Sea area image data set intelligent processing system based on CNN recognition

Abstract

The invention relates to a sea area image data set intelligent processing system based on CNN identification, and provides a scheme for deploying a sea area image processing task on a sea application system, and a sea area image CNN identification model constructed based on CNN training is deployed in a sea area base station system of a sea area image acquisition device. Therefore, the sea area image collected by the device is subjected to image recognition processing in advance, the sea area image meeting the requirement of the ashore server is found out and then uploaded for application, the uploading difficulty and cost of the sea area image set are reduced, and the consumption of network resources is reduced. Besides, the processing and application difficulty of the shore server on the sea image can be reduced, the shore server can obtain the sea area image with the image quality meeting the requirement, and the sea area image is further used for sea area research, so that the image processing pressure of a shore service area is reduced, shore computing resources are saved, and the process and efficiency of the shore server on the sea area research are improved.

Description

Technical Field

[0001] The present disclosure relates to the technical field of image recognition, and in particular to a system, method and electronic device for intelligently processing a sea area image data set based on CNN recognition. Background Art

[0002] Marine big data contains a relatively rich dataset of marine application images. Usually these marine images are collected by marine image acquisition devices (or systems). Marine image acquisition devices are mainly composed of the following parts:

[0003] Ship Surround View Imaging System

[0004] Composition: Install multiple high-definition cameras at key locations on the ship and use image processing technology to stitch them into 360-degree panoramic images.

[0005] Subsea Imaging Systems:

[0006] Composition: Mainly includes optical system and mechanical structure system, designed to be light and compact.

[0007] Seabed Observation Network:

[0008] Composition: It includes a multi-level structure including shore base station, submarine main base station, etc., which are connected by optoelectronic composite communication submarine cables.

[0009] Other systems:

[0010] For example, the combination of digital side-scan sonar and shallow-sediment profiler provides true-scale seabed image data; the frogman handheld reconnaissance system is used by divers to navigate and search in low-visibility conditions.

[0011] In the traditional sea area image acquisition system, sea area images are mainly collected through ship surround image systems, seabed imaging systems, etc., and uploaded to shore servers through shore base stations and seabed main base stations, so that shore users can provide sea area image sampling information and further analyze and study the sea area information based on the sampled images.

[0012] Therefore, for the image processing of sea area images, as shown in the attached Figure 1 As shown in the figure, image processing is mainly performed on the onshore server (there are a large number of low-quality images with unclear pixel values, dark and blurred images in the image collection, which require a lot of energy, time and cost to process in the early stage). The massive sea area images are concentrated on the onshore server for processing, which will inevitably cause the operating pressure of the onshore server to be overloaded, increase the image processing pressure of the onshore server, and occupy the computing resources of the onshore server for processing tasks such as sea area data warehouse and sea area model construction. Summary of the invention

[0013] In order to solve the above problems, the present application proposes a system, method and electronic device for intelligent processing of sea area image datasets based on CNN recognition.

[0014] On the one hand, the present application proposes an intelligent processing system for a sea area image dataset based on CNN recognition, including a sea area image acquisition device and an onshore server, wherein:

[0015] The sea area image acquisition device comprises:

[0016] The sea area image acquisition system is used to collect and transmit the sea area image data set to the sea area base station system;

[0017] The sea area base station system is used to traverse and identify the sea area image features of each sea area image in the sea area image dataset based on the pre-deployed sea area image CNN recognition model, and find the sea area image that meets the current sea area image sampling condition according to the sea area image features and upload it to the onshore server;

[0018] The onshore server is used to configure the sea area image sampling conditions for the sea area base station system for collecting the sea area image this time; and receive and process the sea area image uploaded by the sea area base station system;

[0019] The sea area image acquisition device and the onshore server are communicatively connected via a seabed main base station in the sea area base station system.

[0020] As an optional implementation scheme of the present application, optionally, the sea area base station system includes:

[0021] Controller, used for logic control and calculation;

[0022] An image parsing module, used for parsing the sea area image data set to obtain each sea area image in the sea area image data set;

[0023] A buffer, used for caching the sea area image data set and each of the sea area images obtained by parsing;

[0024] A configuration module, used for pre-configuring and storing the sea area image sampling conditions;

[0025] A sea area image recognition module is used to pre-deploy the sea area image CNN recognition model; and use the sea area image CNN recognition model to traverse and identify each sea area image in the sea area image data set in the buffer, identify its sea area image features, and find the sea area image that meets the current sea area image sampling conditions according to the sea area image features; and output the sea area image that meets the current sea area image sampling conditions to the seabed main base station;

[0026] The seabed main base station is used for data communication, including: receiving the sea area image data set collected by the sea area image collection system; and uploading the sea area image that meets the current sea area image sampling conditions to the onshore server;

[0027] Main power supply, for power supply;

[0028] The image analysis module, the buffer, the configuration module, the sea area image recognition module, the seabed main base station and the main power supply are electrically connected to the controller respectively.

[0029] As an optional implementation scheme of the present application, optionally, the method for generating the sea area image CNN recognition model includes:

[0030] Collect several historical sea area images obtained from historical sampling missions;

[0031] Performing image preprocessing on the historical sea area images to construct a sea area image training set;

[0032] Performing feature engineering on the sea area image training set to extract feature information of each of the historical sea area images in the sea area image training set includes:

[0033] Based on a convolutional neural network, image features on the historical sea area image are extracted, including spatial features, spectral features, and temporal features;

[0034] Based on image detection technology, extracting ROI mark detection features marked by the historical sea area image when it is executed, including ROI contour, SIFT features and SURF features of the ROI area;

[0035] Constructing a feature set based on the extracted feature information of each of the historical sea area images;

[0036] Dividing the feature set into a training set and a validation set according to a preset ratio;

[0037] Inputting the training set into a preset CNN model to perform model feature learning, and training and generating the sea area image CNN recognition model;

[0038] The recognition performance of the sea area image CNN recognition model is verified using the verification set:

[0039] If the verification is qualified, the model file of the sea area image CNN recognition model is sent to the sea area base station system through the onshore server, and the sea area base station system loads and configures the model file in the sea area image recognition module;

[0040] If the verification fails, repeat the above steps.

[0041] As an optional implementation scheme of the present application, optionally, the onshore server is also used for:

[0042] According to the preset sampling frequency, a sampling instruction is issued to the sea area base station system;

[0043] as well as,

[0044] Receive the message data fed back by the sea area base station system and calculate the packet loss rate:

[0045] Packet loss rate = [(number of input messages - number of output messages) / number of input messages] * 100%;

[0046] When the packet loss rate is lower than a preset value, a pause message instruction is sent to the sea area base station system.

[0047] As an optional implementation scheme of the present application, optionally, the sea area base station system is also used for:

[0048] The controller executes the pause message instruction to suspend image sampling and image feature recognition work.

[0049] On the other hand, the present application proposes a method for intelligently processing a sea area image dataset based on CNN recognition, which is implemented based on the intelligent processing system for sea area image dataset based on CNN recognition, and includes the following steps:

[0050] The onshore server sends a sampling instruction to the sea area image acquisition device 5-30 minutes in advance and configures the sea area image sampling conditions for this sea area image acquisition;

[0051] The sea area image acquisition device executes the sampling instruction and performs image acquisition in advance, including:

[0052] The sea area image acquisition system collects and transmits the sea area image data set to the sea area base station system;

[0053] The sea area base station system, based on the pre-deployed sea area image CNN recognition model, traverses and identifies the sea area image features of each sea area image in the sea area image dataset, and finds the sea area image that meets the current sea area image sampling conditions according to the sea area image features and uploads it to the onshore server;

[0054] The onshore server receives and processes the sea area images uploaded by the sea area base station system, and performs post-processing of the sea area images.

[0055] As an optional implementation scheme of the present application, optionally, based on a pre-deployed sea area image CNN recognition model, traversing and identifying sea area image features of each sea area image in the sea area image dataset, and finding the sea area image that meets the current sea area image sampling condition according to the sea area image features, including:

[0056] Through the sea area image recognition module, each sea area image cached in the buffer is traversed, and based on the pre-deployed sea area image CNN recognition model, sea area image features in the sea area image are identified and extracted: image features and ROI mark detection features;

[0057] Reading the sea area image sampling conditions pre-configured and stored in the configuration module, and finding the sea area image that meets the current sea area image sampling conditions according to the sea area image features;

[0058] The sea area image that meets the current sea area image sampling conditions is output to the seabed main base station, and the seabed main base station uploads the sea area image that meets the current sea area image sampling conditions to the onshore server.

[0059] In another aspect, the present application further provides an electronic device, comprising:

[0060] processor;

[0061] a memory for storing processor-executable instructions;

[0062] Wherein, the processor is configured to implement the intelligent processing method of sea area image data set based on CNN recognition when executing the executable instructions.

[0063] Technical effects of the present invention:

[0064] The present invention proposes a solution for deploying sea area image processing tasks on a marine application system. A CNN recognition model of sea area images constructed based on CNN training is deployed in a sea area base station system of a sea area image acquisition device, so that image recognition processing is performed on the sea area images collected by the device in advance, and sea area images that meet the requirements of the onshore server are found and then uploaded to the application, thereby reducing the difficulty and cost of uploading sea area image sets and reducing the consumption of network resources.

[0065] In addition, it can reduce the difficulty of processing and applying ocean images by onshore servers, allowing onshore servers to obtain sea area images with image quality that meets the requirements and further use them for sea area research, thereby reducing the image processing pressure in onshore service areas, saving onshore computing resources, and improving the progress and efficiency of onshore servers in sea area research.

[0066] Further features and aspects of the present disclosure will become apparent from the following detailed description of exemplary embodiments with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate exemplary embodiments, features, and aspects of the disclosure and, together with the description, serve to explain the principles of the disclosure.

[0068] Figure 1 The figure shows a schematic diagram of the pre-processing of traditional sea area images;

[0069] Figure 2 It is a schematic diagram of the pre-processing of the sea area image of the present invention;

[0070] Figure 3 It is a schematic diagram showing the composition structure of the sea area base station system of the present invention;

[0071] Figure 4 A schematic diagram showing the composition structure of a submarine main base station used in the present invention is shown;

[0072] Figure 5 The figure shows a training process diagram of the CNN recognition model for sea area images of the present invention;

[0073] Figure 6 It is a schematic diagram showing the application of the electronic device of the present invention. DETAILED DESCRIPTION

[0074] Various exemplary embodiments, features and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0075] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0076] In addition, in order to better illustrate the present disclosure, numerous specific details are given in the following specific embodiments. It should be understood by those skilled in the art that the present disclosure can also be implemented without certain specific details. In some examples, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the main purpose of the present disclosure.

[0077] Example 1

[0078] like Figure 2 As shown, on one hand, the present application proposes an intelligent processing system for a sea area image data set based on CNN recognition, including a sea area image acquisition device and an onshore server, wherein:

[0079] The sea area image acquisition device comprises:

[0080] The sea area image acquisition system is used to collect and transmit the sea area image data set to the sea area base station system;

[0081] The sea area base station system is used to traverse and identify the sea area image features of each sea area image in the sea area image dataset based on the pre-deployed sea area image CNN recognition model, and find the sea area image that meets the current sea area image sampling condition according to the sea area image features and upload it to the onshore server;

[0082] The onshore server is used to configure the sea area image sampling conditions for the sea area base station system for collecting the sea area image this time; and receive and process the sea area image uploaded by the sea area base station system;

[0083] The sea area image acquisition device and the onshore server are communicatively connected via a seabed main base station in the sea area base station system.

[0084] Combination Figure 1 and Figure 2 As shown, the present invention moves the sea area image pre-processing work originally done by the onshore server to the image acquisition end, thereby reducing the image processing pressure of the onshore server, avoiding operation overload, increasing the image processing pressure of the onshore server, and does not occupy the computing resources of other tasks such as the sea area data warehouse.

[0085] Sea area image acquisition systems, such as ship surround image systems, are mainly used for cameras used for sea area photography, etc. Image data can be transmitted through wired, wireless or optical cable devices.

[0086] The sea area base station system is mainly a communication system for data communication between the sea and the shore, mainly including the seabed main base station and other application systems. The present invention adds a sea area image CNN recognition model to the sea area base station system to pre-process the collected sea area images and then upload them to the shore server through the seabed main base station.

[0087] The data communication between the submarine main base station and the onshore server should be understood in combination with the technical principles of existing base station communication. Figure 4 The existing base station system is shown and will not be described in detail here.

[0088] The sea area image has the following image features and detection features for saliency detection (generally the features for constructing ROI areas and ROI marker detection features):

[0089] 1. Image features:

[0090] Spatial characteristics: Display the geometric outline, coordinates, area and other information of objects in space.

[0091] Spectral characteristics: represented by the grayscale value and RGB value of pixels, which are closely related to the reflectivity of objects to sunlight radiation.

[0092] Temporal characteristics: Due to changes over time, such as the ebb and flow of tides, remote sensing images will differ at different time points in the same area.

[0093] It mainly identifies the format of sea area images, and identifies the spatial features, grayscale values, and image feature information at the current time on the sea area images.

[0094] 2.ROI marking detection features:

[0095] ROI contour: By constructing the ROI area, the ROI area contour of the detected sea area image is identified; the addition of the ROI area can be constructed by the administrator, and specifically, when the historical task is executed, the technician can construct the ROI area of ​​interest on the sea area image. The ROI contour can be extracted by the image contour recognition algorithm.

[0096] SIFT features in the ROI area: remain invariant to rotation, scaling, and brightness changes, and are suitable for fast and accurate matching of massive feature libraries.

[0097] SURF features in the ROI area: They are scale invariant and highly robust to changes in illumination, affine, and perspective, thus improving feature extraction efficiency.

[0098] Salient region detection: Combine image enhancement and superpixel segmentation algorithms to obtain uniform and clear salient maps for specific target detection.

[0099] The main purpose is to identify the features of the region of interest on the sea image, so as to obtain the sea image required for this task. When the model recognizes that the current sea image has the learned ROI mark detection features, the current sea image is used as the target image that meets the requirements of this task and uploaded to the onshore server.

[0100] Therefore, the model recognition based on image features and ROI marker detection features can find the target image required for this task. These features and quantitative detection methods play an important role in applications such as marine image analysis and target recognition, and can be used for the research and application of specific marine tasks, such as marine data warehouses and marine map construction.

[0101] like Figure 3 As shown, as an optional implementation scheme of the present application, optionally, the sea area base station system includes:

[0102] Controller, used for logic control and calculation;

[0103] An image parsing module, used for parsing the sea area image data set to obtain each sea area image in the sea area image data set;

[0104] A buffer, used for caching the sea area image data set and each of the sea area images obtained by parsing;

[0105] A configuration module, used for pre-configuring and storing the sea area image sampling conditions;

[0106] A sea area image recognition module is used to pre-deploy the sea area image CNN recognition model; and use the sea area image CNN recognition model to traverse and identify each sea area image in the sea area image data set in the buffer, identify its sea area image features, and find the sea area image that meets the current sea area image sampling conditions according to the sea area image features; and output the sea area image that meets the current sea area image sampling conditions to the seabed main base station;

[0107] The seabed main base station is used for data communication, including: receiving the sea area image data set collected by the sea area image collection system; and uploading the sea area image that meets the current sea area image sampling conditions to the onshore server;

[0108] Main power supply, for power supply;

[0109] The image analysis module, the buffer, the configuration module, the sea area image recognition module, the seabed main base station and the main power supply are electrically connected to the controller respectively.

[0110] In the present invention, in the sea area base station system, in addition to the original seabed main base station and main power supply, a corresponding image analysis module, a buffer, a configuration module and a sea area image recognition module are also configured. The main power supply is used for power supply, and can also supply power to the power supply in the seabed main base station. The installation and deployment of the sea area base station can refer to the construction plan of the existing sea base station, and is not limited here.

[0111] In the present invention, the controller is a single chip microcomputer or a PLC controller, or an MCU control chip, etc. It can perform logic control on the sea area base station system and execute relevant instructions issued by the shore server.

[0112] After the sea image acquisition system acquires the sea image data set, it can be transmitted to the sea base station system. The system receives the image through the port and caches it in the buffer. The image parsing module can perform image parsing to obtain the corresponding sea image. The subsequent sea image recognition module can use the pre-deployed sea image CNN recognition model to traverse the cached sea images, identify the sea image features in each sea image through the model, and read the pre-configured sea image sampling conditions from the configuration module to determine which sea image features meet the sea image sampling conditions and send the sea image corresponding to the sea image features that meet the conditions to the seabed main base station, and then the seabed main base station uploads the qualified sea image to the shore server. In the configuration module, the pre-configured and stored sea image sampling conditions are mainly configured in advance by the shore server, and the image requirements and feature requirements, formats, etc. of the sea images required for this task are configured. The various feature requirements configured in the specific sea image sampling conditions can be used by the administrator to limit the sea image features on the sea image according to this task. For example, the grayscale value of the sea image, the ROI area and the SIFT features of the ROI area, etc., can be used to limit the corresponding feature values ​​and feature conditions.

[0113] The feature recognition and extraction of the CNN recognition model for sea area images can be understood in conjunction with the training of the CNN recognition model for sea area images below.

[0114] like Figure 4 The following is a schematic diagram of the control system composition of the submarine main base station. The submarine main base station can be understood by referring to the composition and principle of existing base stations, and will not be described in detail here.

[0115] like Figure 5 As shown, as an optional implementation scheme of the present application, optionally, the method for generating the sea area image CNN recognition model includes:

[0116] Collect several historical sea area images obtained from historical sampling missions;

[0117] Performing image preprocessing on the historical sea area images to construct a sea area image training set;

[0118] Performing feature engineering on the sea area image training set to extract feature information of each of the historical sea area images in the sea area image training set includes:

[0119] Based on a convolutional neural network, image features on the historical sea area image are extracted, including spatial features, spectral features, and temporal features;

[0120] Based on image detection technology, extracting ROI mark detection features marked by the historical sea area image when it is executed, including ROI contour, SIFT features and SURF features of the ROI area;

[0121] Constructing a feature set based on the extracted feature information of each of the historical sea area images;

[0122] Dividing the feature set into a training set and a validation set according to a preset ratio;

[0123] Inputting the training set into a preset CNN model to perform model feature learning, and training and generating the sea area image CNN recognition model;

[0124] The recognition performance of the sea area image CNN recognition model is verified using the verification set:

[0125] If the verification is qualified, the model file of the sea area image CNN recognition model is sent to the sea area base station system through the onshore server, and the sea area base station system loads and configures the model file in the sea area image recognition module;

[0126] If the verification fails, repeat the above steps.

[0127] The sea area image has the following image features and detection features for saliency detection (generally the features for constructing ROI areas and ROI marker detection features):

[0128] 1. Image features:

[0129] Spatial characteristics: Display the geometric outline, coordinates, area and other information of objects in space.

[0130] Spectral characteristics: represented by the grayscale value and RGB value of pixels, which are closely related to the reflectivity of objects to sunlight radiation.

[0131] Temporal characteristics: Due to changes over time, such as the ebb and flow of tides, remote sensing images will differ at different time points in the same area.

[0132] 2.ROI marking detection features:

[0133] ROI contour: By constructing the ROI area, the ROI area contour of the detected sea area image is identified; the addition of the ROI area can be constructed by the administrator, and specifically, when the historical task is executed, the technician can construct the ROI area of ​​interest on the sea area image. The ROI contour can be extracted by the image contour recognition algorithm.

[0134] SIFT features in the ROI area: remain invariant to rotation, scaling, and brightness changes, and are suitable for fast and accurate matching of massive feature libraries.

[0135] SURF features in the ROI area: They are scale invariant and highly robust to changes in illumination, affine, and perspective, thus improving feature extraction efficiency.

[0136] Salient region detection: Combine image enhancement and superpixel segmentation algorithms to obtain uniform and clear salient maps for specific target detection.

[0137] These features and quantitative detection methods play an important role in applications such as marine image analysis and target recognition.

[0138] Image features on historical sea area images, such as grayscale values, spatial coordinates, spatial area or. Image feature values ​​corresponding to different time nodes, such as the color of remote sensing images, etc. These features can be extracted through CNN. In previous image processing work, technicians can mark the ROI area on the sea area image. The extracted image features can also extract the image features in the marked ROI area. The ROI area can identify the ROI contour through the image contour recognition algorithm and further extract the SI FT and other features of the ROI area.

[0139] Using image detection technology, the contour of ROI (region of interest) in the image and the SIFT features of the region can be detected and extracted by the following steps:

[0140] The ROI area contour detection can refer to the following steps:

[0141] Image preprocessing: Convert the image to grayscale and apply Gaussian blur to remove noise.

[0142] Contour extraction: Use the Canny edge detection algorithm to obtain the image edge, and then extract the contour through the contour detection function.

[0143] ROI determination: The position and size of the ROI area are determined based on the bounding box of the contour.

[0144] The SIFT feature extraction can refer to the following steps:

[0145] Construct scale space: Perform Gaussian convolution on the ROI area image to establish the scale space.

[0146] Detect extreme points: Detect extreme points in scale space to obtain scale invariance.

[0147] Feature point positioning: Accurately locate extreme points and eliminate unstable feature points.

[0148] Feature description: Extract feature descriptors at feature points and generate 128-dimensional feature vectors.

[0149] The implementation of specific feature engineering can be completed by the administrator, and the optimization iterative training of the corresponding model can be completed.

[0150] The verification methods of model prediction and recognition performance include cross-validation, leave-one-out validation, bootstrapping, ROC curve and AUC value, confusion matrix, accuracy, precision, recall and F1 score, etc. The administrator will use the validation set for verification.

[0151] After the model is constructed, the model file is sent to the sea area base station system through the onshore server, and the sea area base station system loads the model file and configures it in the sea area image recognition module.

[0152] In order to avoid image transmission errors or inefficiencies caused by network quality, the present invention adopts a technical solution of onshore packet loss rate detection. When the network quality is found to be low, the onshore server automatically stops the collection of sea area images.

[0153] As an optional implementation scheme of the present application, optionally, the onshore server is also used for:

[0154] According to the preset sampling frequency, a sampling instruction is issued to the sea area base station system;

[0155] as well as,

[0156] Receive the message data fed back by the sea area base station system and calculate the packet loss rate:

[0157] Packet loss rate = [(number of input messages - number of output messages) / number of input messages] * 100%;

[0158] When the packet loss rate is lower than a preset value, a pause message instruction is sent to the sea area base station system.

[0159] As an optional implementation scheme of the present application, optionally, the sea area base station system is also used for:

[0160] The controller executes the pause message instruction to suspend image sampling and image feature recognition work.

[0161] Packet loss rate definition: Packet loss rate refers to the ratio of the number of data packets lost in the test to the number of data packets sent. Calculation formula: Packet loss rate = [(number of input packets - number of output packets) / number of input packets] * 100%.

[0162] Therefore, the server onshore can send messages to the sea area base station system at regular intervals according to the sampling frequency (system default message data), and calculate the packet loss rate based on the feedback message information. Calculation steps:

[0163] Count the number of input messages: record the total number of input messages received by the server within the sampling time.

[0164] Count the number of output messages: record the total number of output messages successfully sent by the server within the same sampling time.

[0165] Calculate the difference: Subtract the number of output packets from the number of input packets to get the number of lost packets.

[0166] Calculate the ratio: Divide the number of lost packets by the number of input packets, and multiply by 100% to get the packet loss ratio.

[0167] Note: The packet loss rate is related to the length of the data packet and the frequency of packet sending, and is usually tested within the throughput range.

[0168] Through the above steps, the server can accurately calculate the data packet loss rate, thereby evaluating the network performance and taking corresponding optimization measures.

[0169] Obviously, those skilled in the art should understand that the implementation of all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above-mentioned embodiments. Those skilled in the art can understand that the implementation of all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above-mentioned embodiments. Among them, the storage medium can be a disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory (Flash Memory), a hard disk (Hard Disk Drive, abbreviated as: HDD) or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above-mentioned types of memory.

[0170] Example 2

[0171] Based on the implementation principle of Example 1, on the other hand, the present application proposes a method for intelligently processing a sea area image dataset based on CNN recognition, which is implemented based on the intelligent processing system for sea area image dataset based on CNN recognition, and includes the following steps:

[0172] The onshore server sends a sampling instruction to the sea area image acquisition device 5-30 minutes in advance and configures the sea area image sampling conditions for this sea area image acquisition;

[0173] The sea area image acquisition device executes the sampling instruction and performs image acquisition in advance, including:

[0174] The sea area image acquisition system collects and transmits the sea area image data set to the sea area base station system;

[0175] The sea area base station system, based on the pre-deployed sea area image CNN recognition model, traverses and identifies the sea area image features of each sea area image in the sea area image dataset, and finds the sea area image that meets the current sea area image sampling conditions according to the sea area image features and uploads it to the onshore server;

[0176] The onshore server receives and processes the sea area images uploaded by the sea area base station system, and performs post-processing of the sea area images.

[0177] As an optional implementation scheme of the present application, optionally, based on a pre-deployed sea area image CNN recognition model, traversing and identifying sea area image features of each sea area image in the sea area image dataset, and finding the sea area image that meets the current sea area image sampling condition according to the sea area image features, including:

[0178] Through the sea area image recognition module, each sea area image cached in the buffer is traversed, and based on the pre-deployed sea area image CNN recognition model, sea area image features in the sea area image are identified and extracted: image features and ROI mark detection features;

[0179] Reading the sea area image sampling conditions pre-configured and stored in the configuration module, and finding the sea area image that meets the current sea area image sampling conditions according to the sea area image features;

[0180] The sea area image that meets the current sea area image sampling conditions is output to the seabed main base station, and the seabed main base station uploads the sea area image that meets the current sea area image sampling conditions to the onshore server.

[0181] Please understand the above method steps in conjunction with the system interaction description and implementation principles of Example 1, which will not be repeated in this example.

[0182] The modules or steps of the present invention described above can be implemented by a general-purpose computing system, they can be concentrated on a single computing system, or distributed on a network composed of multiple computing systems, and optionally, they can be implemented by a program code executable by a computing system, so that they can be stored in a storage system and executed by the computing system, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0183] Example 3

[0184] like Figure 6 As shown, further, in another aspect, the present application also proposes an electronic device, including:

[0185] processor;

[0186] a memory for storing processor-executable instructions;

[0187] Among them, the processor is configured to implement an intelligent processing system for sea area image data set based on CNN recognition as described in Example 2 when executing the executable instructions.

[0188] The electronic device of the embodiment of the present disclosure includes a processor and a memory for storing processor executable instructions. The processor is configured to implement the intelligent processing system of the sea image data set based on CNN recognition described in the above embodiment 2 when executing the executable instructions.

[0189] Here, it should be noted that the number of processors can be one or more. At the same time, the electronic device of the embodiment of the present disclosure may also include an input system and an output system. Among them, the processor, memory, input system and output system may be connected through a bus or in other ways, which are not specifically limited here.

[0190] As a computer-readable storage medium, the memory can be used to store software programs, computer executable programs and various modules, such as the program or module corresponding to the intelligent processing system of the sea image dataset based on CNN recognition in the embodiment of the present disclosure. The processor executes various functional applications and data processing of the electronic device by running the software programs or modules stored in the memory.

[0191] The input system can be used to receive input numbers or signals. The signal can be a key signal related to user settings and function control of the device / terminal / server. The output system can include display devices such as display screens.

[0192] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. An intelligent processing system for sea area image dataset based on CNN recognition, characterized in that: It includes a sea area image acquisition device and an onshore server, wherein: The sea area image acquisition device comprises: The sea area image acquisition system is used to collect and transmit the sea area image data set to the sea area base station system; The sea area base station system is used to traverse and identify the sea area image features of each sea area image in the sea area image dataset based on the pre-deployed sea area image CNN recognition model, and find the sea area image that meets the current sea area image sampling condition according to the sea area image features and upload it to the onshore server; The onshore server is used to configure the sea area image sampling conditions for the sea area base station system for collecting the sea area image this time; and receive and process the sea area image uploaded by the sea area base station system; The sea area image acquisition device and the onshore server are communicatively connected via a seabed main base station in the sea area base station system.

2. According to claim 1, the intelligent processing system of sea area image data set based on CNN recognition is characterized in that: The sea area base station system comprises: Controller, used for logic control and calculation; An image parsing module, used for parsing the sea area image data set to obtain each sea area image in the sea area image data set; A buffer, used for caching the sea area image data set and each of the sea area images obtained by parsing; A configuration module, used for pre-configuring and storing the sea area image sampling conditions; A sea area image recognition module is used to pre-deploy the sea area image CNN recognition model; and use the sea area image CNN recognition model to traverse and identify each sea area image in the sea area image data set in the buffer, identify its sea area image features, and find the sea area image that meets the current sea area image sampling conditions according to the sea area image features; and output the sea area image that meets the current sea area image sampling conditions to the seabed main base station; The seabed main base station is used for data communication, including: receiving the sea area image data set collected by the sea area image collection system; and uploading the sea area image that meets the current sea area image sampling conditions to the onshore server; Main power supply, for power supply; The image analysis module, the buffer, the configuration module, the sea area image recognition module, the seabed main base station and the main power supply are electrically connected to the controller respectively.

3. The intelligent processing system for sea area image data set based on CNN recognition according to claim 2 is characterized in that: The method for generating the sea area image CNN recognition model comprises: Collect several historical sea area images obtained from historical sampling missions; Performing image preprocessing on the historical sea area images to construct a sea area image training set; Performing feature engineering on the sea area image training set to extract feature information of each of the historical sea area images in the sea area image training set includes: Based on a convolutional neural network, image features on the historical sea area image are extracted, including spatial features, spectral features, and temporal features; Based on image detection technology, extracting ROI mark detection features marked by the historical sea area image when it is executed, including ROI contour, SIFT features and SURF features of the ROI area; Constructing a feature set based on the extracted feature information of each of the historical sea area images; Dividing the feature set into a training set and a validation set according to a preset ratio; Inputting the training set into a preset CNN model to perform model feature learning, and training and generating the sea area image CNN recognition model; The recognition performance of the sea area image CNN recognition model is verified using the verification set: If the verification is qualified, the model file of the sea area image CNN recognition model is sent to the sea area base station system through the onshore server, and the sea area base station system loads and configures the model file in the sea area image recognition module; If the verification fails, repeat the above steps.

4. The intelligent processing system for sea area image data set based on CNN recognition according to claim 1 is characterized in that: The onshore server is also used for: According to the preset sampling frequency, a sampling instruction is issued to the sea area base station system; as well as, Receive the message data fed back by the sea area base station system and calculate the packet loss rate: Packet loss rate = [(number of input messages - number of output messages) / number of input messages] * 100%; When the packet loss rate is lower than a preset value, a pause message instruction is sent to the sea area base station system.

5. The intelligent processing system for sea area image data set based on CNN recognition according to claim 1 is characterized in that: The sea area base station system is also used for: The controller executes the pause message instruction to suspend image sampling and image feature recognition work.

6. A method for intelligently processing a sea area image dataset based on CNN recognition, implemented based on a system for intelligently processing a sea area image dataset based on CNN recognition according to any one of claims 1 to 5, characterized in that: The steps include: The onshore server sends a sampling instruction to the sea area image acquisition device 5-30 minutes in advance and configures the sea area image sampling conditions for this sea area image acquisition; The sea area image acquisition device executes the sampling instruction and performs image acquisition in advance, including: The sea area image acquisition system collects and transmits the sea area image data set to the sea area base station system; The sea area base station system, based on the pre-deployed sea area image CNN recognition model, traverses and identifies the sea area image features of each sea area image in the sea area image dataset, and finds the sea area image that meets the current sea area image sampling conditions according to the sea area image features and uploads it to the onshore server; The onshore server receives and processes the sea area images uploaded by the sea area base station system, and performs post-processing of the sea area images.

7. The method for intelligently processing a sea area image dataset based on CNN recognition according to claim 6 is characterized in that: Based on the pre-deployed sea area image CNN recognition model, traverse and identify the sea area image features of each sea area image in the sea area image dataset, and find the sea area image that meets the current sea area image sampling condition according to the sea area image features, including: Through the sea area image recognition module, each sea area image cached in the buffer is traversed, and based on the pre-deployed sea area image CNN recognition model, sea area image features in the sea area image are identified and extracted: image features and ROI mark detection features; Reading the sea area image sampling conditions pre-configured and stored in the configuration module, and finding the sea area image that meets the current sea area image sampling conditions according to the sea area image features; The sea area image that meets the current sea area image sampling conditions is output to the seabed main base station, and the seabed main base station uploads the sea area image that meets the current sea area image sampling conditions to the onshore server.

8. An electronic device, characterized in that include: processor; a memory for storing processor-executable instructions; Wherein, the processor is configured to implement an intelligent processing method for sea area image data set based on CNN recognition as described in any one of claims 6-7 when executing the executable instructions.

Images