Image recognition and processing system and method based on deep learning

Abstract

The invention provides an image recognition and processing system and method based on deep learning, the system comprises an image acquisition module, a deep learning model module and an image processing module, the image acquisition module is used for capturing image data to be recognized and processed; the deep learning model module is used for performing feature extraction and recognition on the captured image data; the image processing module is used for performing corresponding processing on the image according to the identification result; and the deep learning model module and the image processing module are connected with each other through a data interface so as to realize data interaction and processing. According to the invention, the accuracy and processing efficiency of image recognition can be improved, and visual operation experience and result display are provided.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition and processing, and more specifically, to a system and method for image recognition and processing based on deep learning. Background Art

[0002] Existing image recognition and processing technologies have been widely used in many fields, such as autonomous driving, medical diagnosis, and security monitoring. These technologies usually rely on image acquisition devices to capture image data and analyze and process images through algorithms. However, traditional image processing methods have certain limitations in feature extraction and recognition, especially when dealing with complex scenes or requiring high-precision measurements, it is often difficult to achieve the desired effect. In addition, the friendliness and intuitiveness of the user interaction interface are also areas where existing technologies need to be improved so that users can more conveniently input instructions and view processing results.

[0003] In the process of implementing the embodiments of the present invention, the inventors found that there are at least the following problems or defects in the prior art: Although deep learning technology has made significant progress in the field of image recognition, it still faces challenges in data processing efficiency, accuracy, and user interaction experience in practical applications. In particular, when performing high-precision measurement of images, how to effectively remove outliers, optimize measurement results, and provide a smooth user experience are key issues that the prior art needs to solve. Summary of the invention

[0004] The present invention provides an image recognition and processing system and method based on deep learning.

[0005] In a first aspect of the present invention, there is provided an image recognition and processing system based on deep learning, comprising:

[0006] Image acquisition module, deep learning model module, image processing module, among which,

[0007] The image acquisition module is used to capture image data to be identified and processed; the deep learning model module is used to extract features and identify the captured image data; the image processing module is used to process the image accordingly according to the identification result;

[0008] The deep learning model module and the image processing module are interconnected through a data interface to achieve data interaction and processing.

[0009] Furthermore, the system also includes a user interaction interface, which is used to receive instructions and parameters input by the user and display processed image results.

[0010] Furthermore, when the system needs to perform distance measurement on an image, the user interaction interface obtains the measurement value by the following method, which specifically includes:

[0011] The user specifies the measurement area through an interactive interface;

[0012] The deep learning model module extracts features from the specified area and calculates preliminary measurements;

[0013] The image processing module performs optimization processing based on the preliminary measurement value to obtain the final measurement value;

[0014] The user interface displays the optimized measurement values ​​to the user in a graphical manner.

[0015] Furthermore, the image processing module optimizes the measurement value by the following method, which specifically includes:

[0016] A normal value sequence is obtained by removing outliers and data points exceeding a preset threshold from the preliminary measurement values;

[0017] Based on the optimal window value, the mean sequence is obtained by sliding the window for the normal value sequence;

[0018] Based on the optimal number of interpolation, interpolation is performed between two adjacent data points in the mean sequence to obtain the interpolated sequence;

[0019] The interpolated sequence is deduplicated and de-jittered to obtain the optimized measurement value.

[0020] Furthermore, when the system is first run or receives new image data, the optimal window value and the optimal number of interpolations are obtained by the following method, which specifically includes the following steps:

[0021] S1, removing outliers and data points exceeding a preset threshold in the preliminary measurement values ​​to obtain a normal value sequence;

[0022] S2, set the initial window value, the window value is smaller than the number of data points in the normal value sequence;

[0023] S3, based on the window value, the normal value sequence is calculated by sliding the window to obtain the mean value sequence;

[0024] S4, calculating the number of interpolation values ​​based on the window value, and performing interpolation processing on the mean sequence to obtain an interpolated sequence;

[0025] S5, performing deduplication and de-jittering processing on the interpolated sequence to obtain an optimized measurement value;

[0026] S6. Display the image and measurement results in real time according to the optimized measurement value to determine whether the display is smooth. If so, increase the window value and repeat S3-S6 until the display is not smooth; otherwise, reduce the window value and repeat S3-S6 until the display is smooth, and obtain the optimal window value and the optimal number of interpolations.

[0027] Furthermore, the number of interpolation values ​​is calculated based on the window value by the following formula:

[0028]

[0029] Where j is the number of interpolations, t is the time it takes for the system to process single-cycle image data, n is the window value, and N is the number of data points in the normal value sequence.

[0030] Furthermore, the mean value sequence is obtained by sliding the window of the normal value sequence through the following method, and the method specifically includes:

[0031] Set a sliding window for the normal value sequence. The number of data points covered by the sliding window is the same as the window value. The starting point of the sliding window coincides with the first data point of the normal value sequence. Calculate and save the mean of the corresponding values ​​of the data points covered by the sliding window.

[0032] The sliding window moves backwards, and each time a data point is moved, the mean of the corresponding values ​​of the data points covered by the sliding window is calculated and saved, until the end point of the sliding window coincides with the last data point of the normal value sequence, and the mean of the corresponding values ​​of the last data point covered by the sliding window is calculated and saved;

[0033] Arrange the obtained moving means in the order of sliding window movement to obtain the mean sequence.

[0034] Furthermore, interpolation is performed between two adjacent data points of the mean sequence, specifically, the number of data points inserted between two adjacent data points of the mean sequence is equal to the number of interpolation, and the inserted data points and the two adjacent data points form an arithmetic progression.

[0035] Furthermore, the interpolated sequence is de-jittered, specifically, it is determined whether the difference between the values ​​corresponding to each two adjacent data points in the interpolated sequence is less than a preset threshold. If so, only one of the two adjacent data points is removed, otherwise, both of the two adjacent data points are retained.

[0036] In a second aspect of the present invention, there is provided an image recognition and processing method based on the system according to any one of claims 1 to 9, comprising:

[0037] The system starts and completes initialization;

[0038] The user interaction interface receives user instructions and parameters;

[0039] The image acquisition module captures image data;

[0040] The deep learning model module extracts and recognizes features of image data;

[0041] According to the recognition result, the image processing module processes the image;

[0042] The user interface displays the processed image results.

[0043] According to the above-mentioned embodiment of the present invention, at least the following beneficial effects are achieved: the image recognition and processing system can significantly improve the extraction and recognition accuracy of image features by integrating a deep learning model, thereby maintaining high-accuracy image analysis results even in complex environments. The image processing module inside the system can further optimize the measurement values, and can generate smoother and more accurate measurement sequences by removing outliers and preset threshold filtering of data points, and using sliding windows and interpolation techniques. This method can not only improve the reliability of the data, but also enhance the robustness of the system. In addition, the user interaction interface included in the system design can provide intuitive operation guidance and result display, allowing users to easily specify the measurement area and obtain measurement results. By graphically displaying the optimized measurement values, users can more intuitively understand the effect of image processing, thereby making more accurate judgments and decisions in practical applications. This design of the system can not only improve the user experience, but also make the image recognition and processing process more efficient and convenient. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood by reading the detailed description below with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, in which:

[0045] Figure 1 A schematic diagram of the structure of an image recognition and processing system based on deep learning provided by an embodiment of the present invention;

[0046] Figure 2 A flowchart of an image recognition and processing method based on deep learning provided by one embodiment of the present invention;

[0047] Figure 3 The schematic diagram schematically shows the structure of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0048] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and implement the present invention, and are not intended to limit the scope of the present invention in any way. On the contrary, these embodiments are provided to make the present invention more thorough and complete, and to fully convey the scope of the present invention to those skilled in the art.

[0049] Those skilled in the art know that the embodiments of the present invention can be implemented as a system, device, apparatus, method or computer program product. Therefore, the present invention can be specifically implemented in the following forms, namely: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0050] It should be noted that any number of elements in the drawings is for illustration rather than limitation, and any naming is only for distinction and does not have any limiting meaning.

[0051] Reference below Figure 1 , Figure 1 The structure diagram of the image recognition and processing system based on deep learning provided by one embodiment of the present invention is as follows. Figure 1 As shown, a deep learning-based image recognition and processing system 100 includes:

[0052] An image acquisition module 101, a deep learning model module 102, and an image processing module 103, wherein the image acquisition module is used to capture image data to be identified and processed; the deep learning model module is used to extract features and identify the captured image data; and the image processing module is used to process the image accordingly according to the identification result;

[0053] The deep learning model module and the image processing module are interconnected through a data interface to achieve data interaction and processing.

[0054] It should be noted that the core components of the system include image acquisition module, deep learning model module and image processing module, which work together to realize image recognition and processing. The image acquisition module is responsible for capturing image data, the deep learning model module conducts in-depth analysis of these data, extracts key features and performs recognition, and the image processing module further processes the image based on the recognition results, such as optimization, enhancement or conversion.

[0055] Specifically, the image acquisition module can be any device that can capture images, such as a camera or scanner, while the deep learning model module may use advanced algorithms such as convolutional neural networks (CNN) to identify objects or features in the image. The image processing module may contain a variety of algorithms, such as filtering, edge detection, or image enhancement, to improve image quality or extract more useful information.

[0056] Preferably, the deep learning model module can be further refined into multiple submodules, such as feature extraction submodule, classification submodule and regression submodule, each submodule is responsible for different tasks. For example, the feature extraction submodule focuses on extracting key features from the image, while the classification submodule classifies the objects in the image according to these features. In addition, the parameter settings of the image processing module can be adjusted according to the specific application scenario. For example, in medical image processing, it may be necessary to emphasize the clarity and contrast of the image, while in security monitoring, more attention may be paid to dynamic capture and real-time performance.

[0057] In some embodiments, the system further includes a user interaction interface, which is used to receive instructions and parameters input by a user and to display processed image results.

[0058] It should be noted that the user interface is a bridge for information exchange between the system and the user. It not only receives the user's instructions and parameters, but also displays the processed image results. The design of the user interface is crucial to improving the user experience. It should be intuitive, easy to operate, and able to clearly display all necessary information.

[0059] Specifically, the user interaction interface may include a graphical user interface (GUI) in which various controls, such as buttons, sliders, text boxes, etc., are integrated for users to input instructions and parameters. For example, a user may select an image captured by an image acquisition module by clicking a button, set the size of a measurement area by adjusting a slider, or enter a specific parameter value through a text box. In addition, the interface may also display a progress bar, prompt information, and final image results of image processing.

[0060] Preferably, the user interaction interface can be further optimized to provide a richer interactive experience. For example, the interface can provide a real-time preview function, allowing users to immediately see the expected effect of image processing when adjusting parameters. In addition, the system can set up an intelligent recommendation system to automatically recommend appropriate parameter settings based on the user's operating habits and historical data. More specifically, for the parameter setting of the deep learning model module, a parameter adjustment wizard can be provided to guide users to select appropriate deep learning models and parameters based on the specific features of the image to achieve the best recognition and processing effects.

[0061] In some embodiments, when the system needs to perform distance measurement on an image, the user interaction interface obtains the measurement value by the following method, which specifically includes:

[0062] The user specifies the measurement area through an interactive interface;

[0063] The deep learning model module extracts features from the specified area and calculates preliminary measurements;

[0064] The image processing module performs optimization processing based on the preliminary measurement value to obtain the final measurement value;

[0065] The user interface displays the optimized measurement values ​​to the user in a graphical manner.

[0066] It should be noted that the user interaction interface obtains the measurement value through the following method, which indicates that the system provides an interactive mode, allowing the user to operate through the interface to achieve distance measurement of a specific area in the image.

[0067] Specifically, the user specifies the measurement area through the interactive interface, which involves the user selecting or drawing an area on the image, which is the target area that the user wants the system to measure. The deep learning model module extracts features from the specified area and calculates a preliminary measurement value, which involves the model analyzing the image features of the selected area and calculating a preliminary measurement result based on these features. The image processing module optimizes the preliminary measurement value to obtain the final measurement value, which may include steps such as removing outliers and smoothing data to ensure the accuracy of the measurement result.

[0068] Preferably, the user interface can provide more auxiliary tools, such as rulers, grid lines or reference points, to help users specify the measurement area more accurately. The deep learning model module can use a variety of algorithms and techniques, such as convolutional neural networks (CNNs), to improve the accuracy of feature recognition when performing feature extraction. The image processing module can use more advanced filtering algorithms, such as Kalman filtering or wavelet transforms, to further improve the smoothness and accuracy of the data when optimizing the measurement value. In addition, the system can also provide a variety of measurement modes, such as straight line measurement, curve measurement or area measurement, to adapt to different application scenarios.

[0069] In some embodiments, the image processing module optimizes the measurement value by the following method, which specifically includes:

[0070] A normal value sequence is obtained by removing outliers and data points exceeding a preset threshold from the preliminary measurement values;

[0071] Based on the optimal window value, the mean sequence is obtained by sliding the window for the normal value sequence;

[0072] Based on the optimal number of interpolation, interpolation is performed between two adjacent data points in the mean sequence to obtain the interpolated sequence;

[0073] The interpolated sequence is deduplicated and de-jittered to obtain the optimized measurement value.

[0074] It should be noted that the method of optimizing the measurement value by the image processing module is achieved by performing a series of data processing steps on the preliminary measurement value. This process includes removing outliers and data points exceeding a preset threshold, and then obtaining smoother and more accurate measurement values ​​based on these normal value sequences through sliding window and interpolation techniques.

[0075] Specifically, the image processing module first needs to define what an outlier is and what the preset threshold is. An outlier can be a data point that deviates greatly from other data points in the normal value sequence, and the preset threshold is a pre-set value used to determine whether a data point is abnormal. After determining these parameters, the module will perform a sliding window operation and select a window value, which determines the number of data points covered by the sliding window. The module will then calculate the mean of the data point values ​​in the window and update this mean as the sliding window moves to form a mean sequence.

[0076] Preferably, in order to further optimize the measured value, the image processing module can perform interpolation processing between adjacent data points of the mean sequence according to the optimal interpolation number based on the mean sequence obtained by sliding the window. This interpolation number can be calculated by the system processing time length of single-cycle image data and the window value. For example, if the system processing time length is T, the window value is W, and the number of data points in the normal value sequence is N, then the interpolation number I can be calculated according to the formula

[0077]

[0078] In addition, in order to improve the stability of the measured value, the interpolated sequence can be deduplicated and de-jittered to ensure the continuity and smoothness of the measurement result.

[0079] In some embodiments, when the system is first run or receives new image data, the optimal window value and the optimal number of interpolations are obtained by the following method, which specifically includes the following steps:

[0080] S1, removing outliers and data points exceeding a preset threshold in the preliminary measurement values ​​to obtain a normal value sequence;

[0081] S2, set the initial window value, the window value is smaller than the number of data points in the normal value sequence;

[0082] S3, based on the window value, the normal value sequence is calculated by sliding the window to obtain the mean value sequence;

[0083] S4, calculating the number of interpolation values ​​based on the window value, and performing interpolation processing on the mean sequence to obtain an interpolated sequence;

[0084] S5, performing deduplication and de-jittering processing on the interpolated sequence to obtain an optimized measurement value;

[0085] S6. Display the image and measurement results in real time according to the optimized measurement value to determine whether the display is smooth. If so, increase the window value and repeat S3-S6 until the display is not smooth; otherwise, reduce the window value and repeat S3-S6 until the display is smooth, and obtain the optimal window value and the optimal number of interpolations.

[0086] It should be noted that the preliminary measurement value refers to the raw data obtained during the image recognition process, which may contain outliers or points outside the normal range. Outliers refer to values ​​that are significantly different from most data points, and the preset threshold is a system-set parameter used to identify and exclude these outliers. The window value and the number of interpolations are key parameters that affect the smoothness and accuracy of data processing. The window value determines the range of data points considered when calculating the mean, while the number of interpolations determines how many additional data points are inserted into the mean sequence to improve the continuity of the data.

[0087] Preferably, the operation steps can be further refined or alternative solutions can be provided in the following ways. First, the system can use an adaptive algorithm to dynamically adjust the window value, for example, automatically adjust the window size according to the complexity and change rate of the image data. Second, the calculation of the number of interpolations can be based on the statistical characteristics of the image data, such as variance and standard deviation, to determine the optimal interpolation strategy. In addition, the system can also introduce a machine learning model to predict and adjust these parameters to adapt to different image processing requirements and improve the overall system performance.

[0088] In some embodiments, the number of interpolation values ​​is calculated based on the window value by the following formula:

[0089]

[0090] Where j is the number of interpolations, t is the time it takes for the system to process single-cycle image data, n is the window value, and N is the number of data points in the normal value sequence.

[0091] It should be noted that the calculation of the number of interpolation values ​​based on the window value by the following formula describes a specific calculation method for determining how many additional data points should be inserted between two data points when performing image data interpolation processing. The purpose of this method is to make the image data smoother and reduce noise or discontinuity that may be generated during data acquisition or processing.

[0092] Specifically, n in the formula represents the number of interpolations, t represents the duration of the system processing single-cycle image data, w represents the window value, and N represents the number of data points in the normal value sequence. The window value w is usually a positive integer less than or equal to N, which determines the number of data points covered during the sliding window operation. Through this formula, the number of data points to be inserted can be dynamically calculated based on the window value and the number of data points to achieve the best interpolation effect.

[0093] Preferably, the formula can be further refined to adapt to different application scenarios. For example, if the system processes image data for a short time t, it may be necessary to increase the window value w to obtain more data points for interpolation to ensure the continuity and smoothness of the image data. In addition, if the number of data points in the normal value sequence N is small, it may be necessary to adjust the window value to avoid excessive interpolation and data distortion. In practical applications, different window values ​​and durations can be selected according to the characteristics of the image data and the performance requirements of the system to achieve the best image processing effect.

[0094] In some embodiments, the following method is used to obtain the mean value sequence by sliding the window of the normal value sequence, and the method specifically includes:

[0095] Set a sliding window for the normal value sequence. The number of data points covered by the sliding window is the same as the window value. The starting point of the sliding window coincides with the first data point of the normal value sequence. Calculate and save the mean of the corresponding values ​​of the data points covered by the sliding window.

[0096] The sliding window moves backwards, and each time a data point is moved, the mean of the corresponding values ​​of the data points covered by the sliding window is calculated and saved, until the end point of the sliding window coincides with the last data point of the normal value sequence, and the mean of the corresponding values ​​of the last data point covered by the sliding window is calculated and saved;

[0097] Arrange the obtained moving means in the order of sliding window movement to obtain the mean sequence.

[0098] It should be noted that the image processing module obtains the mean sequence through the sliding window method in order to reduce the impact of random fluctuations on the measurement results and improve the stability and reliability of the data. This method obtains a smoother data sequence by sliding a fixed-size window on the data sequence and calculating the mean of all data points in the window.

[0099] Specifically, the key concepts involved in the sliding window method include window size, data point coverage, and mean calculation. The window size determines the number of data points covered by the sliding window, which is usually set according to the needs of the actual application and the characteristics of the data. For example, if the data changes relatively slowly, a larger window can be selected to obtain a smoother result; if the data changes drastically, a smaller window is required to maintain the sensitivity of the data. The data point coverage refers to the set of continuous data points covered by the sliding window when it moves on the data sequence. The mean calculation is the result of summing the values ​​of all data points in the sliding window and dividing it by the number of data points.

[0100] Preferably, when the sliding window method is implemented, the following steps can be used to further refine the operation: First, determine the window size, which can be determined based on the previous analysis of the data characteristics and experimental results. Secondly, initialize the sliding window position so that its starting point coincides with the first data point of the normal value sequence. Then, calculate the mean of all data points in the sliding window and save this value. Next, move the sliding window backward by a distance of one data point and repeat the process of calculating the mean until the sliding window covers the last data point. In addition, in order to improve computational efficiency, the cumulative sum method can be used to update the mean instead of recalculating the sum of all data points each time. As an alternative, other statistical methods such as moving median or moving weighted average can also be considered to smooth the data to adapt to different application scenarios and data characteristics.

[0101] In some embodiments, interpolation is performed between two adjacent data points in the mean sequence, specifically, the number of data points inserted between two adjacent data points in the mean sequence is equal to the number of interpolations, and the inserted data points form an arithmetic progression with the two adjacent data points.

[0102] It should be noted that the method of interpolating between two adjacent data points of the mean sequence is to add additional data points between the data points to improve the continuity and resolution of the measurement. This method generates new data points based on the existing data points through mathematical interpolation technology, making the entire data sequence smoother and reducing the errors that may be caused by large intervals between data points.

[0103] Specifically, the arithmetic sequence mentioned here is a simple mathematical interpolation method, in which each newly inserted data point forms an arithmetic sequence relationship with two adjacent original data points. When setting parameters, the tolerance of the arithmetic sequence can be determined according to the actual measurement accuracy requirements and the characteristics of the data. For example, if the interval between the original data points is large, a smaller tolerance value can be selected to increase the density of the data points; conversely, if the data points are already dense enough, a larger tolerance value can be selected to reduce the amount of calculation.

[0104] Preferably, in addition to using arithmetic progressions for interpolation, other more complex mathematical models, such as polynomial interpolation or spline interpolation, can also be considered to obtain better interpolation effects. In practical applications, the most appropriate interpolation method can be selected based on the characteristics of the data and the required accuracy requirements. For example, if the data changes relatively slowly, linear interpolation can be used; if the data changes more drastically, a higher-order polynomial interpolation or spline interpolation may be required to better fit the changing trend of the data. In addition, the accuracy and speed of interpolation can be adjusted according to the performance and response time requirements of the system to achieve the best balance.

[0105] In some embodiments, the interpolated sequence is de-jittered, specifically determining whether the difference between the values ​​corresponding to each two adjacent data points in the interpolated sequence is less than a preset threshold. If so, only one of the two adjacent data points is removed, otherwise, both adjacent data points are retained.

[0106] It should be noted that the de-jittering process for the interpolated sequence means that in the image processing process, in order to improve the stability and accuracy of the measurement data, the data sequence generated after interpolation needs to be further optimized to eliminate discontinuities caused by data fluctuations or noise.

[0107] Specifically, de-jittering involves comparing and analyzing the values ​​of every two adjacent data points in the interpolated sequence. If the difference between adjacent data points is less than a preset threshold, this indicates that the change between the two data points is small, which may be caused by noise or small fluctuations. In this case, the system can selectively remove one of the data points to reduce data discontinuity.

[0108] Preferably, the de-jittering process can be further refined into the following steps: First, determine a suitable threshold value, which can be set according to the actual application scenario and data characteristics to ensure that the de-jittering process does not over-smooth the data and thus lose important information. Secondly, the system can adopt a strategy of dynamically adjusting the threshold value to adjust the threshold value in real time according to the fluctuation of the data sequence to adapt to different measurement environments. In addition, it is also possible to consider introducing a machine learning algorithm to identify and predict the noise in the data through training models, so as to perform de-jittering more intelligently.

[0109] The above-mentioned embodiments of the present invention have the following beneficial effects: the image recognition and processing system can accurately identify key information in the image through the advanced feature extraction capability of its deep learning model module, thereby providing high-quality image processing results. The image acquisition module in the system design can ensure the real-time and accuracy of the data, while the introduction of the user interaction interface can enhance the user experience, allowing users to intuitively participate in the image recognition and processing process. In addition, when the system performs distance measurement, it can improve the accuracy and reliability of the measurement by removing outliers and data points through the optimization algorithm.

[0110] Furthermore, the system applies advanced sliding window and interpolation technology in the image processing module, which can not only enhance the continuity and smoothness of the data, but also further improve the stability of the data through deduplication and de-jittering. When the system is running for the first time or receiving new image data, it can automatically adjust the optimal window value and the number of interpolations to ensure the optimality and adaptability of the processing results. The application of these technologies enables the system to provide more accurate and efficient solutions when processing complex image data, meeting the high standards for image recognition and processing in different scenarios.

[0111] like Figure 2 As shown, in some embodiments, an image recognition and processing method 200 based on the system according to any one of claims 1 to 9 includes:

[0112] Step 201, the system starts and completes initialization;

[0113] Step 202, the user interaction interface receives user instructions and parameters;

[0114] Step 203, the image acquisition module captures image data;

[0115] Step 204, the deep learning model module extracts and recognizes features of the image data;

[0116] Step 205, the image processing module processes the image according to the recognition result;

[0117] Step 206: The user interaction interface displays the processed image result.

[0118] It is understandable that the modules described in the image recognition and processing method 200 and the reference Figure 1 Therefore, the operations, features and beneficial effects described above for the image recognition and processing system based on deep learning are also applicable to the image recognition and processing method 200 and the modules contained therein, and will not be repeated here.

[0119] Reference below Figure 3 , which shows a schematic diagram of a structure 300 of an electronic device suitable for implementing some embodiments of the present invention. The electronic devices in some embodiments of the present invention may include, but are not limited to, mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 3 The terminal device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0120] like Figure 3 As shown, the electronic device 300 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. In the RAM 303, various programs and data required for the operation of the electronic device 300 are also stored. The processing device 301, the ROM 302, and the RAM 303 are connected to each other via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0121] Typically, the following devices may be connected to the I / O interface 305: input devices 306 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; output devices 307 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 308 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 309. The communication devices 309 may allow the electronic device 300 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 3 The electronic device 300 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead. Figure 3 Each block shown in the figure may represent one device, or may represent multiple devices as required.

[0122] Furthermore, the storage medium of the embodiment of the present application stores program instructions that can implement all the above methods, wherein the program instructions can be stored in the above storage medium in the form of a software product, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or terminal devices such as a computer, a server, a mobile phone, and a tablet.

[0123] The above descriptions are only some preferred embodiments of the present invention and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present invention is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the above features are replaced with (but not limited to) technical features with similar functions disclosed in the embodiments of the present invention to form a technical solution.

Claims

1. An image recognition and processing system based on deep learning, characterized in that: The system includes an image acquisition module, a deep learning model module, and an image processing module, wherein: The image acquisition module is used to capture image data to be identified and processed; the deep learning model module is used to extract features and identify the captured image data; the image processing module is used to process the image accordingly according to the identification result; The deep learning model module and the image processing module are interconnected through a data interface to achieve data interaction and processing.

2. The image recognition and processing system based on deep learning according to claim 1, characterized in that: The system also includes a user interaction interface, which is used to receive instructions and parameters input by a user and display processed image results.

3. The image recognition and processing system based on deep learning according to claim 2, characterized in that: When the system needs to measure the distance of an image, the user interaction interface obtains the measurement value by the following method, which specifically includes: The user specifies the measurement area through an interactive interface; The deep learning model module extracts features from the specified area and calculates preliminary measurements; The image processing module performs optimization processing based on the preliminary measurement value to obtain the final measurement value; The user interface displays the optimized measurement values ​​to the user in a graphical manner.

4. The image recognition and processing system based on deep learning according to claim 3, characterized in that: The image processing module optimizes the measurement value by the following method, which specifically includes: A normal value sequence is obtained by removing outliers and data points exceeding a preset threshold from the preliminary measurement values; Based on the optimal window value, the mean sequence is obtained by sliding the window for the normal value sequence; Based on the optimal number of interpolation, interpolation is performed between two adjacent data points in the mean sequence to obtain the interpolated sequence; The interpolated sequence is deduplicated and de-jittered to obtain the optimized measurement value.

5. The image recognition and processing system based on deep learning according to claim 4, characterized in that: When the system is first run or receives new image data, the optimal window value and the optimal number of interpolations are obtained by the following method, which specifically includes the following steps: S1, removing outliers and data points exceeding a preset threshold in the preliminary measurement values ​​to obtain a normal value sequence; S2, set the initial window value, the window value is smaller than the number of data points in the normal value sequence; S3, based on the window value, the normal value sequence is calculated by sliding the window to obtain the mean value sequence; S4, calculating the number of interpolation values ​​based on the window value, and performing interpolation processing on the mean sequence to obtain an interpolated sequence; S5, performing deduplication and de-jittering processing on the interpolated sequence to obtain an optimized measurement value; S6. Display the image and measurement results in real time according to the optimized measurement value to determine whether the display is smooth. If so, increase the window value and repeat S3-S6 until the display is not smooth; otherwise, reduce the window value and repeat S3-S6 until the display is smooth, and obtain the optimal window value and the optimal number of interpolations.

6. The image recognition and processing system based on deep learning according to claim 5, characterized in that: The number of interpolation values ​​is calculated based on the window value using the following formula: Where j is the number of interpolations, t is the time it takes for the system to process single-cycle image data, n is the window value, and N is the number of data points in the normal value sequence.

7. The image recognition and processing system based on deep learning according to claim 4 or 5, characterized in that: The following method is used to obtain the mean value sequence by sliding the window of the normal value sequence, and the method specifically includes: Set a sliding window for the normal value sequence. The number of data points covered by the sliding window is the same as the window value. The starting point of the sliding window coincides with the first data point of the normal value sequence. Calculate and save the mean of the corresponding values ​​of the data points covered by the sliding window. The sliding window moves backwards, and each time a data point is moved, the mean of the corresponding values ​​of the data points covered by the sliding window is calculated and saved, until the end point of the sliding window coincides with the last data point of the normal value sequence, and the mean of the corresponding values ​​of the last data point covered by the sliding window is calculated and saved; Arrange the obtained moving means in the order of sliding window movement to obtain the mean sequence.

8. The image recognition and processing system based on deep learning according to claim 4 or 5, characterized in that: Interpolation is performed between two adjacent data points of the mean sequence, specifically, the number of data points inserted between two adjacent data points of the mean sequence is equal to the number of interpolations, and the inserted data points and the two adjacent data points form an arithmetic progression.

9. The image recognition and processing system based on deep learning according to claim 4 or 5, characterized in that: The interpolated sequence is de-jittered, specifically referring to determining whether the difference between the values ​​corresponding to each two adjacent data points in the interpolated sequence is less than a preset threshold. If so, only one of the two adjacent data points is removed, otherwise, both of the two adjacent data points are retained.

10. An image recognition and processing method based on the system according to any one of claims 1 to 9, characterized in that: The method specifically comprises: The system starts and completes initialization; The user interaction interface receives user instructions and parameters; The image acquisition module captures image data; The deep learning model module extracts and recognizes features of image data; According to the recognition result, the image processing module processes the image; The user interface displays the processed image results.

Images