Fiber distribution outlet recognition method, system and device based on color space fusion

Abstract

The application discloses a kind of based on color space fusion's optical fiber distribution socket identification method, system and equipment, it is related to computer vision and optical fiber distribution socket identification field, the method includes to target optical fiber distribution box picture is handled, obtains the corresponding several first target regions of target optical fiber distribution box picture;First target region is the region possibly existing red fiber insertion dust cap;Each first target region is input into the color space fusion network trained in advance, obtains several second target regions;Second target region is the region existing red fiber insertion dust cap;The Euclidean distance of each second target region relative to other second target region is calculated by Minkowski distance formula;According to the hierarchical clustering of Euclidean distance to each second target region obtains clustering result region;Clustering result region is the unplug fiber port position region in target optical fiber distribution box picture.The application can automatically identify the unplug fiber port position in optical fiber distribution box picture.

Description

Technical Field

[0001] This invention relates to the fields of computer vision and fiber optic cable connector recognition technology, and in particular to a method, system and device for fiber optic cable connector recognition based on color space fusion. Background Technology

[0002] Fiber optic technology is now ubiquitous, bringing convenience to users but also creating inconvenience for operators in managing and maintaining fiber optic distribution boxes. Currently, most management of fiber optic distribution boxes relies on manual labeling or visual identification of the corresponding fiber optic ports, which reduces work efficiency and makes it easy to miss some ports during inspection.

[0003] Because there are many users of fiber optic cables, such as Figure 1 As shown, this can easily lead to a situation where multiple fiber optic cables in a single fiber optic distribution box block unplugged fiber ports. This can cause errors in counting or omissions when maintenance personnel are monitoring and tracking fiber plugging, making it inconvenient to update the fiber optic resource management system data later and resulting in a waste of manpower and resources.

[0004] Because the unplugged fiber optic distribution boxes have red dust caps, color features can be used for identification. Currently, most methods for fiber plug identification convert RGB images to HSV color space images, leveraging the similarity between HSV color space and human color perception to identify the red dust caps. However, due to variations in lighting and angle in real-world photography, the accuracy of fiber plug identification is not high in complex situations. Furthermore, if fiber optic cables obstruct the unplugged area, effective detection fails, leading to missed plug counts. Summary of the Invention

[0005] Therefore, the purpose of this invention is to provide a method, system, and device for identifying fiber optic cable connectors based on color space fusion.

[0006] To achieve the above objectives, the present invention provides the following solution:

[0007] In a first aspect, the present invention provides a method for identifying fiber optic cable connectors based on color space fusion, comprising:

[0008] Obtain an image of the target fiber optic distribution box; the target fiber optic distribution box is a fiber optic distribution box with markers affixed to each of its four corners; the markers are used to convert the location of the fiber optic distribution box area in the camera coordinate system to the location of the fiber optic distribution box area in the world coordinate system; the image of the target fiber optic distribution box is an RGB format image.

[0009] The target optical fiber distribution box picture is processed to obtain a plurality of first target regions corresponding to the target optical fiber distribution box picture; the first target region is a region in which a red fiber insertion dust cap may exist;

[0010] Each first target region is input into a pre-trained color space fusion network to obtain a plurality of second target regions; the second target region is a region in which a red fiber insertion dust cap exists;

[0011] The Euclidean distance of each second target region relative to other second target regions is calculated by using a Minkowski distance formula;

[0012] The second target regions are hierarchically clustered according to the Euclidean distance to obtain a plurality of clustering result regions; the clustering result region is an uninserted fiber port position region in the target optical fiber distribution box picture.

[0013] In a second aspect, the present application provides a fiber distribution port recognition system based on color space fusion, comprising:

[0014] A target optical fiber distribution box picture acquisition module is configured to acquire a target optical fiber distribution box picture; the target optical fiber distribution box is an optical fiber distribution box with mark points attached to four corners; the mark points are used to convert the position of the optical fiber distribution box region in the camera coordinate system into the position of the optical fiber distribution box region in the world coordinate system; and the target optical fiber distribution box picture is an RGB format picture;

[0015] A first target region determination module is configured to process the target optical fiber distribution box picture to obtain a plurality of first target regions corresponding to the target optical fiber distribution box picture; the first target region is a region in which a red fiber insertion dust cap may exist;

[0016] A second target region determination module is configured to input each first target region into a pre-trained color space fusion network to obtain a plurality of second target regions; the second target region is a region in which a red fiber insertion dust cap exists;

[0017] An Euclidean distance calculation module is configured to calculate the Euclidean distance of each second target region relative to other second target regions by using a Minkowski distance formula;

[0018] A red fiber insertion dust cap region determination module is configured to hierarchically cluster the second target regions according to the Euclidean distance to obtain a plurality of clustering result regions; the clustering result region is an uninserted fiber port position region in the target optical fiber distribution box picture.

[0019] In a third aspect, the present application provides an electronic device comprising a memory and a processor, the memory being configured to store a computer program, and the processor being configured to execute the computer program to cause the electronic device to perform the method for recognizing a fiber distribution port based on color space fusion according to the first aspect.

[0020] According to the embodiments of the present application, the following technical effects are achieved.

[0021] Therefore, the present application can automatically recognize the positions of the un-plugged fiber ports in the photographed picture of the fiber distribution cabinet and count the number of the un-plugged fiber ports, and can feed back the detection and counting results to the operation and maintenance personnel in real time, so as to facilitate the operation and maintenance personnel to supervise and count the use of the fiber distribution cabinet. BRIEF DESCRIPTION OF DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0023] Figure 1 A physical diagram of the fiber distribution frame provided by the embodiments of the present application is shown in the figure.

[0024] Figure 2 A use condition diagram of the fiber distribution cabinet under normal conditions provided by the embodiments of the present application is shown in the figure.

[0025] Figure 3 A flowchart of the method for recognizing a fiber distribution port based on color space fusion provided by the embodiments of the present application is shown in the figure.

[0026] Figure 4 A perspective transformation principle diagram provided by the embodiments of the present application is shown in the figure.

[0027] Figure 5 A comparison diagram before and after perspective transformation provided by the embodiments of the present application is shown in the figure.

[0028] Figure 6 A structure diagram of the method for recognizing a fiber distribution port based on color space fusion provided by the embodiments of the present application is shown in the figure. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0030] In order to make the above objectives, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0031] Embodiment one

[0032] As shown in the figure, the optical fiber distribution frame includes a plurality of optical fiber distribution boxes, wherein the optical fiber distribution box includes a plurality of optical fiber jacks and a plurality of optical fiber lines; as shown in the figure, the red fiber insertion dust cap is not inserted with an optical fiber, and the optical fiber line includes a yellow optical fiber line and a blue optical fiber line. In actual application, a label paper will also be worn on the optical fiber line, which brings certain difficulty to the actual identification task. Figure 1 Figure 2 As shown in the figure, the optical fiber distribution frame includes a plurality of optical fiber distribution boxes, wherein the optical fiber distribution box includes a plurality of optical fiber jacks and a plurality of optical fiber lines; as shown in the figure, the red fiber insertion dust cap is not inserted with an optical fiber, and the optical fiber line includes a yellow optical fiber line and a blue optical fiber line. In actual application, a label paper will also be worn on the optical fiber line, which brings certain difficulty to the actual identification task.

[0033] As shown in the figure, the optical fiber distribution frame includes a plurality of optical fiber distribution boxes, wherein the optical fiber distribution box includes a plurality of optical fiber jacks and a plurality of optical fiber lines; as shown in the figure, the red fiber insertion dust cap is not inserted with an optical fiber, and the optical fiber line includes a yellow optical fiber line and a blue optical fiber line. In actual application, a label paper will also be worn on the optical fiber line, which brings certain difficulty to the actual identification task. Figure 3

[0034] Step 100: obtaining a target optical fiber distribution box picture; the target optical fiber distribution box is an optical fiber distribution box with mark points pasted at four corners. The mark point is used to convert the position of the optical fiber distribution box region in the camera coordinate system into the position of the optical fiber distribution box region in the world coordinate system; the target optical fiber distribution box picture is an RGB format picture.

[0035] One example: the user uses the mobile application deployed on the Android end to take the optical fiber distribution box picture, and clicks the picture upload button. Through the Flask lightweight Web application framework in Python, the mobile end sends a Post request to the server and uploads the picture data, so as to perform the identification task of the use condition of the optical fiber distribution box.

[0036] As shown in the figure, the four corners of the optical fiber distribution box are pasted with purple marks, which are used for perspective transformation of the optical fiber distribution box pictures taken under different visual angles to determine the use condition of each fiber insertion port in the optical fiber distribution box under a fixed visual angle. Figure 2 The general formula of general perspective transformation is as follows:

[0037]

[0038]

[0039] As shown in the figure, the coordinates of the purple mark points at the four corners of the optical fiber distribution box in the original picture are obtained, assuming that one of the coordinate points is (u, v), and the picture coordinates after perspective transformation are (x, y). The transformation matrix is 3x3 form, and then: Figure 4

[0040] ​​​​Suppose the coordinates of the four corner markers of the original picture are (u0, v0), (u1, v1), (u2, v2), and (u3, v3), and after perspective transformation, they are (x0, y0), (x1, y1), (x2, y2), and (x3, y3). Simply put, it is a transformation from a quadrilateral to a square, and the four sets of corresponding points of the transformation can be represented as: (u0, v0)→(x0, y0), (u1, v1)→(x1, y1), (u2, v2)→(x2, y2), and (u3, v3)→(x3, y3).

[0041] Step 200: processing the target fiber distribution box picture to obtain a plurality of first target regions corresponding to the target fiber distribution box picture; the first target region is a region where a red fiber insertion dust cap may exist.

[0042] In the embodiment of the application, step 200 specifically includes: performing preprocessing operations such as transmission transformation and color space conversion on the target fiber distribution box picture to obtain a plurality of first target regions corresponding to the target fiber distribution box picture.

[0043] This step 200 further includes: (1) extracting the two-dimensional coordinates of the marker points at the four corners of the target fiber distribution box picture, performing perspective transformation on the two-dimensional coordinates of the marker points, and obtaining a fiber distribution box picture under a front view angle as shown in Figure 5 (2) performing LAB color space conversion on the fiber distribution box picture under the front view angle to obtain a fiber distribution box picture converted into the LAB color space, and calculating a LAB color histogram according to the A channel of the fiber distribution box picture converted into the LAB color space, and then calculating the best percentage threshold suitable for the current red dust cap region according to the LAB color histogram. (3) performing binaryzation processing on the best percentage threshold to obtain all regions in the fiber distribution box picture under the front view angle where a red fiber insertion dust cap may exist.

[0044] Step 300: inputting each first target region into a pre-trained color space fusion network to obtain a plurality of second target regions; the second target region is a region where a red fiber insertion dust cap exists.

[0045] The building process of the pre-trained color space fusion network is as follows:

[0046] (1) 1000 pictures of fiber distribution box port conditions are shot and collected, i.e. 1000 original fiber distribution box sample pictures; the original fiber distribution box sample pictures are RGB format pictures. As shown in Figure 1 the picture content is single fiber insertion port region information, and the label of each sample picture region where a red fiber insertion dust cap exists is marked as 1, and otherwise as 0.

[0047] (2) Convert each original fiber distribution box sample picture into LAB color space, YUV color space and YCrCb color space, and take out the A channel in the LAV color space, the V channel in the YUV color space and the Cr channel in the YCrCb color space, respectively, and fuse the V channel, the Cr channel and the A channel to obtain a VCrA color space sample picture.

[0048] (3) Construct a convolutional neural network based on DenseBlock; the convolutional network is used to detect the fiber port situation in the region.

[0049] (4) Input the VCrA color space sample picture and the original RGB color space sample picture (i.e. the original fiber distribution box sample picture) into the model composed of the convolutional neural network based on DenseBlock respectively to train, and obtain the first model and the second model after training.

[0050] Step 300 specifically includes:

[0051] (1) Convert each first target region into LAB color space, YUV color space and YCrCb color space, respectively, and take out the A channel in the LAV color space, the V channel in the YUV color space and the Cr channel in the YCrCb color space, and then fuse the V channel, the Cr channel and the A channel to obtain a VCrA color space picture corresponding to each target region.

[0052] (2) Input each target region and the corresponding VCrA color space picture into the pre-trained color space fusion network to obtain a plurality of second target regions.

[0053] An example: convert the region information that may exist red fiber insertion dust cap obtained in step 200 into VCrA color space, and input the VCrA color space sample and the original RGB color space sample into the first model 1 and the second model 2 respectively.

[0054] Using the correlation of red, blue and yellow in each color space, the weight of the VCrA color space sample and the original RGB color space sample is obtained, and the classification results in the first model 1 and the second model 2 are fused to obtain the output result.

[0055] The region with the output result of 1 is retained to the result region set, and all regions in the region set are regions with red fiber insertion dust cap, and the remaining regions with the output result of 0 are removed.

[0056] Step 400: Calculate the Euclidean distance of each second target region relative to other second target regions by the Minkowski distance formula.

[0057] Step 500: hierarchical clustering is performed on each second target region according to the Euclidean distance to obtain a plurality of clustering result regions; the clustering result region is an un-plugged port position region in the target fiber distribution box picture.

[0058] Step 500 specifically comprises: first, regarding each second target region as a cluster, calculating the distance value between each cluster, and regarding the distance value between each cluster as the similarity. Secondly, a hierarchical nested clustering tree is created by calculating the similarity between different category clusters, wherein each cluster belongs to a different category, and different category clusters are the lowest layer of the tree, and the smaller the distance between each cluster, the higher the similarity. Then, a bottom-up method is adopted, a suitable threshold is selected as the final retention condition, and two regions smaller than the suitable threshold are retained for category combination, that is, a merging operation, and iteration is repeated until all distance values are greater than the suitable threshold.

[0059] Further, the method provided by the embodiment of the application further comprises:

[0060] Each clustering result region is marked with a rectangle by using the rectangle function of the opencv library and counted.

[0061] In the embodiment of the application, the clustering result region is marked and counted, the marked picture is transmitted to the mobile terminal as an output result, and the left upper corner position coordinates, the right lower corner position coordinates of each marked rectangular frame in the output picture and the counting result are saved to the background database.

[0062] The color space fusion and hierarchical clustering method provided by the application can ensure that the use of the fiber distribution box can be fed back to the operation and maintenance personnel in real time, and has stable and high detection accuracy in some complex environments such as fiber line disorder; the color space fusion method involved in the application can reduce the influence of environmental noise on the recognition of the red fiber insertion cap under different light brightness by converting the picture into LAB, YUV and YCrCb color spaces and fusing the split V channel, Cr channel and A channel, and can have high recognition accuracy by performing multi-model fusion classification with the original RGB picture; when performing the clustering task, the hierarchical clustering method can effectively avoid the shielding of the red fiber insertion cap by the fiber line, thereby realizing the counting of the result region.

[0063] Compared with the existing method, the application is more stable for recognizing the use of the fiber distribution box in a complex environment, has high recognition accuracy, enhances the illumination robustness, and has high improvement in practicality and efficiency.

[0064] Embodiment two

[0065] To further illustrate the method provided by the present application, Example Two is provided for illustration. The method provided by Example Two specifically includes:

[0066] First, the obtained input picture is subjected to a transmission transformation, color space conversion and other preprocessing operations. The input picture is an RGB format picture.

[0067] Second, the LAB color histogram of the input picture subjected to the preprocessing operation is obtained, the Lab color space a is set as a single-channel picture 1, the single-channel picture 1 is subjected to a binaryzation processing through a preset percentage threshold to obtain each red target region S i , i is the index of each region, which is set as S = {S1, S2, S3…, S n}.

[0068] Then, each region S i in the region set S obtained after the preprocessing is input into the trained model 1 and model 2 respectively, and the output results of each model are weighted and fused to obtain a final result set P, that is, P = {P1, P2, P3…, P n}, wherein the region i with a prediction of 1 has P i = 1.

[0069] Next, the Euclidean distance of each result region in the final result set P relative to other result regions is calculated through the Minkowski distance formula.

[0070] Finally, hierarchical clustering is performed on each result region in the final result set P according to the Euclidean distance to obtain a plurality of clustering result regions; the clustering result region is a red fiber insertion dustproof cap region in the input picture.

[0071] Further, the region in the reserved picture where the fiber insertion dustproof cap exists includes the following steps:

[0072]

[0073] Wherein x i , x j are the region point position sets of the two regions S i , S j to be calculated, x iu , x ju are each coordinate point of the region point position set, n represents the total number of region sets, and p represents the dimension of the space. Generally, when p = 1, it is Manhattan distance, when p = 2, it is Euclidean distance, and when p = ∞, it is Chebyshev distance. In this embodiment, the value of p is 2.

[0074] The region S i with P i = 1 in the reserved result set PSet it as a reserved area set S', calculate the Euclidean distance of each area in the area set S' to other reserved areas S i ' by Minkowski distance formula. i-1 Further, the hierarchical clustering of each result area comprises the following steps:

[0075] Select a suitable threshold as the final reservation condition, and retain one of the two areas less than the suitable threshold.

[0076] Further, the hierarchical clustering of each result area comprises the following steps:

[0077] The result set P={P1, P2, P3…, P n} as input, and use single-link algorithm according to cluster distance measurement function, and the nearest sample distance in two clusters is the direct distance of the two clusters.

[0078] Calculate the effective contour information by Euclidean distance, and remove the invalid contour information, and finally output the current number of fiber insertion ports of the input picture.

[0079] Embodiment three

[0080] In order to perform the method corresponding to the above-mentioned embodiment one, to realize the corresponding function and technical effect, the following provides a fiber distribution port recognition system based on color space fusion.

[0081] As Figure 6 shown, the fiber distribution port recognition system based on color space fusion provided by the embodiment of the application comprises:

[0082] A target fiber distribution box picture acquisition module 1 is configured to acquire a target fiber distribution box picture; the target fiber distribution box is a fiber distribution box with mark points pasted at four corners; the mark points are used to convert the position of the fiber distribution box region in the camera coordinate system into the position of the fiber distribution box region in the world coordinate system; and the target fiber distribution box picture is an RGB format picture.

[0083] A first target area determination module 2 is configured to process the target fiber distribution box picture to obtain a plurality of first target areas corresponding to the target fiber distribution box picture; and the first target area is a region where a red fiber insertion dustproof cap may exist.

[0084] A second target area determination module 3 is configured to input each first target area into a pre-trained color space fusion network to obtain a plurality of second target areas; and the second target area is a region where a red fiber insertion dustproof cap exists.

[0085] An Euclidean distance calculation module 4 is configured to calculate the Euclidean distance of each second target area to other second target areas by Minkowski distance formula.

[0086] A red fiber insertion dust cap region determining module 5 is configured to perform hierarchical clustering on each second target region according to the Euclidean distance to obtain a plurality of clustering result regions; the clustering result region is an uninserted fiber port position region in the target fiber distribution box picture.

[0087] Embodiment four

[0088] The electronic device provided by the embodiment of the present application comprises a memory and a processor, the memory is used for storing a computer program, and the processor runs the computer program to make the electronic device execute the fiber distribution port recognition method based on color space fusion of the embodiment one.

[0089] Optionally, the electronic device can be a server.

[0090] In addition, the embodiment of the present application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the fiber distribution port recognition method based on color space fusion of the embodiment one.

[0091] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same or similar parts of each embodiment can be referred to each other. For the system disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the related parts can be referred to the method part.

[0092] The principles and implementation manners of the present application are described by using specific examples in the present application. The above embodiment is only used to help understand the method of the present application and its core idea; meanwhile, for the general technical personnel in the art, according to the idea of the present application, the specific implementation manner and application range can be changed. In conclusion, the content of the specification should not be understood as the limitation of the present application.

Claims

1. A method for identifying an optical fiber distribution outlet based on color space fusion, characterized by, The method comprises the following steps: Obtain a target fiber distribution box picture; The target fiber distribution box is a fiber distribution box with mark points pasted at four corners; The mark points are used to convert the position of the fiber distribution box region in the camera coordinate system to the position of the fiber distribution box region in the world coordinate system; and the target fiber distribution box picture is an RGB format picture; Process the target fiber distribution box picture to obtain a plurality of first target regions corresponding to the target fiber distribution box picture; the first target region is a region where a red fiber insertion dust cap may exist; Input each first target region into a pre-trained color space fusion network to obtain a plurality of second target regions; the second target region is a region where a red fiber insertion dust cap exists; Calculate the Euclidean distance of each second target region relative to other second target regions by using the Minkowski distance formula; Hierarchical clustering is performed on each second target region according to the Euclidean distance to obtain a plurality of clustering result regions; the clustering result region is an uninserted fiber port position region in the target fiber distribution box picture.

2. The method according to claim 1, wherein The method comprises the following steps: Process the target fiber distribution box picture to obtain a plurality of first target regions corresponding to the target fiber distribution box picture, specifically including:

3. The method according to claim 2, wherein Perform transmission transformation and color space conversion on the target fiber distribution box picture to obtain a plurality of first target regions corresponding to the target fiber distribution box picture. Perform transmission transformation and color space conversion on the target fiber distribution box picture to obtain a plurality of first target regions corresponding to the target fiber distribution box picture, specifically including: Extract the two-dimensional coordinates of the mark points at the four corners of the target fiber distribution box picture; Perform perspective transformation on the two-dimensional coordinates of the mark points to obtain a fiber distribution box picture at a front view angle; Perform LAB color space conversion on the fiber distribution box picture at the front view angle to obtain a fiber distribution box picture converted into the LAB color space; Calculate a LAB color histogram according to the A channel of the fiber distribution box picture converted into the LAB color space, and calculate a best percentage threshold suitable for the current red dust cap region according to the LAB color histogram; 4. The method of claim 1, wherein the color space fusion-based fiber distribution outlet recognition method is characterized by, Perform binaryzation processing on the best percentage threshold to obtain all regions in the fiber distribution box picture at the front view angle where a red fiber insertion dust cap may exist. The construction process of the pre-trained color space fusion network is as follows: Obtain an original fiber distribution box sample picture; the original fiber distribution box sample picture is an RGB format picture; Mark the region of the red fiber insertion dust cap in each original fiber distribution box sample picture as 1, and mark the region other than the red fiber insertion dust cap in each original fiber distribution box sample picture as 0; Convert each original fiber distribution box sample picture into the LAB color space, the YUV color space and the YCrCb color space, and then take out the A channel in the LAB color space, the V channel in the YUV color space and the Cr channel in the YCrCb color space, and then fuse the V channel, the Cr channel and the A channel to obtain a VCrA color space sample picture; Construct a convolutional neural network based on a DenseBlock; The VCrA color space sample picture and the original fiber distribution box sample picture are respectively input into a model composed of a convolutional neural network of DenseBlock for training to obtain a first trained model and a second trained model; The first trained model and the second trained model are combined to obtain a pre-trained color space fusion network; The output result of the pre-trained color space fusion network is a result obtained by weighted fusion of the output result of the first trained model and the output result of the second trained model.

5. The color space fusion-based fiber distribution outlet identification method of claim 4, wherein, Each first target region is input into the pre-trained color space fusion network to obtain a plurality of second target regions, specifically including: Each first target region is respectively converted into a LAB color space, a YUV color space and a YCrCb color space, and the A channel in the LAV color space, the V channel in the YUV color space and the Cr channel in the YCrCb color space are respectively taken out, and then the taken-out V channel, Cr channel and A channel are fused to obtain a VCrA color space picture corresponding to each target region; Each target region and the corresponding VCrA color space picture are input into the pre-trained color space fusion network to obtain a plurality of second target regions.

6. The method of claim 1, wherein the color space fusion-based fiber distribution outlet recognition method is characterized by, Further comprising: Each clustering result region is rectangularly marked and counted by using an opencv library rectangle function.

7. A fiber distribution outlet identification system based on color space fusion, comprising: It comprises: a target fiber distribution box picture acquisition module for acquiring a target fiber distribution box picture; The target fiber distribution box is a fiber distribution box with mark points attached to four corners; the mark points are used to convert the fiber distribution box region position in a camera coordinate system into a fiber distribution box region position in a world coordinate system; The target fiber distribution box picture is an RGB format picture; a first target region determination module for processing the target fiber distribution box picture to obtain a plurality of first target regions corresponding to the target fiber distribution box picture; the first target region is a region where a red fiber insertion dust cap may exist; a second target region determination module for inputting each first target region into a pre-trained color space fusion network to obtain a plurality of second target regions; the second target region is a region where a red fiber insertion dust cap exists; a Euclidean distance calculation module for calculating the Euclidean distance of each second target region relative to other second target regions by using a Minkowski distance formula; a red fiber insertion dust cap region determination module for performing hierarchical clustering on each second target region according to the Euclidean distance to obtain a plurality of clustering result regions; the clustering result region is an uninserted fiber port position region in the target fiber distribution box picture.

8. An electronic device, comprising: An electronic device comprises a memory for storing a computer program and a processor for running the computer program to enable the electronic device to perform the fiber distribution port recognition method based on color space fusion according to any one of claims 1 to 6.

Images