Image filtering processing system

Abstract

The invention relates to an image filtering processing system, and the system comprises a hierarchical training mechanism which is used for carrying out the multi-layer training of a deep convolutional network, so as to obtain the output of an AI judgment model; an imaging lens and a photoelectric sensing assembly are arranged in the pinhole snapshot mechanism, and the pinhole snapshot mechanism is used for acquiring a snapshot picture corresponding to the current moment and outputting the snapshot picture as a current snapshot picture; and the state judgment component is used for intelligently judging the processed sub-processing type of the current snapshot picture by adopting an AI judgment model based on each processing reference image, the original image, the current snapshot picture and each shooting parameter of the pinhole snapshot mechanism. The image filtering processing system is stable in operation and intelligent in control. The AI judgment model can be adopted to execute intelligent judgment on the sub-processing type passed by the current snapshot picture based on the multiple items of associated data screened in a targeted manner, so that effective identification of real information of the received picture is ensured.

Description

Technical Field

[0001] The present invention relates to the field of filtering processing, and in particular to an image filtering processing system. Background Art

[0002] Image filtering is an image processing technique that can change the appearance and characteristics of an image by applying various filters to it. The following are some common types of image filtering: Mean filtering: Taking the average of the values ​​around a pixel can smooth the image and reduce noise.

[0003] Median filter: Replace the value of the central pixel with the median of the values ​​around the pixel, which is suitable for removing salt and pepper noise.

[0004] Gaussian filtering: Using the weight of the Gaussian function to perform weighted averaging on the values ​​around the pixel can smooth the image and retain more image details.

[0005] Bilateral filtering: It combines the information in the spatial domain and the grayscale value domain. It can retain edge information while smoothing the image. It is suitable for processing images with complex textures and edge information.

[0006] Sobel filter: used for edge detection, it can detect horizontal and vertical edges in images.

[0007] Laplacian filtering: used to enhance the high-frequency features of an image and make the image sharper.

[0008] Canny filter: It is used to detect edges in images and has the advantages of high accuracy and low false alarm rate.

[0009] At present, it is very likely to be clear that the received picture has undergone major types of image processing, such as image filtering processing, but it is impossible to determine the small types of image processing that the received picture has undergone, such as mean filtering, median filtering, Gaussian filtering, bilateral filtering, Sobel filtering, Laplacian filtering or Canny filtering as mentioned above, which makes it difficult to specifically identify the real information of the received picture. Summary of the invention

[0010] In order to solve the technical problems in the related fields, the present invention provides an image filtering processing system, which uses the various shooting parameters of the pinhole capture mechanism, the various processing reference images, the original image and the current captured picture as multiple basic information for intelligent judgment of the sub-processing type of image processing that the current captured picture has undergone, thereby ensuring the validity of the intelligent judgment result, and customizing the structure of the AI ​​judgment model for executing the intelligent judgment of the sub-processing type, which is manifested in: performing multi-layer training on the deep convolutional network to obtain the AI ​​judgment model, and performing each layer of learning on the deep convolutional network is performing each training process on the deep convolutional network, wherein the number of training layers performed on the deep convolutional network is positively correlated with the horizontal resolution of the captured picture and the vertical resolution of the captured picture, thereby providing AI judgment models with different structures for different pinhole capture mechanisms, ensuring that the real information of the received picture is targetedly identified.

[0011] According to the present invention, there is provided an image filtering processing system, the system comprising: A hierarchical training mechanism, used for performing multi-layer training on the deep convolutional network to obtain an AI judgment model output, wherein performing each layer of learning on the deep convolutional network is performing each training process on the deep convolutional network; The pinhole capture mechanism has a built-in imaging lens and a photoelectric sensor component, which is used to obtain the captured image corresponding to the current moment and output it as the current captured image; A first acquisition component is connected to the pinhole capture mechanism and is used to acquire the original image output by the pinhole capture mechanism and obtain the processed reference images after being processed by various sub-processing types under the image filtering processing type, wherein the various sub-processing types include statistical sorting filtering processing, gradient sharpening filtering processing, homomorphic filtering processing and arithmetic mean filtering processing; The second acquisition component is connected to the pinhole capture mechanism and is used to collect various shooting parameters of the pinhole capture mechanism, wherein the various shooting parameters of the pinhole capture mechanism are the capture frame rate, capture picture clarity, capture picture horizontal resolution and capture picture vertical resolution of the pinhole capture mechanism; A state judgment component is connected to the hierarchical training mechanism, the first acquisition component and the second acquisition component respectively, and is used to intelligently judge the sub-processing type of the current captured picture based on each processing reference image, the original image, the current captured picture and each shooting parameter of the pinhole capture mechanism by using an AI judgment model, wherein each processing reference image, the original image and the current captured picture are all in the type name representation mode of the binary data stream, and the sub-processing type of the current captured picture that is intelligently judged to be processed is in the type name representation mode of the ASCII code; The hierarchical training mechanism is used to perform multi-layer training on the deep convolutional network to obtain the AI ​​judgment model output, and each layer of learning on the deep convolutional network is performed for each training process on the deep convolutional network, including: the number of layers of training performed on the deep convolutional network is positively correlated with the horizontal resolution of the captured picture and positively correlated with the vertical resolution of the captured picture; Among them, the first acquisition component is used to collect the original image output by the pinhole capture mechanism and obtain various processed reference images after being processed by various sub-processing types under the image filtering processing type. The various sub-processing types include statistical sorting filtering processing, gradient sharpening filtering processing, homomorphic filtering processing and arithmetic mean filtering processing, including: the original image output by the pinhole capture mechanism is a sensing image directly sensed and output by the photoelectric sensing component of the pinhole capture mechanism.

[0012] It can be seen that the present invention mainly has the following three significant technical effects: Firstly, the various shooting parameters of the pinhole capture mechanism, various processing reference images, the original image and the current captured image are used as multiple basic information for intelligent judgment of the image processing of the sub-processing type of the current captured image, thereby ensuring the validity of the intelligent judgment result; Secondly: Specifically, each set of shooting parameters of the pinhole capture mechanism is the capture frame rate, capture picture clarity, capture picture horizontal resolution and capture picture vertical resolution of the pinhole capture mechanism, and the type name representation mode of each image participating in the intelligent judgment is the numerical representation of the binary data stream of the image; Again: the structural customization of the AI ​​judgment model for executing intelligent judgment of sub-processing types is manifested in: performing multi-layer training on the deep convolutional network to obtain the AI ​​judgment model, and performing each layer of learning on the deep convolutional network is performing each training process on the deep convolutional network, wherein the number of training layers performed on the deep convolutional network is positively correlated with the horizontal resolution of the captured image and positively correlated with the vertical resolution of the captured image, thereby providing AI judgment models with different structures for different pinhole capture mechanisms.

[0013] The image filtering processing system of the present invention is stable in operation and intelligent in control. Since it can use an AI judgment model based on a plurality of targeted selected related data to perform intelligent judgment on the sub-processing type of the current captured image, effective identification of the real information of the received image is ensured. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The embodiments of the present invention will be described below with reference to the accompanying drawings, wherein:

[0015] Figure 1 FIG. 1 is a schematic diagram of the internal structure of an image filtering processing system according to a first embodiment of the present invention.

[0016] Figure 2 FIG. 4 is a schematic diagram of the internal structure of an image filtering processing system according to a second embodiment of the present invention.

[0017] Figure 3 FIG. 4 is a schematic diagram of the internal structure of an image filtering processing system according to a third embodiment of the present invention. DETAILED DESCRIPTION

[0018] The following will describe in detail an embodiment of the image filtering processing system of the present invention with reference to the accompanying drawings.

[0019] Figure 1 This is a schematic diagram of the internal structure of an image filtering processing system according to a first embodiment of the present invention, wherein the system comprises: A hierarchical training mechanism, used for performing multi-layer training on the deep convolutional network to obtain an AI judgment model output, wherein performing each layer of learning on the deep convolutional network is performing each training process on the deep convolutional network; For example, a hierarchical training mechanism is used to perform multi-layer training on a deep convolutional network to obtain an AI judgment model output, and performing each layer of learning on the deep convolutional network is to perform each training process on the deep convolutional network, including: selecting and using a GAL chip to implement the hierarchical training mechanism, which is used to perform multi-layer training on the deep convolutional network to obtain an AI judgment model output, and performing each layer of learning on the deep convolutional network is to perform each training process on the deep convolutional network; The pinhole capture mechanism has a built-in imaging lens and a photoelectric sensor component, which is used to obtain the captured image corresponding to the current moment and output it as the current captured image; A first acquisition component is connected to the pinhole capture mechanism and is used to acquire the original image output by the pinhole capture mechanism and obtain the processed reference images after being processed by various sub-processing types under the image filtering processing type, wherein the various sub-processing types include statistical sorting filtering processing, gradient sharpening filtering processing, homomorphic filtering processing and arithmetic mean filtering processing; The second acquisition component is connected to the pinhole capture mechanism and is used to collect various shooting parameters of the pinhole capture mechanism, wherein the various shooting parameters of the pinhole capture mechanism are the capture frame rate, capture picture clarity, capture picture horizontal resolution and capture picture vertical resolution of the pinhole capture mechanism; A state judgment component is connected to the hierarchical training mechanism, the first acquisition component and the second acquisition component respectively, and is used to intelligently judge the sub-processing type of the current captured picture based on each processing reference image, the original image, the current captured picture and each shooting parameter of the pinhole capture mechanism by using an AI judgment model, wherein each processing reference image, the original image and the current captured picture are all in the type name representation mode of the binary data stream, and the sub-processing type of the current captured picture that is intelligently judged to be processed is in the type name representation mode of the ASCII code; The hierarchical training mechanism is used to perform multi-layer training on the deep convolutional network to obtain the AI ​​judgment model output, and each layer of learning on the deep convolutional network is performed for each training process on the deep convolutional network, including: the number of layers of training performed on the deep convolutional network is positively correlated with the horizontal resolution of the captured picture and positively correlated with the vertical resolution of the captured picture; Among them, the first acquisition component is used to acquire the original image output by the pinhole capture mechanism, and each processed reference image is obtained after being processed by each sub-processing type under the image filtering processing type, and each sub-processing type includes statistical sorting filtering processing, gradient sharpening filtering processing, homomorphic filtering processing and arithmetic mean filtering processing, including: the original image output by the pinhole capture mechanism is a sensing image directly sensed and output by the photoelectric sensing component of the pinhole capture mechanism; And wherein, an AI judgment model is used to intelligently judge the sub-processing type of the current captured picture based on each processed reference image, the original image, the current captured picture and each shooting parameter of the pinhole capture mechanism, wherein each processed reference image, the original image and the current captured picture are all type name representation modes of binary data streams, and the intelligently judged sub-processing type of the current captured picture processed is the type name representation mode of ASCII code, including: performing binary data normalization processing on each processed reference image, the original image, the current captured picture and each shooting parameter of the pinhole capture mechanism, and then synchronously inputting them into the AI ​​judgment model.

[0020] Figure 2 FIG. 4 is a schematic diagram of the internal structure of an image filtering processing system according to a second embodiment of the present invention.

[0021] Compared to Figure 1 According to the second embodiment of the present invention, the image filtering processing system may further include: A positioning server component is connected to the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component respectively, and is used to provide the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component with their respective current instant positioning data; The positioning server is respectively connected to the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component, and is used to provide the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component with their respective current instant positioning data, including: the positioning server includes a plurality of positioning service units, which are respectively connected to the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component, so as to complete the respective supply of the current instant positioning data of the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component; Wherein, the positioning server device includes a plurality of positioning service units, which are used to be connected to the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component respectively, so as to complete the respective supply of the current instant positioning data of the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component respectively, including: the plurality of positioning service units are a plurality of positioning sensors, which are used to be connected to the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component respectively, so as to complete the respective supply of the current instant positioning data of the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component respectively; Wherein, the plurality of positioning service units are a plurality of positioning sensors, which are used to be respectively connected with the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component to respectively supply the current instant positioning data of the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component, including: the structures of the plurality of positioning service units are the same; And wherein, the multiple positioning service units are multiple positioning sensors, which are used to respectively connect with the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component to complete the respective supply of current instant positioning data to the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component, and also include: the multiple positioning sensors have the same positioning range.

[0022] Figure 3 FIG. 4 is a schematic diagram of the internal structure of an image filtering processing system according to a third embodiment of the present invention.

[0023] Compared to Figure 1 According to the third embodiment of the present invention, the image filtering processing system may further include: A user control interface is connected to the state judgment component, the hierarchical training mechanism, the first acquisition component and the plurality of positioning service units of the second acquisition component, respectively, and is used to synchronously control the current working mode of the respective positioning service units of the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component; Wherein, the user control interface is respectively connected to the multiple positioning service units of the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component, and is used to synchronously control the current working mode of the positioning service units of the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component respectively, including: the current working mode of the positioning service units of the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component respectively is a sleep working mode or a running working mode; And wherein, the user control interface is respectively connected to the multiple positioning service units of the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component, and is used to synchronously control the current working mode of the respective positioning service units of the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component, including: the multiple positioning service units of the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component are all based on the Beidou positioning mechanism.

[0024] In addition, in the image filtering and processing system, an AI judgment model is used to intelligently judge the sub-processing type of the current captured picture based on each processing reference image, the original image, the current captured picture and each shooting parameter of the pinhole capture mechanism, wherein each processing reference image, the original image and the current captured picture are all type name representation modes of binary data streams, and the type name representation mode of the sub-processing type of the current captured picture that is intelligently judged to be processed is an ASCII code, and also includes: using a numerical simulation mode to implement testing and simulation of the data processing process of using the AI ​​judgment model to intelligently judge the sub-processing type of the current captured picture based on each processing reference image, the original image, the current captured picture and each shooting parameter of the pinhole capture mechanism.

[0025] Thus, the objects and advantages can be achieved through the present invention. Although preferred specific examples have been disclosed and described in detail herein, the scope and objects of the present invention are not limited thereto. The present invention is defined by the appended claims.

Claims

1. An image filtering processing system, characterized in that: The system comprises: A hierarchical training mechanism, used for performing multi-layer training on the deep convolutional network to obtain an AI judgment model output, wherein performing each layer of learning on the deep convolutional network is performing each training process on the deep convolutional network; Pinhole capture mechanism, with built-in imaging lens and photoelectric sensor components, is used to obtain the snapshot image corresponding to the current moment and output it as the current snapshot image; A first acquisition component is connected to the pinhole capture mechanism and is used to acquire the original image output by the pinhole capture mechanism and obtain the processed reference images after being processed by various sub-processing types under the image filtering processing type, wherein the various sub-processing types include statistical sorting filtering processing, gradient sharpening filtering processing, homomorphic filtering processing and arithmetic mean filtering processing; The second acquisition component is connected to the pinhole capture mechanism and is used to collect various shooting parameters of the pinhole capture mechanism, wherein the various shooting parameters of the pinhole capture mechanism are the capture frame rate, capture picture clarity, capture picture horizontal resolution and capture picture vertical resolution of the pinhole capture mechanism; A state judgment component is connected to the hierarchical training mechanism, the first acquisition component and the second acquisition component respectively, and is used to intelligently judge the sub-processing type of the current captured picture based on each processing reference image, the original image, the current captured picture and each shooting parameter of the pinhole capture mechanism by using an AI judgment model, wherein each processing reference image, the original image and the current captured picture are all in the type name representation mode of the binary data stream, and the sub-processing type of the current captured picture that is intelligently judged to be processed is in the type name representation mode of the ASCII code; The hierarchical training mechanism is used to perform multi-layer training on the deep convolutional network to obtain the AI ​​judgment model output, and each layer of learning on the deep convolutional network is performed for each training process on the deep convolutional network, including: the number of layers of training performed on the deep convolutional network is positively correlated with the horizontal resolution of the captured picture and positively correlated with the vertical resolution of the captured picture; Among them, the first acquisition component is used to collect the original image output by the pinhole capture mechanism and obtain various processed reference images after being processed by various sub-processing types under the image filtering processing type. The various sub-processing types include statistical sorting filtering processing, gradient sharpening filtering processing, homomorphic filtering processing and arithmetic mean filtering processing, including: the original image output by the pinhole capture mechanism is a sensing image directly sensed and output by the photoelectric sensing component of the pinhole capture mechanism.

2. The image filtering processing system according to claim 1, wherein: An AI judgment model is adopted to intelligently judge the sub-processing type of the current captured picture based on each processed reference image, the original image, the current captured picture and each shooting parameter of the pinhole capture mechanism, wherein each processed reference image, the original image and the current captured picture are all type name representation modes of binary data streams, and the intelligently judged sub-processing type of the current captured picture processed is a type name representation mode of ASCII code, including: performing binary data normalization processing on each processed reference image, the original image, the current captured picture and each shooting parameter of the pinhole capture mechanism, and then synchronously inputting them into the AI ​​judgment model.

3. The image filtering processing system according to claim 2, characterized in that: The system further comprises: A positioning server component is connected to the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component respectively, and is used to provide the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component with their respective current instant positioning data; Among them, the positioning server is respectively connected to the state judgment component, the hierarchical training mechanism, the first collection component and the second collection component, and is used to provide the state judgment component, the hierarchical training mechanism, the first collection component and the second collection component with their respective current and real-time positioning data, including: the positioning server includes multiple positioning service units, which are respectively connected to the state judgment component, the hierarchical training mechanism, the first collection component and the second collection component to complete the respective supply of current and real-time positioning data to the state judgment component, the hierarchical training mechanism, the first collection component and the second collection component.

4. The image filtering processing system according to claim 3, wherein: The positioning server device includes multiple positioning service units, which are used to connect with the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component respectively, so as to complete the respective supply of current and real-time positioning data of the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component. The multiple positioning service units are multiple positioning sensors, which are used to connect with the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component respectively, so as to complete the respective supply of current and real-time positioning data of the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component.

5. The image filtering processing system according to claim 4, characterized in that: The multiple positioning service units are multiple positioning sensors, which are used to connect with the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component respectively to complete the respective supply of current instant positioning data to the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component, including: the structures of the multiple positioning service units are the same.

6. The image filtering processing system according to claim 5, characterized in that: The multiple positioning service units are multiple positioning sensors, which are used to respectively connect with the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component to complete the respective supply of current instant positioning data to the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component, and also include: the multiple positioning sensors have the same positioning range.

7. The image filtering processing system according to claim 2, characterized in that: The system further comprises: The user control interface is respectively connected to multiple positioning service units of the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component, and is used to synchronously control the current working mode of the respective positioning service units of the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component.

8. The image filtering processing system according to claim 7, wherein: The user control interface is respectively connected to multiple positioning service units of the state judgment component, the layered training mechanism, the first acquisition component and the second acquisition component, and is used to synchronously control the current working mode of the positioning service units of the state judgment component, the layered training mechanism, the first acquisition component and the second acquisition component, including: the current working mode of the positioning service units of the state judgment component, the layered training mechanism, the first acquisition component and the second acquisition component is a sleep working mode or a running working mode.

9. The image filtering processing system according to claim 7, wherein: A user control interface is respectively connected to the multiple positioning service units of the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component, and is used to synchronously control the current working mode of the respective positioning service units of the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component, including: the multiple positioning service units of the state judgment component, the hierarchical training mechanism, the first acquisition component and the second acquisition component are all based on the Beidou positioning mechanism.

Images