process the first image in the video to be processed to generate a second image suitable for intelligent analysis of the artificial intelligence processing module 300; the output module 500 is connected with the control module 100 and configured to output the image information data processed by the artificial intelligence processing module 300; and the memory module 600 is connected with the USB receiving module 200, the artificial intelligence processing module 300, the preprocessing module 400 and the output module 500 and configured to be connected with the memory 20 of the external device.

[0097] Therefore, according to the technical scheme of the embodiment, the control module 100, the USB receiving module 200, the artificial intelligence processing module 300, the preprocessing module 400 and the output module 500 are integrated in an artificial intelligence processor 10, thereby simplifying the structure of the artificial intelligence analysis device 30, and further saving power consumption and improving operation speed. ​ As shown in the figure, the artificial intelligence analysis device 30 only needs to be configured with an input interface (such as a USB interface 31), an artificial intelligence processor 10 and an output interface (such as an HDMI / DP output interface 32 and a network input interface 33).

[0098] Unless specifically stated otherwise, the relative arrangements of the components and steps illustrated in these embodiments and the numerical expressions and values set forth herein are not limiting. It is to be understood that the dimensions of the various parts shown in the drawings are not necessarily to scale, and that for the purposes of the description herein, like parts bear like reference numbers throughout the drawings. Techniques, methods, and apparatus known to those of ordinary skill in the art can not be discussed in detail, but can be assumed to be part of the enabling description. In all examples shown and discussed herein, any specific values should be interpreted as merely exemplary, and not as a limitation on the scope of the exemplary embodiments. Thus, other example embodiments can have different values. It is to be noted that like reference numerals and letters refer to like items in the drawings, and, once an item is defined in one drawing, it need not be discussed further in subsequent drawings.

[0099] For purposes of the description hereinafter, the terms "upper", "lower", "right", "left", "vertical", "horizontal", "top", "bottom", and derivatives or

[0100] In the description of the application, it is to be understood that the orientation terms such as "front", "back", "upper", "lower", "left", "right", "lateral", "vertical", "horizontal", and "top", "bottom" and derivatives thereof, refer to the orientation or position of the device as shown in the drawings, and are used only for convenience in describing the present application and its components, and are not intended to limit the scope of the application to only the orientation or position as shown in the drawings. The orientation terms "inner", "outer" refer to the inner and outer with respect to the outline of the components themselves.

[0101] The above description is only the preferred embodiment of the present application, but the protection scope of the present application is not limited to this. Any changes or replacements within the technical scope disclosed by the present application can be easily thought by those skilled in the art, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. An artificial intelligence processor (10) characterized by, The application relates to a USB image intelligent processing system, which comprises a control module (100), a USB receiving module (200) and an artificial intelligence processing module (300), wherein the control module (100) is connected with the USB receiving module (200) and the artificial intelligence processing module (300) respectively and is used for controlling the USB receiving module (200) and the artificial intelligence processing module (300); the USB receiving module (200) is used for receiving image information to be processed sent by an image acquisition module of an external device; and the artificial intelligence processing module (300) is used for intelligently analyzing target objects in the image information to be processed. The USB receiving module (200) comprises a plurality of USB controller units (201-20n), the plurality of USB controller units (201-20n) are connected with USB interfaces (31) respectively and are configured to receive a plurality of image information to be processed from different USB interfaces (31), wherein each image information to be processed contains different target objects, and wherein the artificial intelligence processing module (300) comprises a plurality of deep learning processor units (301-30n), wherein the plurality of deep learning processor units (301-30n) are configured to intelligently analyze target objects in a plurality of different image information to be processed respectively and generate intelligent analysis results of the target objects; the control module (100) is further configured to determine the number of deep learning processor units (301-30n) corresponding to the USB controller units (201-20n) according to the number of the USB controller units (201-20n); and the control module (100) is further configured to determine network models used for intelligently analyzing different target objects respectively and the number of deep learning processor units (301-30n) corresponding to the network models. The application further comprises a preprocessing module (400), wherein the preprocessing module (400) is connected with the control module (100) and is configured to pre-process a first image in the image information to be processed and generate a second image suitable for intelligent analysis of the artificial intelligence processing module (300). The preprocessing module (400) comprises a format conversion unit (410) configured to convert a first image format of the first image into a second image format suitable for intelligent analysis of the artificial intelligence processing module (300); a resolution conversion unit (420) configured to convert a first image resolution of the first image into a second image resolution suitable for intelligent analysis of the artificial intelligence processing module (300); and an image enhancement unit (430) configured to suppress noise in the first image and enhance detailed information in the first image. The application further comprises an output module (500), wherein the output module (500) is connected with the control module (100) and is configured to output image information data processed by the artificial intelligence processing module (300). The output module (500) comprises ​ ​ ​ ​ ​ 2. The artificial intelligence processor (10) of claim 1, wherein ​ ​ 3. The artificial intelligence processor (10) of claim 2, characterized by ​ ​ ​ ​ 4. The artificial intelligence processor (10) of claim 1, wherein, ​ ​ 5. The artificial intelligence processor (10) according to claim 4, characterized in that ​ An HDMI / DP output unit (510) configured to be connected with an HDMI / DP output interface (32) of the peripheral device and transmit the processed image information data through the HDMI / DP output interface (32); and / or A network access unit (520) configured to be connected with a network access interface (33) of the peripheral device and transmit the processed image information data through a network.

6. The artificial intelligence processor (10) of claim 1, wherein, Further comprising: A preprocessing module (400), an output module (500), and a memory module (600), wherein The preprocessing module (400) is connected with the control module (100) and configured to preprocess a first image in the image information to be processed to generate a second image suitable for intelligent analysis by the artificial intelligence processing module (300); The output module (500) is connected with the control module (100) and configured to output the image information data processed by the artificial intelligence processing module (300); And The memory module (600) is connected with the USB receiving module (200), the artificial intelligence processing module (300), the preprocessing module (400), and the output module (500) and configured to be connected with a memory (20) of the peripheral device.

7. An artificial intelligence analysis device (30) comprising: A USB interface (31) configured to receive image information to be processed; and An artificial intelligence processor (10), wherein the artificial intelligence processor (10) is connected with the USB interface (31), and characterized in that the artificial intelligence processor (10) comprises a control module (100), a USB receiving module (200), and an artificial intelligence processing module (300), wherein The control module (100) is connected with the USB receiving module (200) and the artificial intelligence processing module (300) respectively to control the USB receiving module (200) and the artificial intelligence processing module (300); The USB receiving module (200) is connected with the USB interface (31) to receive image information to be processed sent by an image acquisition device through the USB interface (31), wherein the image acquisition device is arranged outside the artificial intelligence analysis device and is detachably connected with the artificial intelligence analysis device; and The artificial intelligence processing module (300) is configured to intelligently analyze a target object in the image information to be processed, wherein The USB receiving module (200) comprises a plurality of USB controller units (201-20n) connected with the USB interface (31) respectively and configured to receive a plurality of image information to be processed from different USB interfaces (31), wherein each image information to be processed contains a different target object, and wherein The artificial intelligence processing module (300) comprises a plurality of deep learning processor units (301-30n), wherein the plurality of deep learning processor units (301-30n) are configured to respectively perform intelligent analysis on target objects in a plurality of different to-be-processed image information, and generate intelligent analysis results of the respective target objects; and The control module (100) is further configured to determine the number of deep learning processor units (301-30n) corresponding to the USB controller units (201-20n) according to the number of the USB controller units (201-20n); and The control module (100) is further configured to determine network models used for intelligent analysis of different target objects respectively, and the number of deep learning processor units (301-30n) corresponding to the used network models.

8. The artificial intelligence analysis device (30) according to claim 7, characterized in that The artificial intelligence analysis device (30) further comprises an HDMI / DP output interface (32) for transmitting the processed image information data, and the artificial intelligence processor (10) further comprises an output module (500), wherein the output module (500) is connected with the control module (100) and the HDMI / DP output interface (32), and is configured to output the image information data processed by the artificial intelligence processing module (300); and / or The artificial intelligence analysis device (30) further comprises a network access interface (33) for transmitting the processed image information data through a network, and the artificial intelligence processor (10) further comprises an output module (500), wherein the output module (500) is connected with the control module (100) and the network access interface (33), and is configured to output the image information data processed by the artificial intelligence processing module (300).

Images