Intelligent image fusion processing method and system based on heterogeneous calculation

Abstract

The invention relates to the technical field of image fusion, in particular to an intelligent image fusion processing method and system based on heterogeneous computing, and the method comprises the following steps: obtaining image layer data, extracting a frequency structure alignment channel position matching overlapping relation, and matching a behavior number and an image region removal overlapping number; and screening unbound channel pause path instruction update connection fields, sequencing thread numbers to participate in path connection, and writing in a connection information synchronization thread state registration record set. According to the method, through identification of the frequency overlapping relation between channels and extraction of a layer response structure, calibration and behavior number separation are carried out on an interference section, and path screening and connection state updating are carried out through linkage updating of the behavior number and the path relation; the sequence distribution of resource scheduling is completed through a matching mode of thread calling frequency and a path state, the unified registration of thread control data is completed through a path index and connection information synchronization mechanism, and the interference processing capability and the task regulation and control efficiency in the image fusion process are improved.

Description

Technical Field

[0001] This invention relates to the field of image fusion technology, and in particular to an intelligent image fusion processing method and system based on heterogeneous computing. Background Technology

[0002] Image fusion technology involves the joint processing of image data from multiple sources or multiple time periods. Core aspects include image registration, the formulation and execution of fusion strategies, and fusion quality evaluation. The goal is to generate a more informative output image by integrating complementary information from different images. This technology is widely used in remote sensing, medical imaging, night vision enhancement, intelligent monitoring, and autonomous driving, among other fields. Image fusion technology systems include pixel-level direct fusion, feature-level structural information fusion, and decision-level result fusion. Each method relies on specific information extraction methods and processing rules in practical applications. For example, in pixel-level fusion, multi-scale transformation is often used to extract and synthesize image frequency domain features; in feature-level fusion, the focus is on the joint expression of structural features such as edges and textures; and in decision-level fusion, rules are formulated and discriminative processing is performed based on the results of independent image processing.

[0003] Traditional image fusion processing refers to the process of performing image registration and fusion strategies on multi-source image data using a unified computing resource platform. This process typically relies on a single type of computing architecture, such as a central processing unit (CPU) system, to complete intensive computational tasks. It mainly uses image fusion algorithms such as wavelet transform, principal component analysis, and mutual information maximization to process the spatial or frequency information between multiple images and perform fusion operations. Traditional methods are limited by single-source computing resource allocation and limited task scheduling capabilities, making it difficult to simultaneously meet real-time and fusion quality requirements. Intelligent image fusion processing based on heterogeneous computing is proposed against this backdrop. It relies on a heterogeneous computing system composed of multiple types of processing units, such as graphics processing units (GPUs) and digital signal processors (DSPs), and employs task partitioning, parallel scheduling, and resource collaboration to complete the joint registration and fusion strategy execution of multiple images. The goal is to improve processing efficiency and system responsiveness by sharing the high-load computational tasks in complex image processing workflows through the collaboration of different computing resources.

[0004] Existing technologies do not construct a linkage model between channel behavior and interference response during image fusion processing. This makes it impossible to identify and suppress overlapping conflicts based on inter-channel response characteristics. Channel behavior lacks real-time effectiveness judgment criteria in path connection. The path management process has difficulty distinguishing between bound and unbound task relationships. When threads participate in path allocation, they do not take into account both call frequency and path status, resulting in increased randomness and adaptation deviation in resource scheduling. Path status records are scattered in different control structures and lack synchronous update logic. In multi-task load environments, problems such as disordered connection relationships and delayed resource response are prone to occur, leading to frequent interruptions in fusion path execution and fluctuations in task completion efficiency. Summary of the Invention

[0005] To address the technical problems existing in the prior art, embodiments of the present invention provide an intelligent image fusion processing method based on heterogeneous computing, comprising the following steps:

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an intelligent image fusion processing method based on heterogeneous computing, comprising the following steps:

[0007] S1: Obtain the current frame image channel layer data, extract the frequency response region structure of each channel, align the positions between each channel, match the overlapping response region relationship, and obtain the fusion interference overlapping section identifier set;

[0008] S2: Based on the coordinate data in the set of fused interference overlapping section identifiers, match the channel behavior number with the task image region, check the overlapping range, remove the corresponding number, and obtain the list of unbound channel behaviors;

[0009] S3: Based on the channel number in the unbound channel behavior list, locate the corresponding path item in the path connection table, compare the channel quantity with the behavior number item by item, filter out paths including multiple unbound numbers and stop the instruction, update the remaining path connection field, and obtain the image fusion path connection control status group.

[0010] S4: Based on the path status information in the image fusion path connection control state group, read the number of thread cache calls, sort the numbers according to the call frequency, select threads to participate in the connection, and obtain the heterogeneous thread access mapping table.

[0011] S5: Based on the numbering relationship in the heterogeneous thread access mapping table, locate the path position in the task path record table, write the connection information and synchronize the thread state to obtain the image fusion thread path registration record set.

[0012] As a further embodiment of the present invention, the fusion interference overlap segment identifier set includes channel frequency response structural features, response amplitude growth region annotation, alignment channel coordinate information, and channel overlap relationship mapping; the unbound channel behavior list includes behavior number unbinding record, interference region overlapping behavior identifier, and behavior region corresponding task configuration index; the image fusion path connection control status group includes remaining path connection identifier, task instruction abort path item, and unbound channel quantity comparison result; the heterogeneous thread access mapping table includes thread call frequency sorting, path number allocation sequence, and thread and path access mapping relationship; the image fusion thread path registration record set includes path control table connection registration item, thread scheduling management synchronization information, and task path record index position.

[0013] As a further aspect of the present invention, the matching overlapping response region relationship refers to image alignment of the spatial coordinates and frequency segments of the frequency response enhancement region in the multi-channel image, and identifying the region group in the differential channel that is located within the same frame coordinate range and has overlapping frequency bands.

[0014] The verification of overlapping range refers to calling the image region coordinates corresponding to the channel behavior number in the fusion task, analyzing the coordinate overlap with the overlapping section of the fusion interference, and filtering the behavior numbers partially covered by the interference area.

[0015] As a further aspect of the present invention, the step-by-step comparison of the number of channels and the behavior number refers to comparing the channel number of the path item in the path connection table with the list of unbound channel behaviors;

[0016] The "update remaining path connection field" refers to modifying the connection status field of the path item that has not been terminated, updating the status value according to the current path task, and writing it to the corresponding field in the path connection table.

[0017] As a further aspect of the present invention, the specific steps of S1 are as follows:

[0018] S101: Obtain the channel layer data of the current frame image, extract the frequency response region structure in each channel layer, identify the image region in the layer with continuously increasing response amplitude, filter the identifiable region group by the correspondence between region number and pixel coordinate, and obtain the list of channel response enhancement regions.

[0019] S102: Based on the channel response enhancement region list, align the pixel positions of the region list in the channel, match the response position of the enhancement region in the intra-frame coordinate space of each channel matching layer, extract the repeated positions that appear simultaneously in the channel, and obtain a multi-channel response overlapping region set.

[0020] S103: Based on the multi-channel response overlapping region set, call the channel number and corresponding coordinate index in each group of overlapping regions, the frequency segment and position index position covered in the classification region, and map the region response segments according to the frame number and channel distribution to obtain the fusion interference overlapping segment identifier set.

[0021] As a further aspect of the present invention, the specific steps of S2 are as follows:

[0022] S201: Based on the coordinate data in the set of fusion interference overlapping section identifiers, obtain the channel behavior number information in the current fusion task, analyze the region position index corresponding to each number in the image channel configuration according to the task configuration rules, and obtain the channel behavior number mapping data.

[0023] S202: Call the image region coordinates in the channel behavior number mapping data, analyze the overlap relationship between the coordinates of each channel behavior region and the interference segment, match the coordinate point index and verify that there is a coverage overlap between the regions, calculate the overlap ratio of the interference behavior regions, and obtain the overlapping channel number index set.

[0024] S203: Based on the numbering information in the overlapping channel numbering index set, match the channel binding records in the task path binding list, remove the items with the same number, update the behavior participation status and remove the interference item numbers to obtain the list of non-bound channel behaviors.

[0025] As a further aspect of the present invention, the specific steps of S3 are as follows:

[0026] S301: Based on the channel number in the unbound channel behavior list, locate the path item in the fusion task path connection table that matches the number, extract the positional relationship between the channel number field and the path structure number field, and obtain the path positioning parameter set;

[0027] S302: Based on the path positioning parameter set, filter items whose number field and channel number have no matching relationship within the path item, extract the total number of channel numbers for each filter item, calculate the offset rate between the frequency of occurrence of the channel number and the total number of channels, and obtain a list of channel offset rates for aborted tasks.

[0028] S303: Based on the list of aborted task channel offset rates, disable the task transmission process of matching path items, and update the connection status field of the remaining items in the path connection table to obtain the image fusion path connection control status group.

[0029] As a further aspect of the present invention, the specific steps of S4 are as follows:

[0030] S401: Based on the path status information in the image fusion path connection control state group, remove path items whose connection status field is inactive, associate the task cache call counts corresponding to the threads in the heterogeneous thread pool, associate the path number with the thread number in a unified sequence, and obtain the thread connection path list.

[0031] S402: Based on the thread number in the thread connection path list, the number of times the thread's task cache is called is used as a reference order parameter, and the original position sequence of the path number is preserved to obtain the thread number response sequence;

[0032] S403: Based on the correspondence between the thread number response sequence and the path number, assign each thread number to the corresponding path response item, advance the pairing process between the number and the path, and obtain the heterogeneous thread access mapping table.

[0033] As a further aspect of the present invention, the specific steps of S5 are as follows:

[0034] S501: Based on the relationship between threads and path numbers in the heterogeneous thread access mapping table, query the index position in the current task path record table corresponding to each path number to be written, and locate the corresponding thread number item to obtain the path number index set.

[0035] S502: Based on the thread number and the pointed path position in the path mapping index data, insert the connection information corresponding to the thread number into the corresponding position in the path control table, and synchronously overwrite the status field content on the path item to obtain the path control filling result;

[0036] S503: Based on the changes in thread number and connection status in the path control filling result, write the connection status value into the corresponding field of the thread scheduling management table, update the execution status association between the thread and the path, and obtain the image fusion thread path registration record set.

[0037] A smart image fusion processing system based on heterogeneous computing includes:

[0038] The interference identification module acquires layer data in the current frame image channels, extracts the frequency response region structure in each channel, locates the response amplitude growth features of continuous regions in each channel, calls the pixel coordinates at the same position between channels, compares the corresponding response regions between multiple channels, filters the coordinate regions with overlapping response amplitudes, and obtains a set of identifiers for fused interference overlapping sections.

[0039] The behavior unbinding module extracts the channel behavior number information in the current fusion task based on the coordinate data in the set of overlapping interference segments of the fusion module, queries the image area position corresponding to the number in the task configuration table, analyzes the overlap between the area and the interference coordinates, removes the behavior numbers with overlapping relationships from the path binding list, and obtains a list of unbound channel behaviors.

[0040] The path control module locates the path connection table of the current fusion task based on the channel number in the unbound channel behavior list, checks the channel number content included in the path item one by one, sets the task instruction transmission status of the path with the unbound number to the aborted state, and obtains the image fusion path connection control status group.

[0041] Based on the path status information in the image fusion path connection control status group, the thread mapping module extracts the task cache call records of threads from the heterogeneous thread pool, processes the thread numbers according to the call frequency, maps the sorting results to the enabled path numbers, assigns thread numbers to participate in path response tasks, obtains the heterogeneous thread access mapping table, and transmits it to the control registration module.

[0042] Based on the relationship between threads and path numbers in the heterogeneous thread access mapping table, the control registration module extracts the position index of the associated path number from the task path record table, fills in the thread number to the corresponding position of the path control field, and simultaneously synchronizes the connection status to the thread scheduling management table to obtain the image fusion thread path registration record set.

[0043] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0044] In this invention, interference segments are calibrated and separated by behavior numbering through the identification of frequency overlap between channels and the extraction of layer response structure. Path filtering and connection status updates are performed through the linkage update of behavior number and path relationship. Resource scheduling is allocated sequentially by matching thread call frequency with path status. Thread control data is uniformly registered through path index and connection information synchronization mechanism, thereby improving interference processing capability and task control efficiency during image fusion. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 This is a schematic diagram of the steps of the present invention;

[0047] Figure 2 This is a system module diagram of the present invention.

[0048] Figure 3 This is a detailed flowchart of S1 of the present invention;

[0049] Figure 4 This is a detailed flowchart of the S2 process of the present invention;

[0050] Figure 5 This is a detailed flowchart of the S3 process of the present invention;

[0051] Figure 6 This is a detailed flowchart of the S4 process of the present invention;

[0052] Figure 7 This is a detailed flowchart of S5 of the present invention. Detailed Implementation

[0053] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0054] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0055] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent.

[0056] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0057] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0058] Please see Figure 1 This invention provides an intelligent image fusion processing method based on heterogeneous computing, comprising the following steps:

[0059] S1: Obtain the channel layer data of the current frame image, extract the frequency response region structure of each channel, identify the region with continuous increase in response amplitude in the channel, align the pixel coordinates for the corresponding positions between channels, match the overlapping relationship between the response regions of multiple channels, and obtain the fusion interference overlapping section identifier set.

[0060] S2: Based on the coordinate data in the set of overlapping interference segments, extract the number information of the channel behavior in the current fusion task, match the image area corresponding to the number in the task configuration, compare the overlap range of the interference area and the behavior area, remove the behavior number corresponding to the overlapping coordinates from the path binding list, and obtain the list of unbound channel behaviors.

[0061] S3: Based on the channel number in the unbound channel behavior list, locate the path item associated in the path connection table in the fusion task, compare the number of channels in the path with the unbound number item by item, filter the path item with the unbound number and stop the task instruction transmission, update the connection status field of the remaining path item, and obtain the image fusion path connection control status group.

[0062] S4: Based on the path status information in the image fusion path connection control state group, read the task cache call count of the thread from the heterogeneous thread pool, arrange the thread numbers in order of call frequency, select the thread numbers to participate in the connection path response allocation in the order of arrangement, and obtain the heterogeneous thread access mapping table.

[0063] S5: Based on the relationship between threads and path numbers in the heterogeneous thread access mapping table, extract the position index of the path number to be written in the current task path record table, add the connection information corresponding to the thread number to the path control table, and synchronize the connection status to the thread scheduling management table to obtain the image fusion thread path registration record set;

[0064] The fusion interference overlap section identifier set includes channel frequency response structure features, response amplitude growth area annotation, alignment channel coordinate information, and channel overlap relationship mapping. The unbound channel behavior list includes behavior number unbinding record, interference area overlapping behavior identifier, and behavior area corresponding task configuration index. The image fusion path connection control status group includes remaining path connection identifier, task instruction aborted path item, and unbound channel quantity comparison result. The heterogeneous thread access mapping table includes thread call frequency sorting, path number allocation sequence, and thread and path access mapping relationship. The image fusion thread path registration record set includes path control table connection registration item, thread scheduling management synchronization information, and task path record index position.

[0065] Please see Figure 3 The specific steps of S1 are as follows:

[0066] S101: Obtain the channel layer data of the current frame image, extract the frequency response region structure in each channel layer, identify the image region in the layer with continuously increasing response amplitude, filter the identifiable region group by the correspondence between region number and pixel coordinate, and obtain the list of channel response enhancement regions.

[0067] First, the data channels of the image frame to be processed are extracted sequentially. Each channel layer is divided into horizontal pixel vector groups according to its grayscale value matrix. The size of a single channel layer is set to 512×512 pixels, that is, there are 512 horizontal pixel vectors, numbered from v1 to v2. 512 Each vector contains 512 pixels. A one-dimensional Fast Fourier Transform is then performed on the data within each vector to obtain its frequency response component sequence. The frequency band is divided into 0Hz to 256Hz with a bandwidth of 1Hz, forming the frequency response sequence of each horizontal pixel vector in the corresponding channel. After completion, the frequency response region structure corresponding to each pixel vector in each channel layer is extracted sequentially. When identifying image regions with continuously increasing response amplitude in the layer, the judgment criterion is set as amplitude increases exceeding 5Hz continuously, and the increase in each frequency band must not be less than 5 units of amplitude value. Frequency bands that meet this criterion can be determined as continuously increasing regions. This method can obtain the start and end indices of the frequency response of multiple region segments. Then, the pixel vector numbers and frequency band ranges marked in each channel layer are combined... Each response region is assigned a unique number label, such as T1, T2, etc., and an index table is established to correspond its number with its starting coordinates and frequency range. For example, if number T5 corresponds to column 87 of channel 1, with starting coordinates (0, 86) and a frequency range of 48Hz to 63Hz, then the number, coordinates, and frequency range are entered into the table. Next, based on the channel layer number and its coordinate index, the overlap or intersection of frequency response regions in different channel layers is compared. The criterion is that two numbered regions have overlapping frequency ranges and their coordinate distance (vertical or horizontal) is less than 20 pixels. Numbered regions that meet this condition are grouped into the same response region group, and their coordinate relationships and number sets across multiple channels are recorded. For example, numbers T5, T9, T1, T2, T9, T1, T9 ... 11 If the channels are from different channels but are close in location and have overlapping frequency bands, they can form a numbered response group G1. Finally, based on the information of the numbered group, the identifiable region group is selected to obtain a list of channel response enhancement regions.

[0068] S102: Based on the list of enhanced regions in the channel response, align the pixel positions of the region list in the channel, match the response position of the enhanced region in the layer of each channel in the frame coordinate space, extract the repeated positions that appear in the channel at the same time, and obtain the set of multi-channel response overlapping regions.

[0069] First, the response enhancement regions recorded in multiple channels are mapped to a unified intra-frame coordinate space. The channel number, region number, and their corresponding pixel start and end coordinate ranges are obtained. The starting row and column positions of each region, along with its width and height, are uniformly converted into the form of the top-left and bottom-right corners, forming the rectangular boundary range of each enhancement region within a channel. Then, the boundaries of enhancement regions in all channels are mapped to unified frame coordinates. When the image frame size is set to 1024×1024 pixels, if the enhancement region range of number T1 in channel 1 is row coordinates [150, 170] and column coordinates [300, 330], then the mapped range of this region is a rectangle (150, 300) - (170, 330). At this point, it is necessary to determine whether there is a region in channel 2 that intersects with or completely covers the boundary of this region in channel 1. This judgment process is performed for each numbered region. The overlap judgment criterion is: if two regions have an overlap of more than 10 pixels in the horizontal or vertical pixel intervals... If a cross is found, it is considered a positional repetition. For this determination process, it is necessary to determine the intersection between the minimum and maximum values ​​of the row and column coordinate intervals of each pair of regions. If the numbers T1 and T7 satisfy the overlap condition, their corresponding channel numbers and coordinate values ​​are recorded as a set of overlap candidates. After the enhancement region range of all channels is determined, all regions that satisfy the overlap determination are traversed, and the number of channel numbers under the same coordinate range in each frame is counted. If the same position has enhancement region response in two or more channels, it can be identified as a multi-channel repeated response region. It is necessary to further collect all coordinate segments that satisfy this condition and their corresponding channel number sets, and combine them into a response region coordinate set. For example, if the coordinate segment [320, 335] × [760, 785] has enhancement region segments in channels 1, 3, and 4, and the positional intersection range is 20 × 25 pixels, then the corresponding number of this segment is added to the overlapping region list, and finally the multi-channel response overlapping region set is obtained.

[0070] S103: Based on the multi-channel response overlapping region set, call the channel number and corresponding coordinate index in each group of overlapping regions, the frequency segment and location index position covered in the classified region, and map the region response segments according to the frame number and channel distribution to obtain the fused interference overlapping segment identifier set.

[0071] First, read the channel number, frame number, and coordinate range of each entry in the set one by one, and locate the frame index number of each group of entries. Divide all the repeated areas under that frame number by channel, and record the spatial position range and index position of each channel. Then, locate the layer frequency response information of the corresponding channel number in the image frame. The frequency value range of the corresponding pixel segment has been divided into 5 segments in the previous stage, namely F1∈[0,4], F2∈[5,10], F3∈[11,20], F4∈[21,30], and F5∈[31,40]. For the overlapping area with channel number 2 and coordinate range of [150,180]×[200,240], it is necessary to extract the range of the frequency response value of the pixels in this area. If the segment with the highest proportion of frequency response value in the pixel is F4, then the frequency segment corresponding to this area is F4, and record the channel number as 2, frame number as 0006, frequency segment as F4, and position index as […]. [150, 180] × [200, 240], then perform the same operation on all similar regions in the frame image, and merge them in order of frame number to form a mapping list. The list items include the channel number contained in each group of overlapping regions, its corresponding frequency response segment number, and the pixel coordinate range in the frame image. This step does not involve any model call, but only needs to classify and judge the frequency segment coverage position for the existing image channel response data. If a certain coordinate segment is covered by frequency segment F3 in both channel 1 and channel 4, then it is merged and represented as a region segment: [channel 1|F3], [channel 4|F3], coordinates [420, 450] × [600, 640]. After the location region information of similar frequency response superposition in multiple channels is completely extracted, it is further grouped by channel number and sorted in order of frame number. Finally, all channel frequency segments and their position joint response segments are combined to form a response cross relationship, and the fused interference overlapping segment identifier set is obtained.

[0072] Please see Figure 4 The specific steps of S2 are as follows:

[0073] S201: Based on the coordinate data in the set of overlapping interference segment identifiers, obtain the channel behavior number information in the current fusion task, analyze the region position index corresponding to each number in the image channel configuration according to the task configuration rules, and obtain the channel behavior number mapping data.

[0074] First, determine the frame number, channel number, and region coordinate index contained in each identification information. Locate the image channel behavior list configured in the current fusion task using the channel number. Read the behavior number field contained in the channel behavior data. Associate each number with the current frame sequence number and the corresponding channel index value. Confirm the channel region range bound to that number in the image channel behavior mapping table. Extract the corresponding region position from the frame using the number as the index. For example, channel behavior number T-08 has a region position marked as [260, 300] × [420, 470] in the task configuration table. If the channel number is 3, then channel behavior number T-08 corresponds to the aforementioned coordinate position under channel number 3. Then, temporarily store the number and its bound region information in an intermediate variable. Finally, based on the fusion interference overlap segment... Coordinate values ​​are matched to find if there is a corresponding region with an intersecting number at that location. For example, if channel 3 appears in the interference segment with a region location of [265, 295] × [425, 465], and this location is completely covered by T-08, then it is confirmed that channel behavior number T-08 exists in the interference segment. The region matching process is completed for the correspondence between all interference coordinate data and channel behavior numbers in each frame. After completion, each number, its corresponding channel number, frame number, and spatial location index are written into the result set. If number T-09 is located in channel 2, and its matching region is [100, 150] × [180, 230], then its number needs to be marked in the structure entry corresponding to the region location and number in the corresponding frame. Finally, the matching relationship between all numbers and coordinate positions is numbered and output to obtain the channel behavior number mapping data.

[0075] S202: Call the image region coordinates in the channel behavior number mapping data, analyze the overlap relationship between the coordinates of each channel behavior region and the interference segment, match the coordinate point index and verify that there is overlap between regions, calculate the overlap ratio of the interference behavior region, and obtain the overlapping channel number index set.

[0076] The specific formula for calculating the overlap ratio of interference behavior regions is as follows:

[0077]

[0078] in, This represents the overlap ratio of interference behavior between region v and region y in channel b. This represents the image grayscale response value of region v in channel b at the z-th coordinate point. This represents the image grayscale response value of region y in channel b at the z-th coordinate point. This represents the image energy deviation value at the z-th point in channel b. This represents the response delay value of region v in channel b at point z. This represents the response delay value of region y in channel b at point z. This represents the length of the boundary pixels of region v in channel b. This represents the length of the boundary pixels of region y in channel b. This represents the intensity of the contour grayscale abrupt change in region v within channel b. z represents the intensity of the contour grayscale abrupt change in region y in channel b, z is the coordinate point index, and b represents the channel number being processed.

[0079] Assumption:

[0080]

[0081] Molecular part:

[0082] Point 1:

[0083]

[0084] Point 2:

[0085]

[0086] The total molecular value is: 82.158 + 65.526 = 147.684;

[0087] Denominator: |68·65|+|41―38|=4420+3=4423;

[0088] The overlap ratio is calculated as follows:

[0089]

[0090] The results show that in channel b, the overlap ratio of the interference behavior regions of region v and region y is 0.0334, which is within the standardized overlap ratio range of (0, 1) and close to 0. This indicates that the degree of overlap between the two regions in the interference segment is low and does not meet the threshold value of 0.25 for determining the significance of channel interference overlap.

[0091] The advantage of the formula is that it shows the differences between regional images. Energy deviation Difference in response latency By performing product calculations, the response accuracy of interference behavior region determination is enhanced, and a shape conflict judgment factor is introduced through path length and contour intensity, thereby enabling synchronous control of structural differences and response differences.

[0092] S203: Based on the numbering information in the overlapping channel numbering index set, match the channel binding records in the task path binding list, remove the items with the same number, update the behavior participation status and remove the interfering item numbers to obtain the list of non-bound channel behaviors.

[0093] First, locate the binding record for each number in the current task path binding list. This record stores the bidirectional association between the channel number and the behavior number. The number format can be T-03, T-04, etc. Compare it with the interference number groups listed in the index set, such as T-03, T-07, T-12, and check if the number exists in the binding list item by item. If it exists, separate the binding relationship of that number from the channel behavior item. For example, if T-07 is bound to channel 6 and points to channel behavior path B-06, then after removing this binding relationship, channel 6 will no longer participate in the fusion operation in behavior path B-06. After performing the comparison between each interference number and the binding list, adjust the movement in the behavior status field. The channel numbers of the behavior numbers are marked with a status of N (not participating), while the other numbers retain their original value of P (participating). For example, the status of number T-03 is changed to N, while T-06 remains P. Then, all behavior numbers with a status of N are removed from the channel behavior items to form a new set of numbers to be used. The behavior list is then traversed one by one according to the channel distribution. After removing the interfering numbers, the remaining numbers are arranged in the original distribution order. For example, after removing numbers T-03 and T-07, numbers T-01, T-06, and T-08 are retained to form a number set and stored under the corresponding number index of the current frame. Finally, the number data after removing the interfering behavior numbers is used as the output to obtain the list of unbound channel behaviors.

[0094] Please see Figure 5 The specific steps of S3 are as follows:

[0095] S301: Based on the channel number in the unbound channel behavior list, locate the path item in the fusion task path connection table that matches the number, extract the positional relationship between the channel number field and the path structure number field, and obtain the path positioning parameter set;

[0096] First, check if each item's number exists in the channel field column of the fusion task path connection table. This field is generally represented by the number Tn. If numbers such as T-05, T-08, and T-12 exist in the list, then match each one in the channel column of the connection table. After finding the corresponding path connection record, read the character interval or field index position between the channel number field and the path structure number field in each record. This position difference is used for subsequent mapping relationship calculations. The path structure number field generally appears as Pn or Rn. For example, T-08 matches the 3rd field in the path item, while the path structure number is the 6th field. The interval between the channel number field and the path structure number field... The value is 3. Then, in the matching items, the path structure number position of each record is scanned, and its corresponding position in the path item is located according to the order of appearance of the channel number in the list. The distance value or relative order index between the position and the path field number position is recorded, and a structural sequence of the position feature of the channel number in the path table is generated. After all the numbers are matched, the structural position indexes corresponding to all channel numbers are expanded in lexicographical order. Each item indicates the number value and its relative field position or position offset value in the path item. For example, the position offset of T-05 is 2, T-08 is 3, and T-12 is 1. Finally, based on the position information between the fields of all numbers, the path positioning parameter set is obtained.

[0097] S302: Based on the path positioning parameter set, filter items whose number field and channel number have no matching relationship within the path item, extract the total number of channel numbers for each filter item, calculate the offset rate between the frequency of occurrence of the channel number and the total number of channels, and obtain the list of channel offset rates for aborted tasks.

[0098] The formula for calculating the offset rate between the frequency of channel number occurrence and the total number of channels is as follows:

[0099]

[0100] Where, Δ i a represents the offset between the frequency of the channel number of the i-th path item and the total number of channels. ij λ represents the hit frequency of the j-th channel number in path item i in the current task. ij c represents the interference enhancement factor of channel j. ij d represents the number of times channel j appears in the remaining path items. ij e represents the offset correction term for channel j during the behavior stripping phase. i ω represents the total number of channel numbers in path item i. ij The connection weight of channel j is represented by m, and the total number of channel numbers involved in the calculation in the path item is represented by m.

[0101] Example of calculation process (path item i = 2, channel number j = 1, 2, 3):

[0102] Let: a 21 =4, a 22 =6, a 23 =5;

[0103] λ 21 =0.42, λ 22 =0.35, λ 23 =0.40;

[0104] c 21 =16,c 22 =9,c 23 =9;

[0105] d 21 =1,d 22 =2,d 23 =2;

[0106] ω 21 =1.8, ω 22 =1.5, ω 23 =1.7;

[0107] e2 = 5, m = 3;

[0108] Calculate the numerator:

[0109]

[0110] Calculate the denominator:

[0111] e2+∑ω 2j =5 + (1.8 + 1.5 + 1.7) = 5 + 5.0 = 10.0;

[0112] Calculate the difference between the means of the terms on the right:

[0113]

[0114] Substitution:

[0115]

[0116] Termination judgment benchmark threshold: The termination threshold τ = 0.28 is set by accumulating the mean and standard deviation of the training sample offset rate. The sample mean is 0.22 and the standard deviation is 0.06.

[0117] Interpretation of results and numerical significance:

[0118] The calculated offset rate Δ2 = 0.393 is greater than τ = 0.28. It is determined that the second path item should be included in the list of offset rates for the aborted task channel, and the number should be removed from the path connection instruction queue. The result will be used as the basis for updating the image fusion path control status group.

[0119] Explanation of the innovative aspects of the formula:

[0120] The advantage of the formula is that by introducing the interference enhancement factor, behavior stripping correction term and connection weight in joint calculation, it can measure the dynamic fluctuation characteristics of channel number, make quantitative judgment on channel stability shift, and improve the accuracy of path task termination judgment.

[0121] S303: Based on the list of aborted task channel offset rates, disable the task transmission process of matching path items, and update the connection status field of the remaining items in the path connection table to obtain the image fusion path connection control status group.

[0122] First, the path record with the matching channel number is searched in the path connection table. The channel number value in the matching field is compared with the offset list item. If the channel number field Tn of the path record exists in the list, the transmission status flag field of the corresponding record is set to disabled, for example, the status field value is changed from "1" to "0" to suspend data transmission of the path in the current frame or the current task cycle. Then, all path records that do not match the list number are read one by one. Their fields include path number Pm, connection number Ck and connection status field Sv. For these records, it is determined whether their connection status field is valid. The state range, i.e. whether Sv is within a preset range, such as between 0.3 and 0.7, is determined. If it is not within this range, a correction operation is performed on its state field, and the state value is readjusted to the default connection state identifier 0.5. Otherwise, the original state is maintained. Furthermore, a one-to-one mapping is formed between the path number field and the state field in these remaining path records, and the updated state value is used to overwrite the original state field in the path connection table to update the connection state field. Finally, all path records with updated state fields are integrated within the current fusion task scope, and the path number and corresponding connection state value of all path items are summarized and output to obtain the image fusion path connection control state group.

[0123] Please see Figure 6 The specific steps of S4 are as follows:

[0124] S401: Based on the path status information in the image fusion path connection control state group, remove path items whose connection status field is inactive, associate the task cache call count corresponding to the thread in the heterogeneous thread pool, associate the path number with the thread number in a unified sequence, and obtain the thread connection path list.

[0125] First, iterate through each path item in the path status field set, retrieving its connection status value Cs. For path items with values ​​of 0 or negative, remove the path record directly from the path set list. After removal, retain the set R-act of all path items with status values ​​greater than 0. Then, for the path number Ri in this set, match it with the thread number Tj in the heterogeneous thread pool, and obtain the task cache call count Fj corresponding to the thread number. This call count is an accumulated record of the thread's historical tasks, which can be quickly read through the task index number. For example, the call count of thread number T-5 is 87 times, and the call count of thread number T-3 is 124 times. Further, merge the path number Ri and the call count Fj according to the index of thread number Tj, that is, use the path number as the left index, the thread number as the middle item, and the call count as the right value to form a one-to-one corresponding triple sequence. Finally, output the set in sequence form, presenting the path number and thread number as table entries in the same structure, to obtain the thread connection path list.

[0126] S402: Based on the thread number in the thread connection path list, the number of times the thread's task cache is called is used as a reference order parameter, and the original position sequence of the path number is preserved to obtain the thread number response sequence;

[0127] First, extract the thread ID Tk and its task cache call count Fk corresponding to each path item in the list. Using the call count as a sorting reference parameter, read the corresponding items according to the original order of the path IDs Pk, while preserving the binding relationship between the thread ID Tk and the path ID Pk in this order. Without disrupting the path sequence structure, compare the call counts of Fk corresponding to Tk in order and construct a sequence mapping index. Then, form an ID-sequence mapping item with each thread ID and its sorting number, as a set of response level mapping relationships. For example, if the original path order is P1, P2, P3, P4, and the call counts of T1, T2, T3, T4 are 92, 135, 84, and 110 respectively, then the position order of P1 to P4 is maintained, and a call count index table of T1→92, T2→135, T3→84, and T4→110 is generated. Based on this index, T1, T2, T3, and T4 are included in the response number sequence group according to the path order, while the number order is not changed, and finally the thread number response sequence is obtained.

[0128] S403: Based on the correspondence between thread number response sequence and path number, assign each thread number to the corresponding path response item, advance the pairing process between number and path, and obtain the heterogeneous thread access mapping table.

[0129] First, the thread number Ti and its sequence number are read sequentially from the thread number response sequence. Then, the path number Pi in the original path connection order is compared, and each thread number Ti and its position are bound to the path item pointed to by Pi. During this process, the one-to-one correspondence between threads and paths is maintained. Subsequently, the paired number pairs are checked item by item to ensure that thread Ti is not reused in multiple path items. For example, when the path number sequence is P1 to P4, the thread number sequence is T3, T1, T4, T2. The pairings T3→P1, T1→P2, T4→P3, and T2→P4 are completed respectively. The order of thread numbers is determined by the call frequency and does not involve position changes. This process does not involve structural changes to the path content, but only adjusts the application position of the thread number. Finally, the above mapping relationship between numbers and paths is solidified into a number pair mapping list. Through one-to-one mapping, these are organized into a unified structure list for subsequent thread access control, ultimately resulting in a heterogeneous thread access mapping table.

[0130] Please see Figure 7 The specific steps of S5 are as follows:

[0131] S501: Based on the relationship between threads and path numbers in the heterogeneous thread access mapping table, query the index position in the current task path record table for each path number to be written, locate the corresponding thread number item, and obtain the path number index set.

[0132] First, each pair of thread IDs Tj and path IDs Pj is sequentially read from the mapping table. While reading Pj, it serves as the primary index entry. The corresponding index position Ij for that path ID is then retrieved from the current task path record table. To ensure accurate identification, fuzzy matching is not used. Instead, Pj is compared item by item with the path ID field in the record table. Items with completely identical field values ​​are recorded in the table. Simultaneously, Tj is used as its location identifier. The thread field corresponding to Ij in the current task path record table is then checked to confirm whether a one-to-one match exists between Tj and Pj. If a pairing relationship exists, the path number Pj and its index Ij in the task path record table are included in the valid matching set. For example, if there are binding pairs T05→P12, T06→P13, and T07→P16 in the access mapping table, then the path numbers P12, P13, and P16 are retrieved from the path record table. Assuming their position indices in the table are 4, 5, and 9, and their corresponding thread fields are T05, T06, and T07, respectively, the binding relationship requirement is met. Then, 4, 5, and 9 are recorded as path number indices in the set, and finally, the path number index set is obtained.

[0133] S502: Based on the thread number and the pointed path position in the path mapping index data, insert the connection information corresponding to the thread number into the corresponding position in the path control table, and synchronously overwrite the status field content on the path item to obtain the path control filling result;

[0134] First, read the thread ID and its corresponding path position index combination Ti→Ij from the index data. After reading Ti, confirm its corresponding path ID Pj in the heterogeneous thread access mapping table. Then, check the structure of the path control table to confirm the actual row number of index Ij in the table. Place Ti as the item to be written in the thread ID field of the row corresponding to index Ij in the path control table. At the same time, enter the connection information section of the path control table and fill in the call information value Vi of the thread. Vi is the ratio of the number of times the thread successfully participated in path execution to the number of times it bound to the path in the past five frames of tasks. If Ti is T08, its five-frame intrinsic parameters With a connection count of 3 and a path binding count of 5, Vi = 0.6. In the path control table, enter Vi = 0.6 in the corresponding row T08. Next, jump to the status field section of the path control table. Based on the current connection status value Sj of path number Pj, update the field. If Sj is "undetermined," convert it to "bound." If the original value is "idle," retain the original value. If it is "occupied," mark it as conflicted. Through the above logic, fill the corresponding path number position with the thread number and connection information, and perform conditional replacement processing on the status field to finally obtain the path control filling result.

[0135] S503: Based on the changes in thread number and connection status in the path control filling result, write the connection status value into the corresponding field of the thread scheduling management table, update the execution status association between the thread and the path, and obtain the image fusion thread path registration record set;

[0136] First, the thread number Tk and connection status field Sk corresponding to each path number Pk in the path control table are read line by line. When a change in the value of Sk is detected, it is recognized that the path has been bound to a thread. The thread index area in the thread scheduling management table is called to find the record row Ik corresponding to Tk. The connection status field area of ​​that row is entered, and the current value of Sk is written to overwrite the original field value. At the same time, the path number Pk occupied by the thread in the path control table is located and recorded in the execution path field of the thread scheduling management table. When a thread is bound to multiple path numbers, the corresponding path numbers are appended in the order of writing. During this process, to ensure that the binding between each written path and the thread is valid, it is necessary to check whether the value of Sk is "bound". If it is "conflicted" or "idle", the binding is checked. If the condition is "", then the writing is skipped. This judgment is completed by directly matching the characters of the field values ​​without introducing a judgment formula. If Tk is T12, Sk is "bound", and Pk is P38, then the connection status field of the corresponding record row of T12 in the thread scheduling management table is updated to "bound", and P38 is appended to its path field list. At the same time, the current operation frame number Fn is recorded for subsequent call statistics. If Fn is the current frame 75, then the frame number is marked "P38@75" in the path field record. The entire process covers all entries in the path control table and repeats the operation until the last row ends. Through this process, each thread number and its bound path number form a one-to-one correspondence in the thread scheduling management table, and finally the image fusion thread path registration record set is obtained.

[0137] Please see Figure 2 A smart image fusion processing system based on heterogeneous computing, comprising:

[0138] The interference identification module acquires layer data in the current frame image channels, extracts the frequency response region structure in each channel, locates the response amplitude growth features of continuous regions in each channel, calls the pixel coordinates at the same position between channels, compares the corresponding response regions between multiple channels, filters the coordinate regions with overlapping response amplitudes, and obtains a set of identifiers for fused interference overlapping sections.

[0139] The behavior unbinding module extracts the channel behavior number information in the current fusion task based on the coordinate data in the set of overlapping interference segment identifiers. It then queries the image area location corresponding to the number in the task configuration table, analyzes the overlap between the area and the interference coordinates, removes the behavior numbers with overlapping relationships from the path binding list, and obtains a list of unbound channel behaviors.

[0140] The path control module locates the path connection table of the current fusion task based on the channel number in the unbound channel behavior list, checks the channel number content included in each path item, sets the task instruction transmission status of the unbound number path to the aborted state, and obtains the image fusion path connection control status group.

[0141] The thread mapping module extracts the task cache call records of threads from the heterogeneous thread pool based on the path status information in the image fusion path connection control status group, processes the thread numbers in the order of call frequency, maps the sorting results to the enabled path numbers, assigns thread numbers to participate in path response tasks, obtains the heterogeneous thread access mapping table, and passes it to the control registration module.

[0142] The control registration module extracts the position index of the associated path number from the task path record table based on the relationship between the thread and path number in the heterogeneous thread access mapping table, fills in the thread number to the corresponding position of the path control field, and synchronizes the connection status to the thread scheduling management table to obtain the image fusion thread path registration record set.

[0143] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An intelligent image fusion processing method based on heterogeneous computing, characterized in that, Includes the following steps: S1: Obtain the current frame image channel layer data, extract the frequency response region structure of each channel, align the positions between each channel, match the overlapping response region relationship, and obtain the fusion interference overlapping section identifier set; S2: Based on the coordinate data in the set of fused interference overlapping section identifiers, match the channel behavior number with the task image region, check the overlapping range, remove the corresponding number, and obtain the list of unbound channel behaviors; S3: Based on the channel number in the unbound channel behavior list, locate the corresponding path item in the path connection table, compare the channel quantity with the behavior number item by item, filter out paths including multiple unbound numbers and stop the instruction, update the remaining path connection field, and obtain the image fusion path connection control status group. S4: Based on the path status information in the image fusion path connection control state group, read the number of thread cache calls, sort the numbers according to the call frequency, select threads to participate in the connection, and obtain the heterogeneous thread access mapping table. S5: Based on the numbering relationship in the heterogeneous thread access mapping table, locate the path position in the task path record table, write the connection information and synchronize the thread state to obtain the image fusion thread path registration record set.

2. The intelligent image fusion processing method based on heterogeneous computing according to claim 1, characterized in that, The fusion interference overlap segment identifier set includes channel frequency response structural features, response amplitude growth region annotation, alignment channel coordinate information, and channel overlap relationship mapping. The unbound channel behavior list includes behavior number unbinding records, interference region overlapping behavior identifiers, and behavior region corresponding task configuration indexes. The image fusion path connection control status group includes remaining path connection identifiers, task instruction aborted path items, and unbound channel quantity comparison results. The heterogeneous thread access mapping table includes thread call frequency sorting, path number allocation sequence, and thread and path access mapping relationship. The image fusion thread path registration record set includes path control table connection registration items, thread scheduling management synchronization information, and task path record index positions.

3. The intelligent image fusion processing method based on heterogeneous computing according to claim 1, characterized in that, The matching overlapping response region relationship refers to image alignment of the spatial coordinates and frequency segments of the frequency response enhancement region in the multi-channel image, and identification of the region group in the differential channel that is located within the same frame coordinate range and has overlapping frequency bands; The verification of overlapping range refers to calling the image region coordinates corresponding to the channel behavior number in the fusion task, analyzing the coordinate overlap with the overlapping section of the fusion interference, and filtering the behavior numbers partially covered by the interference area.

4. The intelligent image fusion processing method based on heterogeneous computing according to claim 1, characterized in that, The item-by-item comparison of channel quantity and behavior number refers to comparing the channel number of the path item in the path connection table with the list of unbound channel behaviors; The "update remaining path connection field" refers to modifying the connection status field of the path item that has not been terminated, updating the status value according to the current path task, and writing it to the corresponding field in the path connection table.

5. The intelligent image fusion processing method based on heterogeneous computing according to claim 1, characterized in that, The specific steps of S1 are as follows: S101: Obtain the channel layer data of the current frame image, extract the frequency response region structure in each channel layer, identify the image region in the layer with continuously increasing response amplitude, filter the identifiable region group by the correspondence between region number and pixel coordinate, and obtain the list of channel response enhancement regions. S102: Based on the channel response enhancement region list, align the pixel positions of the region list in the channel, match the response position of the enhancement region in the intra-frame coordinate space of each channel matching layer, extract the repeated positions that appear simultaneously in the channel, and obtain a multi-channel response overlapping region set. S103: Based on the multi-channel response overlapping region set, call the channel number and corresponding coordinate index in each group of overlapping regions, the frequency segment and position index position covered in the classification region, and map the region response segments according to the frame number and channel distribution to obtain the fusion interference overlapping segment identifier set.

6. The intelligent image fusion processing method based on heterogeneous computing according to claim 1, characterized in that, The specific steps of S2 are as follows: S201: Based on the coordinate data in the set of fusion interference overlapping section identifiers, obtain the channel behavior number information in the current fusion task, analyze the region position index corresponding to each number in the image channel configuration according to the task configuration rules, and obtain the channel behavior number mapping data. S202: Call the image region coordinates in the channel behavior number mapping data, analyze the overlap relationship between the coordinates of each channel behavior region and the interference segment, match the coordinate point index and verify that there is a coverage overlap between the regions, calculate the overlap ratio of the interference behavior regions, and obtain the overlapping channel number index set. S203: Based on the numbering information in the overlapping channel numbering index set, match the channel binding records in the task path binding list, remove the items with the same number, update the behavior participation status and remove the interference item numbers to obtain the list of non-bound channel behaviors.

7. The intelligent image fusion processing method based on heterogeneous computing according to claim 6, characterized in that, The specific steps for S3 are as follows: S301: Based on the channel number in the unbound channel behavior list, locate the path item in the fusion task path connection table that matches the number, extract the positional relationship between the channel number field and the path structure number field, and obtain the path positioning parameter set; S302: Based on the path positioning parameter set, filter items whose number field and channel number have no matching relationship within the path item, extract the total number of channel numbers for each filter item, calculate the offset rate between the frequency of occurrence of the channel number and the total number of channels, and obtain a list of channel offset rates for aborted tasks. S303: Based on the list of aborted task channel offset rates, disable the task transmission process of matching path items, and update the connection status field of the remaining items in the path connection table to obtain the image fusion path connection control status group.

8. The intelligent image fusion processing method based on heterogeneous computing according to claim 1, wherein the specific steps of S4 are as follows: S401: Based on the path status information in the image fusion path connection control state group, remove path items whose connection status field is inactive, associate the task cache call counts corresponding to the threads in the heterogeneous thread pool, associate the path number with the thread number in a unified sequence, and obtain the thread connection path list. S402: Based on the thread number in the thread connection path list, the number of times the thread's task cache is called is used as a reference order parameter, and the original position sequence of the path number is preserved to obtain the thread number response sequence; S403: Based on the correspondence between the thread number response sequence and the path number, assign each thread number to the corresponding path response item, advance the pairing process between the number and the path, and obtain the heterogeneous thread access mapping table.

9. The intelligent image fusion processing method based on heterogeneous computing according to claim 1, characterized in that, The specific steps of S5 are as follows: S501: Based on the relationship between threads and path numbers in the heterogeneous thread access mapping table, query the index position in the current task path record table corresponding to each path number to be written, and locate the corresponding thread number item to obtain the path number index set. S502: Based on the thread number and the pointed path position in the path mapping index data, insert the connection information corresponding to the thread number into the corresponding position in the path control table, and synchronously overwrite the status field content on the path item to obtain the path control filling result; S503: Based on the changes in thread number and connection status in the path control filling result, write the connection status value into the corresponding field of the thread scheduling management table, update the execution status association between the thread and the path, and obtain the image fusion thread path registration record set.

10. An intelligent image fusion processing system based on heterogeneous computing, characterized in that, The system is used to implement the intelligent image fusion processing method based on heterogeneous computing as described in any one of claims 1-9, and the system comprises: The interference identification module acquires layer data in the current frame image channels, extracts the frequency response region structure in each channel, locates the response amplitude growth features of continuous regions in each channel, calls the pixel coordinates at the same position between channels, compares the corresponding response regions between multiple channels, filters the coordinate regions with overlapping response amplitudes, and obtains a set of identifiers for fused interference overlapping sections. The behavior unbinding module extracts the channel behavior number information in the current fusion task based on the coordinate data in the set of overlapping interference segments of the fusion module, queries the image area position corresponding to the number in the task configuration table, analyzes the overlap between the area and the interference coordinates, removes the behavior numbers with overlapping relationships from the path binding list, and obtains a list of unbound channel behaviors. The path control module locates the path connection table of the current fusion task based on the channel number in the unbound channel behavior list, checks the channel number content included in the path item one by one, sets the task instruction transmission status of the path with the unbound number to the aborted state, and obtains the image fusion path connection control status group. Based on the path status information in the image fusion path connection control status group, the thread mapping module extracts the task cache call records of threads from the heterogeneous thread pool, processes the thread numbers according to the call frequency, maps the sorting results to the enabled path numbers, assigns thread numbers to participate in path response tasks, obtains the heterogeneous thread access mapping table, and transmits it to the control registration module. Based on the relationship between threads and path numbers in the heterogeneous thread access mapping table, the control registration module extracts the position index of the associated path number from the task path record table, fills in the thread number to the corresponding position of the path control field, and simultaneously synchronizes the connection status to the thread scheduling management table to obtain the image fusion thread path registration record set.

Images