Intelligent image processing method and system for bionic human eye Internet of Things
The intelligent image processing method and system for bionic human eye IoT sensors address the challenges of high-intensity manual inspection and difficult data collection by capturing images at multiple angles, performing task segmentation, and integrating front-end and back-end processing to improve efficiency and reduce costs.
Patent Information
- Application Number
- CN202410733037.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-07
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2044-06-07
AI Technical Summary
Existing industrial instruments require high-intensity manual inspection and face difficulties in data collection, leading to inefficient data processing and increased costs.
An intelligent image processing method and system using a bionic human eye IoT sensor for capturing images at multiple spatial angles, combining front-end and back-end processing mechanisms for task segmentation, and performing multi-source joint processing analysis to improve data collection and analysis efficiency.
Enhances data collection and analysis efficiency while reducing inspection costs by utilizing a bionic human eye IoT sensor for comprehensive and accurate image acquisition and processing.
Smart Images

Figure CN118314448B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of image processing technology, and in particular to an intelligent image processing method and system for a bionic human eye Internet of Things. Background Art
[0002] The bionic human eye vision system has the ability to efficiently perceive and understand complex scenes, and can be widely used in monitoring, security, intelligent transportation and other fields. In the existing industrial environment, there are a large number of traditional industrial instruments such as electricity, water, oil, gas, and heat. These traditional instruments need to be inspected and recorded manually, and there are technical problems such as high intensity of instrument inspection and difficult data collection. Summary of the invention
[0003] The present invention provides an intelligent image processing method and system for a bionic human eye Internet of Things to solve the technical problems of high intensity of instrument inspection and difficult data collection in the prior art, and achieve the technical effect of improving data collection and analysis efficiency and reducing inspection costs.
[0004] In a first aspect, the present invention provides an intelligent image processing method for a bionic human eye Internet of Things, wherein the method comprises:
[0005] Capturing and acquiring target acquisition images, wherein the target acquisition images are acquired through scene self-adjustment based on the bionic human eye Internet of Things sensor equipped with the smart instrument, and are image sets at multiple spatial angles;
[0006] Reading the image quality standard, combining the front-end processing mechanism and the back-end processing mechanism to perform task segmentation based on the image quality standard, and determining the image processing task, wherein the image processing task includes the front-end processing task and the back-end processing task;
[0007] Based on the front-end processing module, the target acquisition image is subjected to preliminary processing based on the front-end processing task to determine a valid target image, wherein the front-end processing module includes a preprocessing layer and a neuromorphic processing layer, wherein the preliminary processing includes a one-step image processing based on the image processing type and a two-step image processing based on multivariate feature extraction and image splicing reconstruction;
[0008] In combination with the communication port, the effective target image is transmitted to a back-end processing module, wherein the back-end processing module is embedded in a platform device connected to the sensor communication, and has a one-to-many communication relationship with the bionic human eye Internet of Things sensor;
[0009] Based on the back-end processing module, a multi-source joint processing analysis is performed on the received valid target image to determine an assembled target image, wherein the multi-source joint processing analysis is a process of performing sequential image processing based on parallelism and correlation by performing clustering processing;
[0010] The effective target image and the assembled target image are displayed visually based on a mapped human-computer interaction interface.
[0011] The method in the above implementation uses a bionic human eye IoT sensor based on an intelligent instrument assembly for image acquisition, realizes image sets at multiple spatial angles, and improves the comprehensiveness and accuracy of image acquisition. Combining the front-end processing mechanism and the back-end processing mechanism to perform task segmentation based on image quality standards makes the image processing task more intelligent and adaptive, and improves the accuracy and efficiency of image processing. The back-end processing module is embedded in the platform device connected to the sensor communication, realizing a one-to-many communication relationship, improving processing efficiency and communication stability. In addition, the visual display of effective target images and assembly target images provides an intuitive image display for the human-computer interaction interface, improving user experience and operating efficiency.
[0012] In a feasible implementation, performing preliminary processing based on the front-end processing task on the target acquisition image, the method further includes:
[0013] Identify the target acquisition image and the front-end processing task, and determine the image processing type, wherein each image processing type is identified by amplitude modulation and order;
[0014] Determine a one-step processing type based on the image processing type, perform one-step image processing in combination with the preprocessing layer, and determine a preprocessed image, wherein one step is used to define a processing sequence, and the one-step image processing is a preprocessing step for image quality features;
[0015] A two-step processing type based on the image processing type is determined, the pre-processed image is transferred to the neuromorphic processing layer, multivariate feature extraction and image splicing reconstruction are performed, and the effective target image is determined.
[0016] In this method, the type of image processing required is first determined according to the target and task, so that corresponding processing strategies can be adopted for different processing requirements. Then, one-step processing and two-step processing are performed according to different processing types, and finally an effective target image is obtained.
[0017] In a feasible implementation, multivariate feature extraction and image splicing reconstruction are performed, and the method further includes:
[0018] Perform boundary overlap positioning on the pre-processed image, perform splicing processing on the tracking images at multiple spatial angles, and determine a global pre-processed image;
[0019] Identify the preprocessed image, perform recognition extraction and feature reconstruction of the image under multiple angles, and determine the reconstructed image features, wherein the reconstructed features include static dimensions and dynamic dimensions;
[0020] Based on the global preprocessed image and the reconstructed image features, image reconstruction is performed to determine the effective target image.
[0021] That is to say, the pre-processed image is overlapped and located to ensure that the image can be accurately spliced when it is tracked at multiple spatial angles, ensuring the continuity and integrity of the image. According to the overlapped boundary positioning results, the image is tracked and spliced at multiple spatial angles, so that the image can fully display the target features at different angles.
[0022] In a feasible implementation, front-end and back-end interactive management is performed based on image characteristics and channel transmission characteristics, wherein the image characteristics at least include image sensitivity, and the channel transmission characteristics include channel basic characteristics and channel real-time characteristics.
[0023] In a feasible implementation, performing multi-source joint processing analysis on the received valid target image, the method further includes:
[0024] Identify the backend processing task, perform hierarchical clustering on the returned valid target image, and determine multiple groups of images, wherein the backend processing task includes a cluster task type and an independent task type, and the valid target image is identified with a sensor code;
[0025] Traversing the plurality of groups of images, sorting the plurality of groups of images based on task parallelism and relevance, and determining a task processing order;
[0026] The task processing order is mapped to the multiple groups of images, and image processing is performed to determine the assembled target image.
[0027] In a feasible implementation, the assembled target image includes a fusion image of multiple bionic human eye Internet of Things sensors and a temporal and spatial trend distribution image.
[0028] In a feasible implementation, the execution is based on visual display of the mapped human-computer interaction interface, and the method further includes:
[0029] Traversing the valid target image, determining the industrial data features based on the front-end processing task, and generating a single point feature single column;
[0030] Mapping the effective target image and the single-point feature in a single column, and visualizing them on a display interface of a mapped smart meter;
[0031] Traversing the target image of the assembly, determining the industrial data features based on the backend processing task as a single global feature column;
[0032] The assembled target image and the global feature are mapped to a single column and visualized on a platform device display interface.
[0033] In a feasible implementation, after visually displaying the mapping human-computer interaction interface, the method further includes:
[0034] Generate a local management plan based on the effective target image and the single point feature list;
[0035] Generate a comprehensive management plan based on the overall target image and the global feature list;
[0036] Conducting decision collision analysis on the local management plan and the overall management plan, adjusting the plan in combination with the management avoidance principle, and determining a preliminary management plan, wherein the management avoidance principle is determined based on the priority of management elements;
[0037] Based on the pre-management plan, industrial management of the operating area is carried out.
[0038] In a second aspect, the present invention further provides an intelligent image processing system for a bionic human eye Internet of Things, wherein the system comprises:
[0039] An image acquisition unit, wherein the image acquisition unit is used to capture a target acquisition image, wherein the target acquisition image is acquired through scene self-adjustment based on a bionic human eye IoT sensor equipped with a smart meter, and is an image set at multiple spatial angles;
[0040] A task allocation unit, the task allocation unit is used to read the image quality standard, combine the front-end processing mechanism and the back-end processing mechanism to perform task segmentation based on the image quality standard, and determine the image processing task, the image processing task includes the front-end processing task and the back-end processing task;
[0041] A front-end processing unit, the front-end processing unit is used to perform preliminary processing based on the front-end processing task on the target acquisition image based on the front-end processing module to determine a valid target image, the front-end processing module includes a preprocessing layer and a neuromorphic processing layer, wherein the preliminary processing includes a one-step image processing based on the image processing type and a two-step image processing based on multivariate feature extraction and image splicing reconstruction;
[0042] A relay transmission unit, the relay transmission unit is used to transmit the effective target image to a back-end processing module in combination with a communication port, the back-end processing module is embedded in a platform device connected to the sensor communication, and has a one-to-many communication relationship with the bionic human eye Internet of Things sensor;
[0043] A back-end processing unit, the back-end processing unit is used to perform multi-source joint processing analysis on the received valid target image based on the back-end processing module to determine the assembled target image, wherein the multi-source joint processing analysis is a process of performing sequential image processing based on parallelism and correlation by performing clustering processing;
[0044] A visualization unit is used to perform visualization display on the effective target image and the assembly target image based on a mapping human-computer interaction interface.
[0045] The present invention discloses an intelligent image processing method and system for a bionic human eye Internet of Things, comprising: a bionic human eye Internet of Things sensor assembled on an intelligent instrument performs collection through scene self-adjustment, captures a target collection image, and obtains an image set at multiple spatial angles; reads an image quality standard, combines a front-end processing mechanism with a back-end processing mechanism to perform task segmentation based on the image quality standard, and determines an image processing task, including a front-end processing task and a back-end processing task; based on a front-end processing module, performs preliminary processing on the target collection image to determine a valid target image, the front-end processing module includes a preprocessing layer and a neuromorphic processing layer; in combination with a communication port, transmits the valid target image to a back-end processing module, the back-end processing module is embedded in a platform device connected to the sensor communication, and has a one-to-many communication relationship with the bionic human eye Internet of Things sensor; based on the back-end processing module, performs multi-source joint processing and analysis on the received valid target image to determine an assembled target image; performs visual display based on a mapping human-computer interaction interface on the valid target image and the assembled target image. The intelligent image processing method and system for the bionic human eye Internet of Things disclosed in the present invention solve the technical problems of high intensity of instrument inspection and difficult data collection, and achieve the technical effect of improving data collection and analysis efficiency and reducing inspection costs. BRIEF DESCRIPTION OF THE DRAWINGS
[0046] Figure 1 It is a schematic diagram of the flow of the intelligent image processing method for bionic human eye Internet of Things of the present invention;
[0047] Figure 2 The figure is a schematic diagram of the structure of the intelligent image processing system for the bionic human eye Internet of Things of the present invention.
[0048] Explanation of the reference numerals: image acquisition unit 11 , task allocation unit 12 , front-end processing unit 13 , relay transmission unit 14 , back-end processing unit 15 , visualization unit 16 . DETAILED DESCRIPTION
[0049] The technical solution provided in the embodiments of the present invention is to solve the technical problems of high instrument inspection work intensity and difficult data collection in the prior art. The overall idea adopted is as follows:
[0050] First, the target acquisition images are captured and obtained. These images are collected by the bionic human eye IoT sensor equipped with the intelligent instrument through scene self-adjustment, forming an image set at multiple spatial angles. Then, the image quality standard is read, and the front-end processing mechanism and the back-end processing mechanism are combined to perform task segmentation according to the image quality standard to determine the image processing tasks, including the front-end processing tasks and the back-end processing tasks. Next, the target acquisition images are preliminarily processed based on the front-end processing module to determine the effective target image. The front-end processing module includes a preprocessing layer and a neuromorphic processing layer. Subsequently, the effective target image is transmitted to the back-end processing module using the communication port. The back-end processing module is embedded in the platform device connected to the sensor communication, and there is a one-to-many communication relationship between it and the bionic human eye IoT sensor. In the back-end processing module, the received effective target images are subjected to multi-source joint processing and analysis, and the final assembly target image is determined. Finally, the mapping human-computer interaction interface is used to visualize the effective target image and the assembly target image.
[0051] The above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods of the specification to better understand the above technical solution. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments used only to explain the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In addition, it should be noted that, for the convenience of description, only the parts related to the present invention are shown in the drawings, rather than all of them.
[0052] Embodiment 1
[0053] Figure 1 The figure is a flow chart of the intelligent image processing method for the bionic human eye Internet of Things of the present invention; wherein the intelligent image processing method for the bionic human eye Internet of Things includes:
[0054] Capturing and acquiring target acquisition images, wherein the target acquisition images are acquired through scene self-adjustment based on the bionic human eye Internet of Things sensor equipped with the smart instrument, and are image sets at multiple spatial angles;
[0055] In other words, the target image is collected based on the bionic human eye IoT sensor installed in the smart meter, which involves obtaining image sets at multiple spatial angles through scene self-adjustment. For example, in different light environments at multiple spatial angles, the parameters of the image acquisition device, such as exposure time, aperture size, ISO, etc., are sensed and adjusted to adapt to the current light environment to achieve clear capture of the target.
[0056] By using the principle of bionic human eyes and intelligent scene self-adjustment to obtain clear images at different angles, it helps to obtain more comprehensive and accurate target image data.
[0057] Reading the image quality standard, combining the front-end processing mechanism and the back-end processing mechanism to perform task segmentation based on the image quality standard, and determining the image processing task, wherein the image processing task includes the front-end processing task and the back-end processing task;
[0058] Optionally, after image acquisition, task segmentation is performed based on the read image quality standard in combination with the front-end processing mechanism and the back-end processing mechanism to determine the image processing task. The image processing task includes the front-end processing task and the back-end processing task. The front-end processing module performs preliminary processing on the acquired target image according to the front-end processing task to determine the effective target image. The back-end processing module performs multi-source joint processing analysis according to the back-end processing task to determine the assembled target image.
[0059] Specifically, the image quality standard of the target application scenario is obtained, and the image quality standard involves aspects such as image clarity, contrast, and color accuracy. Exemplarily, the image quality standard is set by a dual threshold method. Based on features such as image brightness and contrast, the image is divided into three qualities by two quality thresholds: low-quality image, medium-quality image, and high-quality image. Then, for low-quality images, they are discarded and re-captured; medium-quality images are assigned to the front-end processing module for image enhancement; for high-quality images, they are directly assigned to the back-end processing module for analysis.
[0060] Furthermore, the two quality thresholds set by the dual threshold method are determined based on the front-end processing mechanism and the back-end processing mechanism. Specifically, based on the upper limit of the processing capacity of the front-end processing mechanism, a first quality threshold is determined, and the first quality threshold is used to distinguish between low-quality images and medium-quality images; based on the upper limit of the tolerance of the back-end processing mechanism, a second threshold is determined, and the second threshold is used to distinguish between medium-quality images and high-quality images.
[0061] Furthermore, according to the image quality and the above-mentioned image quality standards, the image processing tasks are divided into front-end processing tasks and back-end processing tasks, and the specific content and execution flow of each task are determined. The image processing tasks are determined automatically, thereby improving the efficiency and accuracy of image processing.
[0062] Based on the front-end processing module, the target acquisition image is subjected to preliminary processing based on the front-end processing task to determine a valid target image, wherein the front-end processing module includes a preprocessing layer and a neuromorphic processing layer, wherein the preliminary processing includes a one-step image processing based on the image processing type and a two-step image processing based on multivariate feature extraction and image splicing reconstruction;
[0063] Optionally, the front-end processing module includes a preprocessing layer and a neuromorphic processing layer, which are used to perform preliminary processing of the front-end processing task on the target acquisition image to determine the effective target image. Among them, the preprocessing layer involves operations such as image denoising, enhancement, and resizing, while the neuromorphic processing layer is used to perform image feature extraction and morphological transformation to help improve the quality and usability of the image and provide a better data basis for subsequent analysis and processing.
[0064] In some embodiments, performing preliminary processing based on the front-end processing task on the target acquisition image, the method further includes:
[0065] Identify the target acquisition image and the front-end processing task, and determine the image processing type, wherein each image processing type is identified by amplitude modulation and order;
[0066] Determine a one-step processing type based on the image processing type, perform one-step image processing in combination with the preprocessing layer, and determine a preprocessed image, wherein one step is used to define a processing sequence, and the one-step image processing is a preprocessing step for image quality features;
[0067] A two-step processing type based on the image processing type is determined, the pre-processed image is transferred to the neuromorphic processing layer, multivariate feature extraction and image splicing reconstruction are performed, and the effective target image is determined.
[0068] Specifically, identifying the target acquisition image and the front-end processing task, and determining the image processing type helps to more accurately determine the processing steps that need to be performed. Each image processing type is identified with modulation and order, where modulation refers to the intensity or degree of processing, and order refers to the order or priority of processing. Exemplarily, first, based on the quality characteristics of the target acquisition image, the target acquisition image is matched with the front-end processing task to determine which tasks are applicable to the current image. Then, based on the matching results, the image processing type is determined, and its modulation and order are identified. Image processing types include denoising, enhancement, background segmentation, and resizing, and each process has different modulation and order requirements.
[0069] Optionally, based on the determined image processing type, the type of one-step processing to be performed is determined. This includes processing types such as denoising, enhancement, and resizing. The determined one-step processing type is combined with a preprocessing layer to perform one-step image processing to obtain a preprocessed image. The preprocessed image is an image that has been preliminarily processed and is an input for subsequent processing steps. By determining the one-step processing type based on the image processing type and performing one-step image processing in combination with the preprocessing layer, the quality and applicability of the image can be improved.
[0070] Optionally, based on the image processing type determined previously, the type of two-step processing to be performed is determined, including processing types such as multivariate feature extraction and image stitching reconstruction. Then, the preprocessed image is transferred to the neuromorphic processing layer to perform multivariate feature extraction and image stitching reconstruction. The neuromorphic processing layer is built based on technologies such as neural networks to extract image features and reconstruct images. After multivariate feature extraction and image stitching reconstruction, the effective target image is determined for subsequent analysis and application.
[0071] By determining the type of two-step processing based on the image processing type and transferring the pre-processed image stream to the neuromorphic processing layer for processing, the image can be subjected to deeper feature extraction and reconstruction, thereby improving the quality and information content of the image.
[0072] In some implementations, multivariate feature extraction and image stitching reconstruction are performed, and the method further includes:
[0073] Perform boundary overlap positioning on the pre-processed image, perform splicing processing on the tracking images at multiple spatial angles, and determine a global pre-processed image;
[0074] Identify the preprocessed image, perform recognition extraction and feature reconstruction of the image under multiple angles, and determine the reconstructed image features, wherein the reconstructed features include static dimensions and dynamic dimensions;
[0075] Based on the global preprocessed image and the reconstructed image features, image reconstruction is performed to determine the effective target image.
[0076] Optionally, the preprocessed image is overlapped to determine the position and shape of the image boundary. Then, based on the image registration method at multiple spatial angles, the feature points in the image are extracted, and the image splicing process is tracked according to the relative position relationship between these feature points to determine the global preprocessed image, which is the overall preprocessed image. Through these steps, the consistency and integrity of the image at different angles are achieved, thereby obtaining a more accurate and reliable global preprocessed image.
[0077] Specifically, in some implementations, the preprocessed image is identified and image feature recognition, extraction and feature reconstruction are performed at multiple angles. First, the preprocessed image is identified to determine the image that needs to be feature extracted and reconstructed. Exemplarily, the image that needs to be feature extracted and reconstructed is the information display area image and features of the target instrument; then, feature extraction is performed on the image at multiple angles. And based on the features extracted at multiple angles, feature reconstruction is performed, and the extracted features are combined and converted into reconstructed features, wherein the reconstructed features include static dimensions (such as color, shape, texture, etc.) and dynamic dimensions (such as motion, change, etc.). Through the above steps, the preprocessed image is fully feature extracted and reconstructed, thereby obtaining reconstructed features containing rich information.
[0078] Furthermore, based on the image reconstruction algorithm, the global preprocessed image and the reconstructed image features are combined to perform image reconstruction. Specifically, image reconstruction includes operations such as image synthesis and filling in missing parts. Through image reconstruction, the final effective target image is obtained. The effective target image is an image that contains the optimal features of the multi-angle image after deep processing, and has higher quality and information content.
[0079] In combination with the communication port, the effective target image is transmitted to a back-end processing module, wherein the back-end processing module is embedded in a platform device connected to the sensor communication, and has a one-to-many communication relationship with the bionic human eye Internet of Things sensor;
[0080] Optionally, the effective target image is transmitted to the back-end processing module using a communication port, which involves encapsulating the image data and sending it to the platform device where the back-end processing module is located via a network or other communication methods.
[0081] Optionally, the back-end processing module is embedded in the platform device, and establishes a one-to-many communication relationship with the bionic human eye Internet of Things sensor through the platform device, so as to effectively manage and process the transmitted image data.
[0082] Based on the back-end processing module, a multi-source joint processing analysis is performed on the received valid target image to determine an assembled target image, wherein the multi-source joint processing analysis is a process of performing sequential image processing based on parallelism and correlation by performing clustering processing;
[0083] In some embodiments, performing multi-source joint processing analysis on the received valid target image, the method further comprises:
[0084] Identify the backend processing task, perform hierarchical clustering on the returned valid target image, and determine multiple groups of images, wherein the backend processing task includes a cluster task type and an independent task type, and the valid target image is identified with a sensor code;
[0085] Traversing the plurality of groups of images, sorting the plurality of groups of images based on task parallelism and relevance, and determining a task processing order;
[0086] The task processing order is mapped to the multiple groups of images, and image processing is performed to determine the assembled target image.
[0087] Specifically, the received valid target images are subjected to multi-source joint processing and analysis. First, the backend processing tasks are identified, where the backend processing tasks include cluster task types and independent task types, corresponding to the overall analysis of multiple sensor data and the extended analysis of single sensor data, respectively. Then, hierarchical clustering is performed on the returned valid target images to determine multiple groups of images. Multiple groups of images correspond to subtasks of multiple task types.
[0088] In addition, the valid target image is marked with a sensor code to distinguish image data from different sources and determine the sensor and instrument corresponding to the image data.
[0089] Optionally, the parallelism and correlation between tasks are analyzed. Some tasks may be performed simultaneously, while some tasks may need to rely on the results of other tasks, so that the multiple groups of images are sorted and the order of task processing is determined. For example, tasks that can be processed in parallel and do not rely on the results of other tasks are processed first, and then other tasks are processed in sequence. This optimizes the processing flow and improves processing efficiency.
[0090] Optionally, according to the order of task processing, multiple groups of images are processed to finally determine the target image. This includes operations such as image synthesis, feature extraction, data fusion, and statistical analysis. Through these steps, joint processing and analysis of multi-source images are achieved to finally determine the target image. This helps to integrate image data from different sensors or tasks and provide more comprehensive and accurate analysis results.
[0091] In some embodiments, the assembled target image includes a fused image of multiple bionic human eye Internet of Things sensors and a temporal and spatial-based trend distribution image.
[0092] Optionally, the overall target image includes a fusion image of multiple bionic human eye IoT sensors and a temporal and spatial trend distribution image. The overall target image combines information from multiple sensors and includes the distribution trend of temporal and spatial information, which helps to provide more comprehensive analysis and cognition.
[0093] Among them, the fused image is the result of processing the images captured by multiple bionic human eye IoT sensors. By fusing the images captured by multiple sensors, the clarity, accuracy and information content of the image can be improved, thereby better expressing the characteristics and status of the monitored target. The trend distribution image based on time and space describes the change trend of the monitored target in time and space.
[0094] The effective target image and the assembled target image are displayed visually based on a mapped human-computer interaction interface.
[0095] In some embodiments, the execution is based on visual display of a mapped human-computer interaction interface, and the method further includes:
[0096] Traversing the valid target image, determining the industrial data features based on the front-end processing task, and generating a single point feature single column;
[0097] Mapping the effective target image and the single-point feature in a single column, and visualizing them on a display interface of a mapped smart meter;
[0098] Traversing the target image of the assembly, determining the industrial data features based on the backend processing task as a single global feature column;
[0099] The assembled target image and the global feature are mapped to a single column and visualized on a platform device display interface.
[0100] Optionally, first, by traversing the valid target image, determine the single-point feature column, and then map it to the display interface of the smart meter. Next, traverse the assembly target image, determine the global feature column, and map it to the display interface of the platform device. Among them, the single-point feature column contains isolated features of multiple targets. The global feature column contains the global features of multiple targets. Presenting the processed industrial data features in an intuitive way helps users understand and analyze the data.
[0101] In some embodiments, after visually displaying the mapping human-computer interaction interface, the method further includes:
[0102] Generate a local management plan based on the effective target image and the single point feature list;
[0103] Generate a comprehensive management plan based on the overall target image and the global feature list;
[0104] Conducting decision collision analysis on the local management plan and the overall management plan, adjusting the plan in combination with the management avoidance principle, and determining a preliminary management plan, wherein the management avoidance principle is determined based on the priority of management elements;
[0105] Based on the pre-management plan, industrial management of the operating area is carried out.
[0106] Optionally, based on multiple valid target images and corresponding single-point feature lists, isolated management analysis of multiple targets is performed, and a local management plan is generated, including management measures such as monitoring, control, scheduling, maintenance, etc. of the targets and their influence ranges, so as to ensure their stable operation within the specified range and optimize the operating status and efficiency of single-point targets.
[0107] Optionally, based on the overall feature information in the assembly target image and the global feature single column, such as the overall operating status, overall trend, etc., a management plan for the whole is formulated, including measures such as monitoring, control, and scheduling of the overall target, so as to optimize the overall operating status and efficiency. Exemplarily, based on the overall operating status, the instrument failure rate and damage degree are evaluated, and the global maintenance cycle is adjusted.
[0108] Optionally, conduct a comprehensive analysis of the local management plan and the overall management plan to identify possible conflicts or inconsistencies, including conflicts in resource allocation, priority setting, operating procedures, etc.
[0109] Optionally, adjust the plan according to the management avoidance principle, such as resource reallocation, process optimization, etc. The management avoidance principle is determined based on the priority of management elements, including ensuring safety, maximizing efficiency, minimizing costs, etc. The determined pre-management plan ensures the effectiveness and feasibility of the management plan.
[0110] In some embodiments, front-end and back-end interactive management is performed based on image characteristics and channel transmission characteristics, wherein the image characteristics include at least image sensitivity, and the channel transmission characteristics include channel basic characteristics and channel real-time characteristics.
[0111] The image characteristics include image sensitivity, etc. Image sensitivity refers to the image's ability to respond to input signals, that is, the image's sensitivity to changes in the input signal. According to the image's sensitivity, the transmission parameters can be adjusted to ensure the quality and integrity of the image during transmission.
[0112] Channel transmission characteristics include basic characteristics and real-time characteristics of the channel. Basic characteristics include the transmission rate and signal-to-noise ratio of the channel, while real-time characteristics include the delay and jitter of the channel. According to the basic characteristics and real-time characteristics of the channel, appropriate transmission methods and parameters can be selected to ensure the real-time and stability of the image.
[0113] Optionally, targeted classified transmission based on image characteristics is performed according to available channel information. In other words, appropriate transmission methods and parameters are selected according to image characteristics and channel conditions to maximize transmission efficiency and ensure image quality and integrity during transmission.
[0114] In summary, the intelligent image processing method for bionic human eye Internet of Things provided by the present invention has the following technical effects:
[0115] By capturing and acquiring the target acquisition image, the target acquisition image is acquired through scene self-adjustment based on the bionic human eye IoT sensor equipped with the intelligent instrument, and is an image set under multiple spatial angles; the image quality standard is read, and the task segmentation based on the image quality standard is combined with the front-end processing mechanism and the back-end processing mechanism to determine the image processing task, which includes the front-end processing task and the back-end processing task; based on the front-end processing module, the target acquisition image is subjected to preliminary processing based on the front-end processing task to determine the effective target image, and the front-end processing module includes the preprocessing layer and the neuromorphic processing layer; combined with the communication port, the effective target image is transmitted to the back-end processing module, which is embedded in the platform device connected to the sensor communication, and has a one-to-many communication relationship with the bionic human eye IoT sensor; based on the back-end processing module, the received effective target image is subjected to multi-source joint processing and analysis to determine the assembled target image; the effective target image and the assembled target image are subjected to visual display based on the mapping human-computer interaction interface. The technical effect of improving data acquisition and analysis efficiency and reducing inspection costs is achieved.
[0116] Embodiment 2
[0117] Figure 2 Schematic diagram of the structure of the intelligent image processing system for the bionic human eye Internet of Things of the present invention. Figure 1 The flow chart of the intelligent image processing method for the bionic human eye Internet of Things of the present invention can be shown as follows: Figure 2 The structure shown is implemented.
[0118] Based on the same concept as the intelligent image processing method for the bionic human eye Internet of Things in the above embodiment, the present invention also provides an intelligent image processing system for the bionic human eye Internet of Things, including:
[0119] An image acquisition unit 11 is used to capture a target image. The target image is acquired through scene self-adjustment based on a bionic human eye IoT sensor installed on a smart meter, and is an image set at multiple spatial angles.
[0120] The task allocation unit 12 is used to read the image quality standard, perform task segmentation based on the image quality standard in combination with the front-end processing mechanism and the back-end processing mechanism, and determine the image processing task, wherein the image processing task includes the front-end processing task and the back-end processing task;
[0121] A front-end processing unit 13 is used to perform preliminary processing based on the front-end processing task on the target acquisition image based on the front-end processing module to determine a valid target image, wherein the front-end processing module includes a preprocessing layer and a neuromorphic processing layer, wherein the preliminary processing includes a one-step image processing based on the image processing type and a two-step image processing based on multivariate feature extraction and image splicing reconstruction;
[0122] The relay transmission unit 14 is used to transmit the effective target image to the back-end processing module in combination with the communication port. The back-end processing module is embedded in the platform device connected to the sensor communication and has a one-to-many communication relationship with the bionic human eye Internet of Things sensor.
[0123] A back-end processing unit 15 is used to perform multi-source joint processing analysis on the received valid target image based on the back-end processing module to determine the assembled target image, wherein the multi-source joint processing analysis is a process of performing sequential image processing based on parallelism and correlation by performing clustering processing;
[0124] The visualization unit 16 is used to perform visualization display of the effective target image and the assembly target image based on a mapping human-computer interaction interface.
[0125] The front-end processing unit 13 includes:
[0126] A processing classification unit, used to identify the target acquisition image and the front-end processing task, and determine the image processing type, wherein each image processing type is identified by amplitude modulation and order;
[0127] A one-step image processing unit, used to determine a one-step processing type based on the image processing type, perform one-step image processing in combination with the preprocessing layer, and determine a preprocessed image, wherein one step is used to define a processing sequence, and the one-step image processing is a preprocessing step for image quality features;
[0128] A reconstruction unit is used to determine a two-step processing type based on the image processing type, transfer the pre-processed image flow to the neuromorphic processing layer, perform multivariate feature extraction and image splicing reconstruction, and determine the effective target image.
[0129] In some implementations, the reconstruction unit in the front-end processing unit 13 includes:
[0130] A global preprocessing unit, used for performing boundary overlap positioning on the preprocessed image, performing splicing processing on the tracking images at multiple spatial angles, and determining a global preprocessed image;
[0131] A feature recognition, extraction and reconstruction unit, used to recognize the preprocessed image, perform recognition, extraction and feature reconstruction of the image features at multiple angles, and determine the reconstructed image features, wherein the reconstructed features include static dimensions and dynamic dimensions;
[0132] The effective image reconstruction unit is used to perform image reconstruction based on the global preprocessed image and the reconstructed image features to determine the effective target image.
[0133] Furthermore, the backend processing unit 15 includes:
[0134] A hierarchical clustering unit, used to identify the backend processing task, hierarchically cluster the returned valid target image, and determine multiple groups of images, wherein the backend processing task includes a cluster task type and an independent task type, and the valid target image is identified with a sensor code;
[0135] An image sorting unit, configured to traverse the plurality of groups of images and determine a task processing order for the plurality of groups based on task parallelism and relevance;
[0136] A mapping processing unit is used to map the task processing order with the multiple groups of images, and perform image processing to determine the assembled target image.
[0137] Furthermore, the visualization unit 16 includes:
[0138] A single point feature unit, used to traverse the valid target image, determine the industrial data features based on the front-end processing task, and generate a single point feature column;
[0139] A single-point feature visualization unit, used for mapping the effective target image and the single-point feature to be visualized on a display interface of a mapped smart meter;
[0140] A global feature unit, used for traversing the assembly target image, determining the industrial data features based on the backend processing task as a global feature single column;
[0141] The global feature visualization unit is used to map the assembled target image and the global feature list, and visualize them on the platform device display interface.
[0142] Furthermore, the visualization unit 16 further includes:
[0143] A local management solution unit, configured to generate a local management solution based on the valid target image and the single point feature list;
[0144] A comprehensive management solution unit, used for generating a comprehensive management solution based on the overall target image and the global feature list;
[0145] A scheme adjustment unit, configured to perform decision collision analysis on the local management scheme and the overall management scheme, adjust the scheme in combination with a management avoidance principle, and determine a preliminary management scheme, wherein the management avoidance principle is determined based on the priority of management elements;
[0146] The area management unit is used to perform industrial management of the operation area based on the pre-management plan.
[0147] Furthermore, the system also includes an interaction management unit, which is used to:
[0148] Based on the image characteristics and channel transmission characteristics, front-end and back-end interactive management is performed, wherein the image characteristics at least include image sensitivity, and the channel transmission characteristics include channel basic characteristics and channel real-time characteristics.
[0149] It should be understood that the embodiments mentioned in this specification focus on their differences from other embodiments. The specific embodiments in the aforementioned embodiment one are also applicable to the intelligent image processing system for the bionic human eye Internet of Things described in embodiment two. For the sake of brevity of the specification, they will not be further elaborated here.
[0150] It should be understood that the embodiments disclosed in the present invention and the above description can enable those skilled in the art to use the present invention to implement the present invention. At the same time, the present invention is not limited to the above-mentioned embodiments. It should be understood that those skilled in the art can still modify the technical solutions recorded in the above-mentioned embodiments, or replace some of the technical features therein by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.
Claims
1. An intelligent image processing method for bionic human eye Internet of Things, characterized in that: The method comprises: Capturing and acquiring target acquisition images, wherein the target acquisition images are acquired through scene self-adjustment based on the bionic human eye Internet of Things sensor equipped with the smart instrument, and are image sets at multiple spatial angles; Reading the image quality standard, combining the front-end processing mechanism and the back-end processing mechanism to perform task segmentation based on the image quality standard, and determining the image processing task, wherein the image processing task includes the front-end processing task and the back-end processing task; Based on the front-end processing module, the target acquisition image is subjected to preliminary processing based on the front-end processing task to determine a valid target image, wherein the front-end processing module includes a preprocessing layer and a neuromorphic processing layer, wherein the preliminary processing includes a one-step image processing based on the image processing type and a two-step image processing based on multivariate feature extraction and image splicing reconstruction; In combination with the communication port, the effective target image is transmitted to a back-end processing module, wherein the back-end processing module is embedded in a platform device connected to the sensor communication, and has a one-to-many communication relationship with the bionic human eye Internet of Things sensor; Based on the back-end processing module, a multi-source joint processing analysis is performed on the received valid target image to determine an assembled target image, wherein the multi-source joint processing analysis is a process of performing sequential image processing based on parallelism and correlation by performing clustering processing; Performing visual display on the effective target image and the assembled target image based on a mapping human-computer interaction interface; Performing preliminary processing based on the front-end processing task on the target acquisition image, the method further includes: Identify the target acquisition image and the front-end processing task, and determine the image processing type, wherein each image processing type is identified by amplitude modulation and order; Determine a one-step processing type based on the image processing type, perform one-step image processing in combination with the preprocessing layer, and determine a preprocessed image, wherein one step is used to define a processing sequence, and the one-step image processing is a preprocessing step for image quality features; Determine a two-step processing type based on the image processing type, transfer the pre-processed image to the neuromorphic processing layer, perform multivariate feature extraction and image splicing reconstruction, and determine the effective target image; Performing multi-source joint processing analysis on the received effective target image, the method further includes: Identify the backend processing task, perform hierarchical clustering on the returned valid target image, and determine multiple groups of images, wherein the backend processing task includes a cluster task type and an independent task type, and the valid target image is identified with a sensor code; Traversing the plurality of groups of images, sorting the plurality of groups of images based on task parallelism and relevance, and determining a task processing order; Mapping the task processing order with the multiple groups of images, and performing image processing to determine the assembled target image; The execution is based on visual display of the mapping human-computer interaction interface, and the method further includes: Traversing the valid target image, determining the industrial data features based on the front-end processing task, and generating a single point feature single column; Mapping the effective target image and the single-point feature in a single column, and visualizing them on a display interface of a mapped smart meter; Traversing the target image of the assembly, determining the industrial data features based on the backend processing task as a single global feature column; Mapping the assembled target image and the global feature single column, and visualizing them on the platform device display interface; Among them, multi-dimensional feature extraction and image splicing reconstruction are performed, and the method also includes: Perform boundary overlap positioning on the pre-processed image, perform splicing processing on the tracking images at multiple spatial angles, and determine a global pre-processed image; Identify the preprocessed image, perform recognition extraction and feature reconstruction of the image under multiple angles, and determine the reconstructed image features, wherein the reconstructed features include static dimensions and dynamic dimensions; Based on the global preprocessed image and the reconstructed image features, image reconstruction is performed to determine the effective target image; Wherein, the front-end and back-end interactive management is performed based on the image characteristics and the channel transmission characteristics, wherein the image characteristics at least include image sensitivity, and the channel transmission characteristics include channel basic characteristics and channel real-time characteristics; The assembled target image includes a fusion image of multiple bionic human eye IoT sensors and a temporal and spatially based trend distribution image; Wherein, after visual display based on mapping the human-computer interaction interface, the method further includes: Generate a local management plan based on the effective target image and the single point feature list; Generate a comprehensive management plan based on the overall target image and the global feature list; Conducting decision collision analysis on the local management plan and the overall management plan, adjusting the plan in combination with the management avoidance principle, and determining a preliminary management plan, wherein the management avoidance principle is determined based on the priority of management elements; Based on the pre-management plan, industrial management of the operating area is carried out.
2. An intelligent image processing system for the bionic human eye Internet of Things, characterized in that: The system is used to execute the method according to claim 1, and the system comprises: An image acquisition unit, wherein the image acquisition unit is used to capture a target acquisition image, wherein the target acquisition image is acquired through scene self-adjustment based on a bionic human eye IoT sensor equipped with a smart meter, and is an image set at multiple spatial angles; A task allocation unit, the task allocation unit is used to read the image quality standard, combine the front-end processing mechanism and the back-end processing mechanism to perform task segmentation based on the image quality standard, and determine the image processing task, the image processing task includes the front-end processing task and the back-end processing task; A front-end processing unit, the front-end processing unit is used to perform preliminary processing based on the front-end processing task on the target acquisition image based on the front-end processing module to determine a valid target image, the front-end processing module includes a preprocessing layer and a neuromorphic processing layer, wherein the preliminary processing includes a one-step image processing based on the image processing type and a two-step image processing based on multivariate feature extraction and image splicing reconstruction; A relay transmission unit, the relay transmission unit is used to transmit the effective target image to a back-end processing module in combination with a communication port, the back-end processing module is embedded in a platform device connected to the sensor communication, and has a one-to-many communication relationship with the bionic human eye Internet of Things sensor; A back-end processing unit, the back-end processing unit is used to perform multi-source joint processing analysis on the received valid target image based on the back-end processing module to determine the assembled target image, wherein the multi-source joint processing analysis is a process of performing sequential image processing based on parallelism and correlation by performing clustering processing; A visualization unit is used to perform visualization display on the effective target image and the assembly target image based on a mapping human-computer interaction interface.
Citation Information
Patent Citations
Neuromorphic memory and inference engine stacked with image sensor
CN114092309A
Coding method and device, digital retina system, electronic equipment and storage medium
CN114745556A
Video quality evaluation method and device for simulating human eye characteristics based on Transform
CN115713708A
Image processing system and image processing method
CN117475121A
Electronic device and method for supporting image data deblurring
CN117795974A
