A panoramic parking digital camera access system, method, electronic device and storage medium

Abstract

The embodiment of the present invention discloses a panoramic parking digital camera access system, method, electronic device and storage medium, wherein the system includes a baseboard, a core board and N-way cameras, the core board is provided with a central processing unit, a video processor, and N-way ipex connection ports; the N-way cameras are connected to the N-way ipex connection ports through N-way interfaces; the N-way cameras obtain image data, and transmit the image data to the video processor through the N-way interface and the N-way ipex connection ports, the video processor processes the image data, and transmits the processed image data to the central processing unit, the central processing unit executes the panoramic parking algorithm based on the processed image data, and outputs the display screen corresponding to the panoramic parking algorithm. The implementation of the embodiment of the present application can reduce the interference and abnormal number of video signals, improve the picture quality; and can improve the efficiency of the panoramic parking algorithm and improve the quality of the displayed content.

Description

Technical Field

[0001] The present invention relates to the field of data processing, specifically to the field of panoramic parking, and in particular to a panoramic parking digital camera access system, method, electronic equipment and storage medium. Background Art

[0002] At present, the camera signal is connected to the core board through the socket from the bottom board, and then processed by the core board through the camera signal. The current camera is powered by the bottom board. Since the camera power supply and video signal are transmitted on the same line, the power supply needs special processing and cannot be compatible with existing cameras, resulting in the inability to support some models of digital cameras; or, after special processing of the power supply, although it can achieve compatibility with some cameras, since the power supply and camera signal are on the same line, signal interference and other signals will still appear, causing signal loss, such as video anomalies, picture jitter, screen distortion, etc. In addition, in the current panoramic parking process, the panoramic parking algorithm is inefficient and the image quality is poor, resulting in longer waiting time for users and poor content displayed on the vehicle display screen. Summary of the invention

[0003] The embodiments of the present invention provide a panoramic parking digital camera access system, method, electronic device and storage medium, which can avoid interference with video signals, improve the efficiency of the panoramic parking algorithm, and improve the display effect of the parking picture.

[0004] On the one hand, the present invention is an embodiment to provide a panoramic parking digital camera access system, the system includes a baseboard, a core board and N cameras, the core board is provided with a central processing unit, a video processor, and N ipex connection ports, N ≥ 2 and N is a positive integer;

[0005] The N-way cameras are connected to the N-way ipex connection ports through N-way interfaces; the N-way ipex connection ports are communicatively connected to the video processor, and the video processor is communicatively connected to the central processing unit;

[0006] When the N cameras are connected to the N ipex connection ports via N interfaces, the N cameras acquire image data, and transmit the image data to the video processor via the N interfaces and the N ipex connection ports. The video processor processes the image data and transmits the processed image data to the central processing unit. The central processing unit executes a panoramic parking algorithm based on the processed image data and outputs a display screen corresponding to the panoramic parking algorithm.

[0007] In a possible implementation, the N-way ipex connection port is an N-way IPEX patch socket, and each of the N-way interfaces includes: an ipex line end seat, a camera coaxial transmission line and a FAKRA connector. The N-way camera is connected to the N-way ipex connection port through the N-way interface, including: the N-way cameras are respectively connected to the N-way FAKRA connectors, the N-way FAKRA connectors are respectively connected to the first ends of the N-way camera coaxial transmission lines, the second ends of the N-way camera coaxial transmission lines are respectively connected to the N ipex line ends, and the N ipex line ends are respectively connected to the N-way IPEX patch sockets.

[0008] In a possible implementation, the base plate includes an LVDS video signal output socket; the base plate is connected to the core board through a first pin connector and a second pin connector, and the central processing unit executes a panoramic parking algorithm based on the processed image data, and outputs a display screen corresponding to the panoramic parking algorithm, including: the central processing unit obtains video output data after executing the panoramic parking algorithm; the central processing unit transmits the video output data to the base plate through the first pin connector and the second pin connector; the base plate performs video encoding on the video output data to obtain the display screen in the target format; the base plate transmits the display screen to the vehicle display screen through the LVDS video signal output socket; the vehicle display screen displays the display screen.

[0009] On the other hand, the present invention provides a panoramic parking method, which is applied to a panoramic parking digital camera access system, and the method comprises:

[0010] Step 1: Obtain different views taken by N cameras at the same time;

[0011] Step 2: For each view obtained by different cameras, use the encoder to obtain a set of encoding parameters Pi = (P 1，i , P 2，i ,P c，i ,P S，i ), where P 1，i , P 2，i are the feature parameters of the global target feature and the local target feature of the i-th camera, respectively, c，i is the position characteristic parameter of the i-th camera, P S，i is the shape parameter of the target in the i-th camera;

[0012] Step 3: Weight the encoding parameters of the N views to obtain a set of normalized encoder parameters P = (P1, P2, P c ,P S );

[0013] Step 4: Set the parameter P = (P1, P2, P c ,P S ) Input the decoder for three-dimensional reconstruction to obtain the obstacle target in the path;

[0014] Step 5: Plan the parking path based on the obstacle detection results in the path.

[0015] It should be noted that in each view acquired by the camera, four parameters that best represent the features of the view are selected, namely, global target feature points, local target feature points, the position parameters of the camera and the shape of the target. The global target features reflect the characteristics of the target in the view as a whole, local target selection, the local area with the richest features in the target, the camera's position encoding reflects the positional characteristics of the view, and the shape of the target can provide better supervision for 3D reconstruction.

[0016] In a possible implementation, for each view acquired by different cameras, an encoder is used to obtain a set of encoding parameters Pi=(P 1，i , P 2，i ,P c，i ,P S，i ),include:

[0017] For each view acquired by different cameras, split it into non-overlapping regions, add position embedding to the non-overlapping regions and linearly project them to the image Ti, concatenate the image Ti with the labeled token Tt of the learnable task, and encode the concatenated image into Ti1 and Tt1 using the ViT network;

[0018] Use the pre-trained prediction cropping module to predict the global target features and local target features of the target in the Ti1 feature map;

[0019] Use MLP to regress a set of encoder parameters Pi = (P 1，i , P 2，i ,P c，i ,P S，i ), where P 1，i , P 2，i are the feature parameters of the global target feature and the local target feature of the i-th camera, respectively, which are obtained by an MLP regression module based on the global target feature and the local target feature of the target in the predicted Ti1 feature map; P c，i is the position characteristic parameter of the i-th camera, P S，i is the shape parameter of the target in the i-th camera.

[0020] It should be noted that the ViT network is a transformer model directly applied to a sequence of image blocks, which can obtain better image classification results on large-scale data sets. The present invention encodes the view of each camera based on the ViT network; in the ViT network, we need to project the image blocks to obtain image block embeddings, and preset an embedding of a learnable task for the image block embedding sequence. In addition, the classification head uses MLP as the classification head to process the image. In addition, the position embedding is added to the image block embedding to retain the spatial position information between the input image blocks. This provides accurate position encoding to obtain better three-dimensional reconstruction effects.

[0021] In a possible implementation, the encoding parameters of the N views are weighted to obtain a set of normalized encoder parameters P of the N views = (P1, P2, P c ,P S )include:

[0022] Use MLP1 to learn task tokens Tt1 and camera parameters P c Get output score

[0023] S1 i =MLP1(Tt1,P c，i ),

[0024] Then the global target feature P1 of multiple views is fused as:

[0025] Use MLP2 to pass local target feature token T 2，i and camera parameters P c Get output score: S2 i =MLP2(T 2，i , P c，i ), then the local target features P2 of multiple views are fused as:

[0026] Among them, the local target feature token T 2，i It is obtained through the prediction cropping module, MLP1 and MLP2 are network layers of the multi-layer perceptron, and N is the number of cameras.

[0027] It should be noted that when fusing N views, we adjust the weight of the global target through the MLP layer, so that more and more correct information can be selected for representation. When decoding according to the final fused encoder parameters, the 3D reconstruction effect obtained is better.

[0028] It should be noted that the decoder may adopt any existing known decoding method.

[0029] In a possible implementation, after acquiring different views captured by N cameras at the same time and before processing the views, the method further includes:

[0030] Get the image parameters of the pictures of N cameras, including the average brightness value, median brightness value, and average saturation value;

[0031] Inputting image parameters of the pictures of the N cameras into a preset adjustment model, and adjusting the image parameters of the pictures of the N cameras based on an output of the adjustment model;

[0032] Among them, the preset adjustment model is obtained based on reinforcement learning; the target reward function of reinforcement learning is constructed based on the accuracy of target recognition in image processing.

[0033] It should be noted that the N cameras have different parameters for the pictures taken due to different positions and lighting conditions. In order to obtain a better fusion effect, the image parameters are adjusted before the view is processed. The adjustment model is used to obtain the final parameters of the N views, and the adjustment of each picture is set according to the final parameters. Here, the average brightness value, median brightness, and average saturation value are selected as adjustment indicators, and the reinforcement learning method is used to select the adjustment model that best suits the current situation. This method adaptively adjusts the image condition according to the actual situation and makes targeted changes to the picture.

[0034] Those skilled in the art may also select other algorithms that can achieve better adjustment effects.

[0035] On the other hand, the present invention provides a panoramic parking device, wherein the panoramic parking device is configured with a panoramic parking digital camera access system, and the panoramic parking device is used to perform a panoramic parking method, wherein the method comprises: step 1, obtaining different views taken by N cameras at the same time; step 2, for each view obtained by different cameras, using an encoder to obtain a set of encoding parameters Pi=(P 1，i , P 2，i ,P c，i ,P s，i ), where P 1，i , P 2，i are the feature parameters of the global target feature and the local target feature of the i-th camera, respectively, c，i is the position characteristic parameter of the i-th camera, P S，i is the shape parameter of the target in the i-th camera; Step 3, weight the encoding parameters of the N views to obtain a set of encoder parameters P after normalization of the N views = (P1, P2, P c ,P S ); Step 4: Set the parameter P = (P1, P2, P c ,PS ) Input the decoder for three-dimensional reconstruction to obtain the obstacle target in the path; Step 5, plan the parking path based on the obstacle detection results in the path.

[0036] On the other hand, the present invention provides an electronic device, comprising: a processor, suitable for implementing one or more instructions; and a computer storage medium, wherein the computer storage medium stores one or more instructions, wherein the one or more instructions are suitable for being loaded by the processor and executing the above-mentioned panoramic parking method.

[0037] On the other hand, the present invention provides a computer storage medium, characterized in that the computer storage medium stores one or more instructions, and the one or more instructions are suitable for being loaded by a processor and executing the above-mentioned panoramic parking method.

[0038] The beneficial effects of this application are as follows:

[0039] (1) The present invention transmits the power supply of the digital camera by the core board PCBA2 through the connection between the camera and the core board, and the LVDS coaxial signal is directly input into the core board PCBA2 without passing through the base board PCBA1. In this way, the power supply requirement of the digital camera can be met, and the continuity of the video signal impedance and the integrity of the video signal can be ensured, thereby solving the technical problem of incompatibility with digital cameras, reducing the interference of digital signals, and improving the picture quality.

[0040] (2) By fusing N views, we can achieve a balance between efficiency and reconstruction accuracy, thereby improving the efficiency of path planning during the parking process. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

[0042] Figure 1A is an architecture diagram of a panoramic parking digital camera access system provided by an embodiment of the present invention;

[0043] Figure 1B is an architecture diagram of a panoramic parking digital camera access system from another perspective provided by an embodiment of the present invention;

[0044] Figure 2 It is a flow chart of a panoramic parking digital camera access system provided by an embodiment of the present invention;

[0045] Figure 3It is a structural schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0047] like Figure 1A , Figure 1B As shown, Figure 1A and Figure 1B 1 is an architecture diagram of a panoramic parking digital camera access system provided in an embodiment of the present application, specifically an architecture diagram obtained from two perspectives. A panoramic parking method provided in an embodiment of the present application can be used in a system architecture as shown in FIG. 1 , which includes a base board PCBA 1, a core board PCBA 2, and a four-way interface.

[0048] The bottom plate PCBA 1 includes: a power socket 9, a signal socket 10, an LVDS video signal output socket 11, an LVDS video signal input socket 12, a dip switch 6, a pin header connector A 13, a pin header connector B 14, and a TF card socket 16.

[0049] The core board PCBA 2 includes: a 4-way IPEX patch seat 5, a SoC processor 7, and a video processor 8; the camera coaxial transmission line 3 includes: an IPEX line end seat 15, and a 4-way FAKRA connector 4.

[0050] Specifically, the vehicle supplies power to the bottom plate through the power socket 9, and after the power is processed in the bottom plate PCBA 1, it is supplied to the core board PCBA2 through the bottom plate and core board pin header connection seat A 13 and the bottom plate and core board pin header connection seat B14. The vehicle provides various required signals to the bottom plate through the signal socket 10, and after the signals are processed in the bottom plate PCBA 1, they are supplied to the core board PCBA2 for processing through the bottom plate and core board pin header connection seat A 13 and the bottom plate and core board pin header connection seat B14. The 4-way digital camera is respectively connected to the 4-way FAKRA connector 4, the IPEX line terminal 15 of the camera coaxial transmission line 3 is connected to the IPEX patch seat 5 in the core board PCBA 2, and the coaxial signal of the digital camera is input into the video processor 8 in the core board PCBA2 through the camera coaxial transmission line 3. After the signal is processed by the video processor 8, the signal is input into the SoC processor 7 for data processing, which may include fusion, algorithm processing, splicing, etc.; the SoC processor 7 transmits the processed video signal to the base board PCBA 1 through the pin connector A 13 and the pin connector B14 between the base board and the core board, and then generates the corresponding format video encoding through the base board PCBA1 and inputs it to the original car screen display through the LVDS video signal output socket 11.

[0051] Through the connection transmission between the camera and the core board, the power supply of the digital camera from the core board PCBA 2 and the LVDS coaxial signal are also directly input into the core board PCBA2 without passing through the bottom board PCBA1. This can not only meet the power supply requirements of the digital camera, but also ensure the continuity of the video signal impedance, ensure the integrity of the video signal, solve the technical problem of incompatibility with digital cameras, reduce the interference of digital signals, and improve the picture quality.

[0052] Based on the above description, an embodiment of the present invention proposes a panoramic parking digital camera access system, the system includes a baseboard, a core board and N cameras, the core board is provided with a central processing unit, a video processor, and N ipex connection ports, N ≥ 2 and N is a positive integer;

[0053] The N-way cameras are connected to the N-way ipex connection ports through N-way interfaces; the N-way ipex connection ports are communicatively connected to the video processor, and the video processor is communicatively connected to the central processing unit;

[0054] When the N cameras are connected to the N ipex connection ports via N interfaces, the N cameras acquire image data, and transmit the image data to the video processor via the N interfaces and the N ipex connection ports. The video processor processes the image data and transmits the processed image data to the central processing unit. The central processing unit executes a panoramic parking algorithm based on the processed image data and outputs a display screen corresponding to the panoramic parking algorithm.

[0055] Wherein, N can be specifically 4; the bottom board is provided with a power socket 9, a signal socket 10, an LVDS video signal output socket 11, an LVDS video signal input socket 12, a code switch 6, a pin header connection socket A 13, a pin header connection socket B14, and a TF card socket 16. The vehicle supplies power to the bottom board through the power socket 9, and after the power is processed in the bottom board PCBA 1, it is supplied to the core board PCBA2 through the bottom board and the core board pin header connection socket A 13 and the bottom board and the core board pin header connection socket B14.

[0056] In a possible implementation, the N-way ipex connection port is an N-way IPEX patch socket, and each of the N-way interfaces includes: an ipex line end seat, a camera coaxial transmission line and a FAKRA connector. The N-way camera is connected to the N-way ipex connection port through the N-way interface, including: the N-way cameras are respectively connected to the N-way FAKRA connectors, the N-way FAKRA connectors are respectively connected to the first ends of the N-way camera coaxial transmission lines, the second ends of the N-way camera coaxial transmission lines are respectively connected to the N ipex line ends, and the N ipex line ends are respectively connected to the N-way IPEX patch sockets.

[0057] In a possible implementation, the base plate includes an LVDS video signal output socket; the base plate is connected to the core board through a first pin connector and a second pin connector, and the central processing unit executes a panoramic parking algorithm based on the processed image data, and outputs a display screen corresponding to the panoramic parking algorithm, including: the central processing unit obtains video output data after executing the panoramic parking algorithm; the central processing unit transmits the video output data to the base plate through the first pin connector and the second pin connector; the base plate performs video encoding on the video output data to obtain the display screen in the target format; the base plate transmits the display screen to the vehicle display screen through the LVDS video signal output socket; the vehicle display screen displays the display screen.

[0058] On the other hand, the present invention provides a panoramic parking method, which is applied to a panoramic parking digital camera access system, such as Figure 2 As shown, the method includes:

[0059] Step S201, obtaining different views captured by N cameras at the same time;

[0060] Step S202: for each view acquired by different cameras, use an encoder to obtain a set of encoding parameters Pi = (P 1，i , P 2，i ,P c，i ,P S，i ), where P 1，i , P 2，i are the feature parameters of the global target feature and the local target feature of the i-th camera, respectively, c，i is the position characteristic parameter of the i-th camera, P S，i is the shape parameter of the target in the i-th camera;

[0061] Step S203: Weight the encoding parameters of the N views to obtain a set of normalized encoder parameters P of the N views = (P1, P2, P c ,PS );

[0062] Step S204: Set the parameter P=(P1, P2, P c ,P S ) Input the decoder for three-dimensional reconstruction to obtain the obstacle target in the path;

[0063] Step S205: planning a parking path based on the obstacle detection result in the path.

[0064] Among them, the central processing unit can send the parking path planning to the on-board display screen for display.

[0065] It should be noted that in each view acquired by the camera, four parameters that best represent the features of the view are selected, namely, global target feature points, local target feature points, the position parameters of the camera and the shape of the target. The global target features reflect the characteristics of the target in the view as a whole, local target selection, the local area with the richest features in the target, the camera's position encoding reflects the positional characteristics of the view, and the shape of the target can provide better supervision for 3D reconstruction.

[0066] In a possible implementation, for each view acquired by different cameras, an encoder is used to obtain a set of encoding parameters Pi=(P 1，i , P 2，i ,P c，i ,P S，i ),include:

[0067] For each view acquired by different cameras, split it into non-overlapping regions, add position embedding to the non-overlapping regions and linearly project them to the image Ti, concatenate the image Ti with the labeled token Tt of the learnable task, and encode the concatenated image into Ti1 and Tt1 using the ViT network;

[0068] Use the pre-trained prediction cropping module to predict the global target features and local target features of the target in the Ti1 feature map;

[0069] Use MLP to regress a set of encoder parameters Pi = (P 1，i , P 2，i ,P c，i ,P S，i ), where P 1，i , P 2，i are the feature parameters of the global target feature and the local target feature of the i-th camera, respectively, which are obtained by an MLP regression module based on the global target feature and the local target feature of the target in the predicted Ti1 feature map; P c，i is the position characteristic parameter of the i-th camera, P S，i is the shape parameter of the target in the i-th camera.

[0070] It should be noted that the ViT network is a transformer model directly applied to a sequence of image blocks, which can obtain better image classification results on large-scale data sets. The present invention encodes the view of each camera based on the ViT network; in the ViT network, we need to project the image blocks to obtain image block embeddings, and preset an embedding of a learnable task for the image block embedding sequence. In addition, the classification head uses MLP as the classification head to process the image. In addition, the position embedding is added to the image block embedding to retain the spatial position information between the input image blocks. This provides accurate position encoding to obtain better three-dimensional reconstruction effects.

[0071] In a possible implementation, the encoding parameters of the N views are weighted to obtain a set of normalized encoder parameters P of the N views = (P1, P2, P c ,P S )include:

[0072] Use MLP1 to learn task tokens Tt1 and camera parameters P c Get output score

[0073] S1 i =MLP1(Tt1,P c，i ),

[0074] Then the global target feature P1 of multiple views is fused as:

[0075] Use MLP2 to pass local target feature token T 2，i and camera parameters P c Get output score: S2 i =MLP2(T 2，i ,P c，i ), then the local target features P2 of multiple views are fused as:

[0076] Among them, the local target feature token T 2，i It is obtained through the prediction cropping module, MLP1 and MLP2 are network layers of the multi-layer perceptron, and N is the number of cameras.

[0077] It should be noted that when fusing N views, we adjust the weight of the global target through the MLP layer, so that more and more correct information can be selected for representation. When decoding according to the final fused encoder parameters, the 3D reconstruction effect obtained is better.

[0078] It should be noted that the decoder may adopt any existing known decoding method.

[0079] In a possible implementation, after acquiring different views captured by N cameras at the same time and before processing the views, the method further includes:

[0080] Get the image parameters of the pictures of N cameras, including the average brightness value, median brightness value, and average saturation value;

[0081] Inputting image parameters of the pictures of the N cameras into a preset adjustment model, and adjusting the image parameters of the pictures of the N cameras based on an output of the adjustment model;

[0082] Among them, the preset adjustment model is obtained based on reinforcement learning; the target reward function of reinforcement learning is constructed based on the accuracy of target recognition in image processing.

[0083] It should be noted that the N cameras have different parameters for the pictures taken due to different positions and lighting conditions. In order to obtain a better fusion effect, the image parameters are adjusted before the view is processed. The adjustment model is used to obtain the final parameters of the N views, and the adjustment of each picture is set according to the final parameters. Here, the average brightness value, median brightness, and average saturation value are selected as adjustment indicators, and the reinforcement learning method is used to select the adjustment model that best suits the current situation. This method adaptively adjusts the image condition according to the actual situation and makes targeted changes to the picture.

[0084] Those skilled in the art may also select other algorithms that can achieve better adjustment effects.

[0085] The embodiment of the present invention also provides a computer storage medium (Memory), which is a memory device in an electronic device for storing programs and data. It is understandable that the computer storage medium here can include both built-in storage media in the electronic device and, of course, extended storage media supported by the electronic device. The computer storage medium provides a storage space, which stores the operating system of the electronic device. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer storage medium here can be a high-speed RAM memory, or a non-volatile memory (non-volatile memory), such as at least one disk storage; optionally, it can also be at least one computer storage medium located away from the aforementioned processor.

[0086] Figure 33 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The device comprises: at least one processor 301, such as a central processing unit (CPU), at least one memory 302, and at least one bus 303.

[0087] The memory 302 may store program instructions, and the processor 301 may be used to call the program instructions to execute a panoramic parking digital camera access system and method.

[0088] A person skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electrically-erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM), a solid state disk (SSD) or other optical disk storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0089] It should be noted that all steps of the embodiments of the present application are performed under legal and compliant conditions, that is, all steps of the embodiments of the present application are performed with authorization.

[0090] The above disclosure is only the preferred embodiment of the present invention, which certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope of the present invention.

Claims

1. A panoramic parking digital camera access system, characterized in that: The system includes a baseboard, a core board and N cameras, wherein the core board is provided with a central processing unit, a video processor and N ipex connection ports, where N≥2 and N is a positive integer; The N-way cameras are connected to the N-way ipex connection ports through N-way interfaces; the N-way ipex connection ports are communicatively connected to the video processor, and the video processor is communicatively connected to the central processing unit; When the N-way cameras are connected to the N-way ipex connection ports through the N-way interfaces, the N-way cameras acquire image data, transmit the image data to the video processor through the N-way interfaces and the N-way ipex connection ports, the video processor processes the image data, transmits the processed image data to the central processing unit, and the central processing unit executes a panoramic parking algorithm based on the processed image data, and outputs a display screen corresponding to the panoramic parking algorithm; The N-way ipex connection port is an N-way IPEX patch seat, each of the N-way interfaces includes: an ipex line terminal, a camera coaxial transmission line and a FAKRA connector, and the N-way camera is connected to the N-way ipex connection port through the N-way interface, including: The N cameras are respectively connected to the N FAKRA connectors, the N FAKRA connectors are respectively connected to the first ends of the N camera coaxial transmission lines, the second ends of the N camera coaxial transmission lines are respectively connected to the N ipex line ends, and the N ipex line ends are respectively connected to the N IPEX patch seats; The base plate includes an LVDS video signal output socket; the base plate is connected to the core board through a first row of pin connectors and a second row of pin connectors, and the central processor executes a panoramic parking algorithm based on the processed image data and outputs a display screen corresponding to the panoramic parking algorithm, including: The central processing unit executes the panoramic parking algorithm to obtain video output data; The central processor transmits the video output data to the bottom board through the first row of pin connectors and the second row of pin connectors; The backplane performs video encoding on the video output data to obtain the display picture in a target format; The base plate transmits the display image to the vehicle display screen through the LVDS video signal output socket; The vehicle-mounted display screen displays the display screen.

2. A panoramic parking method, applied to the system of claim 1, characterized in that: The method comprises: Step 1: Obtain different views taken by N cameras at the same time; Step 2: For each view obtained by different cameras, use the encoder to obtain a set of encoding parameters Pi = (P 1，i , P 2，i ,P c，i ,P S，i ), where P 1，i , P 2，i are the feature parameters of the global target feature and the local target feature of the i-th camera, respectively, c，i is the position characteristic parameter of the i-th camera, P S，i is the shape parameter of the target in the i-th camera; Step 3: Weight the encoding parameters of the N views to obtain a set of normalized encoder parameters P = (P1, P2, P c ,P S ); Step 4: Set the parameter P = (P1, P2, P c ,P S ) Input the decoder for three-dimensional reconstruction to obtain the obstacle target in the path; Step 5: Plan the parking path based on the obstacle detection results in the path.

3. The panoramic parking method according to claim 2, characterized in that: For each view obtained by different cameras, an encoder is used to obtain a set of encoding parameters Pi = (P 1，i , P 2，i ,P c，i ,P S，i ),include: For each view acquired by different cameras, split it into non-overlapping regions, add position embedding to the non-overlapping regions and linearly project them to the image Ti, concatenate the image Ti with the labeled token Tt of the learnable task, and encode the concatenated image into Ti1 and Tt1 using the ViT network; Use the pre-trained prediction cropping module to predict the global target features and local target features of the target in the Ti1 feature map; Use MLP to regress a set of encoder parameters Pi = (P 1，i , P 2，i ,P c，i ,P S，i ), where P 1,i ,P 2,i are the feature parameters of the global target feature and the local target feature of the i-th camera, respectively, which are obtained by an MLP regression module based on the global target feature and the local target feature of the target in the predicted Ti1 feature map; P c，i is the position characteristic parameter of the i-th camera, P S，i is the shape parameter of the target in the i-th camera.

4. The panoramic parking method according to claim 3, characterized in that: The encoding parameters of the N views are weighted to obtain a set of normalized encoder parameters P of the N views = (P1, P2, P c ,P S )include: Use MLP1 through the task token Tt1 and camera parameters P c Get output score S1 i =MLP1(Tt1,P c，i ), Then the global target feature P1 of multiple views is fused as: Use MLP2 to pass local target feature token T 2，i and camera parameters P c Get output score: S2 i =MLP2(T 2，i ,P c，i ), Then the local target features P2 of multiple views are fused as follows: Among them, the local target feature token T 2，i It is obtained through the prediction cropping module, MLP1 and MLP2 are network layers of the multi-layer perceptron, and N is the number of cameras.

5. The method according to claim 4, characterized in that After obtaining different views captured by N cameras at the same time and before step 2, the method further includes: Get the image parameters of the pictures of N cameras, including the average brightness value, median brightness value, and average saturation value; Inputting image parameters of the pictures of the N cameras into a preset adjustment model, and adjusting the image parameters of the pictures of the N cameras based on an output of the adjustment model; Among them, the preset adjustment model is obtained based on reinforcement learning; the target reward function of reinforcement learning is constructed based on the accuracy of target recognition in image processing.

6. A panoramic parking device, wherein the panoramic parking device is equipped with a panoramic parking digital camera access system, and the panoramic parking device is used to execute a panoramic parking method as claimed in claim 2, characterized in that: The method comprises: Step 1: Obtain different views taken by N cameras at the same time; Step 2: For each view obtained by different cameras, use the encoder to obtain a set of encoding parameters Pi = (P 1，i , P 2，i ,P c，i ,P S，i ), where P 1，i , P 2，i are the feature parameters of the global target feature and the local target feature of the i-th camera, respectively, c，i is the position characteristic parameter of the i-th camera, P S，i is the shape parameter of the target in the i-th camera; Step 3: Weight the encoding parameters of the N views to obtain a set of normalized encoder parameters P = (P1, P2, P c ,P S ); Step 4: Set the parameter P = (P1, P2, P c ,P S ) Input the decoder for three-dimensional reconstruction to obtain the obstacle target in the path; Step 5: Plan the parking path based on the obstacle detection results in the path.

7. An electronic device, characterized in that: include: a processor adapted to implement one or more instructions; as well as, A computer storage medium storing one or more instructions, wherein the one or more instructions are suitable for being loaded by the processor and executing the panoramic parking method as described in any one of claims 2 to 5.

8. A computer storage medium, characterized in that: The computer storage medium stores one or more instructions, and the one or more instructions are suitable for being loaded by a processor and executing the panoramic parking method as described in any one of claims 2-5.

Images