CN110827243B - Method and device for detecting abnormity of coverage area of grid beam - Google Patents
Method and device for detecting abnormity of coverage area of grid beam Download PDFInfo
- Publication number
- CN110827243B CN110827243B CN201911025801.2A CN201911025801A CN110827243B CN 110827243 B CN110827243 B CN 110827243B CN 201911025801 A CN201911025801 A CN 201911025801A CN 110827243 B CN110827243 B CN 110827243B
- Authority
- CN
- China
- Prior art keywords
- grid beam
- image
- mask
- image block
- coverage area
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30204—Marker
Abstract
The embodiment of the application provides a method and a device for detecting the abnormity of a grid beam coverage area, wherein the method comprises the following steps: acquiring an image including a coverage area of a grid beam to be detected; extracting a grid beam area which meets a preset condition in the image to obtain a grid beam image block; respectively calculating the similarity of the abnormal grid beam image block and each grid beam image block; and obtaining an abnormal detection result of the coverage area of the grid beam to be detected according to the similarity. According to the embodiment of the application, the image of the coverage area of the grid beam to be detected is collected, the image block of the grid beam is extracted, and then the abnormal detection result is obtained according to the similarity between the extracted image block of the grid beam and the image block of the abnormal grid beam. Compared with a mode of mainly removing detection through manpower, the method and the device do not need to consume a large amount of labor cost, and detection efficiency is high.
Description
Technical Field
The application belongs to the technical field of machine learning, and particularly relates to a grid beam coverage area anomaly detection method and device.
Background
The grid beams, which can also be called cross beams, are beams with the same height and are not divided into primary beams and secondary beams, and are intersected in the same position to form a cross shape. The grid beams or the cross beams can be arranged in some sloping field areas to play a role of slope protection.
At present, when anomaly detection is performed on a grid beam coverage area, a manual mode is mainly adopted to check and detect, for example, whether the coverage area in the middle of the grid beam is exposed is manually detected. When the area that needs to detect is more, artifical the detection needs to spend a large amount of human costs, and detection efficiency is very low.
Disclosure of Invention
The embodiment of the application provides a method and a device for detecting the abnormal conditions of a grid beam coverage area, and aims to solve the problems that a large amount of labor cost is required to be consumed and the detection efficiency is low when the abnormal conditions of the grid beam coverage area are detected manually.
In a first aspect, an embodiment of the present application provides a method for detecting an anomaly in a coverage area of a grid beam, including:
acquiring an image including a coverage area of a grid beam to be detected;
extracting a grid beam area which meets a preset condition in the image to obtain a grid beam image block;
respectively calculating the similarity of the abnormal grid beam image block and each grid beam image block;
and obtaining an abnormal detection result of the coverage area of the grid beam to be detected according to the similarity.
It can be seen that the abnormal detection result is obtained by collecting the image including the coverage area of the grid beam to be detected, extracting the image block of the grid beam and then according to the similarity between the extracted image block of the grid beam and the image block of the abnormal grid beam. Compared with a mode of mainly removing detection through manpower, the method and the device do not need to consume a large amount of labor cost, and detection efficiency is high.
With reference to the first aspect, in a possible implementation manner, the extracting a grid beam region in the image, where the grid beam region meets a preset condition, to obtain a grid beam image block includes:
converting the image into a target picture;
extracting a plurality of mask images from the target picture; wherein the target picture comprises at least one pixel part, and the number of the mask images is equal to that of the pixel parts of the target picture;
selecting a target mask image from a plurality of mask images;
and extracting the grid beam image block which meets the preset condition from a target mask image.
With reference to the first aspect, in a possible implementation manner, the selecting a target mask map from the plurality of mask maps includes:
cutting each mask image to obtain mask image blocks;
inputting each mask image block into a pre-trained convolutional neural network model, and obtaining a probability value which is output by the convolutional neural network model and represents that each mask image block belongs to the grid beam category;
obtaining the probability value of each mask image block representing the grid beam category according to the probability value of each mask image block;
and taking the mask graph with the maximum probability value as the target mask graph.
With reference to the first aspect, in a possible implementation manner, the extracting, from a target mask map, the grid beam image block that meets the preset condition includes:
labeling the grid beam region in the target mask map by a contour extraction method based on the contour of the concerned region, wherein the contour of the concerned region is the contour of the grid beam region;
and extracting the marked grid beam area to obtain the grid beam image block.
With reference to the first aspect, in a possible implementation manner, the converting the image into the target picture includes:
according to the set quantity of the clustering centers, obtaining an initial point for the image by using a K-means algorithm;
and after initializing the cluster center parameters of the Gaussian mixture model by using the initial point as a clustering center, adjusting the cluster center parameters of the Gaussian mixture model through a maximum expectation algorithm to obtain a clustering result, wherein the clustering result is the target picture.
With reference to the first aspect, in a possible implementation manner, the extracting a grid beam region in the image, where the grid beam region meets a preset condition, to obtain a grid beam image block further includes:
denoising the target mask image by using a morphological method;
the extracting the grid beam image block meeting the preset condition from the target mask image comprises:
and extracting the grid beam image block which meets the preset condition from the denoised target mask image.
With reference to the first aspect, in a possible implementation manner, the separately calculating the similarity between the abnormal grid beam image block and each grid beam image block includes:
and inputting each grid beam image block and the abnormal grid beam image block to a pre-trained deep learning model to obtain the similarity output by the deep learning model.
With reference to the first aspect, in a possible implementation manner, the obtaining an anomaly detection result of the coverage area of the grid beam to be detected according to the similarity includes:
and after the grid beam image blocks with the similarity exceeding the preset threshold are labeled, combining the grid beam image blocks according to the positions during extraction to obtain a picture representing an abnormal detection result.
In a second aspect, an embodiment of the present application provides a device for detecting an anomaly in a coverage area of a grid beam, including a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor implements the method according to any one of the above first aspects when executing the computer program.
In a third aspect, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the computer program implements the method according to any one of the above first aspects.
In a fourth aspect, an embodiment of the present application provides a computer program product, which when running on a terminal device, causes a grid beam coverage area abnormality detection apparatus to execute the method described in any one of the above first aspects.
It is to be understood that, for the beneficial effects of the second aspect to the fourth aspect, reference may be made to the relevant description in the first aspect, and details are not described herein again.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments or the prior art descriptions will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without inventive exercise.
Fig. 1 is a schematic block diagram of a flow of a method for detecting an anomaly in a grid beam coverage area according to an embodiment of the present application;
fig. 2 is a schematic block diagram of a specific process of step S102 according to an embodiment of the present disclosure;
fig. 3 is a schematic block diagram of a specific flow of step S203 provided in an embodiment of the present application;
fig. 4 is a schematic block diagram of another specific flow of step S102 provided in the embodiment of the present application;
FIG. 5 is a diagram of an image to be processed according to an embodiment of the present disclosure;
fig. 6 is a mask diagram of an image to be processed according to an embodiment of the present application;
fig. 7 is a denoised target mask map provided in an embodiment of the present application;
fig. 8 is a picture of an abnormal grid beam with distance marked provided in the embodiment of the present application;
fig. 9 is a block diagram of a structure of an anomaly detection device for a grid beam coverage area according to an embodiment of the present application;
fig. 10 is a schematic structural diagram of an abnormality detection apparatus for a grid beam coverage area according to an embodiment of the present application.
Detailed Description
In the following description, for purposes of explanation and not limitation, specific details are set forth, such as particular system structures, techniques, etc. in order to provide a thorough understanding of the embodiments of the present application.
According to the embodiment of the application, the image comprising the coverage area of the grid beam to be detected is collected, the image blocks of the grid beam in the image are extracted, and the abnormal detection result of the coverage area of the grid beam to be detected is obtained according to the similarity between the extracted image blocks of the grid beam and the abnormal image blocks of the grid beam, so that the cost is reduced, and the detection efficiency is improved.
In order to better describe the technical solutions provided by the embodiments of the present application, the following description will be given by specific embodiments.
Referring to fig. 1, a schematic block diagram of a flow of a method for detecting an anomaly in a coverage area of a grid beam provided in an embodiment of the present application may include the following steps:
and S101, acquiring an image including a coverage area of the grid beam to be detected.
It can be understood that the above images can be obtained by shooting through an image acquisition device by an unmanned aerial vehicle or an unmanned vehicle.
And S102, extracting a grid beam area meeting preset conditions in the image to obtain a grid beam image block.
It should be noted that the image includes a coverage area of the grid beams to be detected, the coverage area of the grid beams to be detected includes grid beam areas, and the grid beam areas are specifically rectangular areas or square areas.
The preset condition may refer to the size of the outline of the attention area, that is, the size range of the outline of the attention area is set to filter out part of irrelevant area positions, and then the corresponding grid beam area is extracted according to the set size of the outline of the attention area. For example, the contour of the region of interest is set to be the contour of the region surrounded by the grid beams, the grid beam region is automatically labeled from the mask map representing the grid beams by a contour extraction method, and then the labeled grid beam region is extracted to obtain the grid beam image block.
In specific application, the image may be converted into a target image, a mask map representing the grid beam is extracted from the pixel image, and then a grid beam region is extracted from the mask map representing the grid beam according to the set outline size of the region of interest, so as to obtain a grid beam image block.
And step S103, respectively calculating the similarity of the abnormal grid beam image block and each grid beam image block.
The abnormal grid beam image block is an image block corresponding to a preset abnormal condition of the grid beam, the preset abnormal condition is artificially defined, for example, if the middle coverage area of the grid beam is artificially defined to be exposed to be abnormal, the abnormal grid beam image block is the grid beam image block with the exposed middle coverage area.
In specific application, the similarity can be calculated through a pre-trained deep learning model, namely the extracted grid beam image blocks and the abnormal grid beam image blocks are input into the deep learning model, and a similarity value output by the deep learning model is obtained. Of course, the calculation method of the similarity is not limited to the above-mentioned method, and for example, the similarity may be calculated by using the methods of SSIM and Hist.
That is, in some embodiments, the specific process of separately calculating the similarities of the abnormal grid beam image block and the respective grid beam image blocks described above may include: and inputting each grid beam image block and each abnormal grid beam image block into a pre-trained deep learning model to obtain the similarity of the deep learning model output.
Specifically, the extracted grid beam image blocks and abnormal grid beam image blocks are input into the depth learning model, and the layers of the depth learning model extract and process the features of the grid beam image blocks and the abnormal grid beam image blocks, and then the distances (i.e., similarities) between the features of the grid image blocks and the abnormal grid image blocks are determined.
The deep learning model is trained in advance, and the training process specifically comprises the following steps: acquiring a public data set, wherein the public data set comprises a plurality of abnormal image block data, extracted image block data and a distance between each abnormal image block and each extracted image block; inputting each abnormal image block and each extracted image block of the public data set into a pre-built deep learning model, extracting and processing the characteristics of the two image blocks by each layer of the deep learning model, calculating the distance (similarity) between the corresponding characteristics, summing and averaging to obtain a distance average value, comparing the distance average value with the distance between each abnormal image block and each extracted image block, judging whether the two distance values are close to each other, if the distance value difference is smaller, representing that the two image blocks are more similar, iterating for multiple times until the Loss value is the smallest, and obtaining the trained deep learning model.
And step S104, obtaining an abnormal detection result of the coverage area of the grid beam to be detected according to the similarity.
It should be noted that the higher the similarity between the extracted grid beam image block and the abnormal grid beam image block is, the higher the probability that the grid beam region corresponding to the grid beam image block is abnormal is. In a specific application, a threshold may be set, and when the similarity is greater than or equal to the threshold, it is considered that the grid beam region corresponding to the grid beam image block is abnormal, whereas when the similarity is less than the threshold, it is considered that the grid beam region corresponding to the grid beam image block is not abnormal. And accordingly, obtaining the abnormity detection result of each grid beam image block, thereby obtaining the abnormity detection result of the whole coverage area of the to-be-detected grid beam.
In a specific application, the abnormality detection result may be embodied as a picture, and the picture is composed of image blocks labeled one by one, that is, after the result that whether each image block is abnormal or not is obtained, the image blocks are labeled according to the similarity, and then the labeled image blocks are combined according to the position of each image block when extracted, so as to obtain the picture representing the abnormality detection result.
That is to say, in some embodiments, the specific process of obtaining the anomaly detection result of the grid beam coverage area to be detected according to the similarity may include: and after the grid beam image blocks with the similarity exceeding the preset threshold are labeled, combining the grid beam image blocks according to the positions during extraction to obtain a picture representing an abnormal detection result.
Specifically, after the similarity is calculated, the abnormal grid beams with the similarity exceeding a threshold value can be marked in a manual mode, and the preset threshold value is an empirical value; after labeling, all the image blocks are combined to obtain a picture, and abnormal conditions of the coverage area of the grid beams can be seen from the picture.
It can be seen that the abnormal detection result is obtained by collecting the image including the coverage area of the grid beam to be detected, extracting the image block of the grid beam and then according to the similarity between the extracted image block of the grid beam and the image block of the abnormal grid beam. Compared with a mode of mainly removing detection through manpower, the method and the device do not need to consume a large amount of labor cost, and detection efficiency is high.
Referring to a schematic block diagram of a specific flow of step S102 shown in fig. 2, in step S102, that is, the step of extracting the grid beam region meeting the preset condition in the image, a specific process of obtaining the grid beam image block may include:
step S201, converting the image into a target picture.
Specifically, according to the set number of clustering centers, an initial point is obtained by using a K-means algorithm for the image; and after initializing the cluster center parameters of the Gaussian mixture model by using the initial point as a cluster center, adjusting the cluster center parameters of the Gaussian mixture model by using a maximum expectation algorithm to obtain a clustering result, wherein the clustering result is a target picture.
The number of the clustering centers is set artificially, and the number of the clustering centers is the same as the number of types included in the image. For example, if the image mainly includes 3 types, such as green plants, exposed grid beam middle coverage areas, and non-exposed grid beam middle coverage areas, the number of the clustering centers is set to 3 manually.
Step S202, extracting a plurality of mask images from a target image; the target picture comprises at least one pixel part, and the number of the mask pictures is equal to that of the pixel parts of the target picture.
It should be noted that the target picture may include one or more color portions or pixel portions, and the number of extracted mask maps is the same as the number of color portions or pixel portions. For example, the target picture includes three color portions of red, green, and blue, white is set to represent a red pixel portion, a green pixel portion, and a blue pixel portion, and black represents other colors, based on which 3 mask images are extracted from the target picture.
Step S203, selecting a target mask image from the plurality of mask images, wherein the target mask image is the mask image with the highest probability of belonging to the grid beam class in the plurality of mask images.
In some embodiments, referring to the specific flowchart schematic block diagram of step S203 shown in fig. 3, the specific process of selecting the target mask map from the plurality of mask maps may include:
and step S301, cutting each mask image to obtain a mask image block.
It should be noted that the number of mask image blocks obtained by cutting each mask image may be the same, for example, each mask image is cut into 4 mask image blocks; of course, it may also be different, for example, one mask is divided into 4 mask image blocks, and the other mask is divided into 5 mask image blocks. In addition, the size of the masked image blocks is generally uniform.
Step S302, inputting each mask image block to a pre-trained convolutional neural network model, and obtaining a probability value which is output by the convolutional neural network model and represents that each mask image block belongs to the grid beam category.
The training process of the convolutional neural network model comprises the following steps: acquiring a large number of pictures including a grid beam coverage area through equipment such as an unmanned aerial vehicle; selecting and marking grid beam pictures in a manual mode to form a training data set; converting the picture of the training data set into a target picture, extracting a plurality of mask pictures from the target picture, cutting the mask pictures into mask image blocks, inputting the mask image blocks into a pre-constructed convolutional neural network model, predicting the probability value of the image blocks belonging to the grid beam category through a softmax function layer of the convolutional neural network model, judging whether the prediction category is consistent with the labeled category, and sequentially carrying out iterative training for a plurality of times until the model tends to be convergent to obtain the trained convolutional neural network model.
Step S303, obtaining the probability value of each mask image sign belonging to the grid beam category according to the probability value of each mask image block.
And step S304, taking the mask map with the maximum probability value as a target mask map.
Specifically, after the probability values of the mask image blocks are calculated, the probability values of the mask image blocks of each mask image are added to obtain the probability value of each mask image, and the mask image with the highest probability value is selected as a target mask image which can represent the grid beam most.
For example, the a mask map is sliced to obtain 4 mask image blocks, which are a1, a2, A3, and a4, and the B mask map is sliced to obtain 4 mask image blocks, which are B1, B2, B3, and B4; respectively inputting A1, A2, A3 and A4, and B1, B2, B3 and B4 into the trained convolutional neural network model to obtain the probability value of each mask image block; finally, the probability values of A1, A2, A3 and A4 are added to obtain the probability value of an A mask image, and the probability values of B1, B2, B3 and B4 are added to obtain the probability value of a B mask image. And comparing the probability values of the mask map A and the mask map B, and taking the mask map with a larger probability value as a target mask map.
In other embodiments, when each mask image is divided into different numbers of mask image blocks, after the probability value of each mask image block is calculated, the probability values of the mask image blocks of each mask image may be added to obtain an addition sum, the addition sum is divided by the number of the mask image blocks to obtain a probability value of the mask image, and the mask image with the highest probability value is selected as the target mask image based on the size of the probability value.
For example, the a mask map is divided into 4 mask image blocks, a1, a2, A3 and a4 respectively, and the B mask map is divided into 5 mask image blocks, B1, B2, B3, B4 and B5 respectively. The probability values of the mask image blocks are calculated respectively, then the probability values of A1, A2, A3 and A4 are added, and then the sum is divided by 4, namely, (A1+ A2+ A3+ A4)/4. Similarly, the probability value (B1+ B2+ B3+ B4+ B5)/5 of the B mask map is calculated. Then, the probability values between the A mask map and the B mask map are compared, and the mask map with the larger probability value is taken as the target mask map.
In other embodiments, the mask map may not be cut, but the whole mask map is input into a pre-trained convolutional neural network model to obtain probability values representing that the mask map belongs to the grid beam categories, and then the mask map with the highest probability value is taken as the target mask map. Compared with a method of performing probability calculation after cutting the mask map into a plurality of image blocks, the method of performing probability calculation based on the whole mask map has a lower accuracy, and the method of performing probability calculation after cutting the mask map into a plurality of image blocks has a higher accuracy.
And step S204, extracting the grid beam image block meeting the preset conditions from the target mask image.
It should be noted that the preset condition may refer to a preset contour size of the attention area, that is, a size range of the contour of the attention area is set to filter out a part of irrelevant area positions, and then the corresponding grid beam area is extracted according to the set contour size of the attention area. For example, the contour of the attention area is set as the contour of the grid beam area, the grid beam area is labeled from a mask map representing the grid beam by a contour extraction method, and then the labeled grid beam area is extracted to obtain the grid beam image block.
In some embodiments, the specific process of extracting the grid beam image block meeting the preset condition from the target mask map may include: automatically labeling the grid beam region in the target mask map by a contour extraction method based on the contour of the concerned region, wherein the contour of the concerned region is the contour of the region surrounded by the grid beams; and extracting the outline of the marked area surrounded by the grid beam to obtain the image block of the grid beam.
Specifically, irrelevant area positions are filtered out by setting the size of the outline of the attention area, and then the outline of the attention area is extracted from the target mask image by using an outline extraction method, such as an OpenCv findContours API, so that a plurality of grid beam image blocks are obtained.
Referring to the specific flow schematic block diagram of step S102 shown in fig. 4, in step S102, that is, the step of extracting the grid beam region meeting the preset condition in the image, the specific process of obtaining the grid beam image block may include:
and step S401, converting the image into a target picture.
Step S402, extracting a plurality of mask images from a target image; the target picture comprises at least one pixel part, and the number of the mask pictures is equal to that of the pixel parts of the target picture.
Step S403, selecting a target mask map from the plurality of mask maps, where the target mask map is a mask map with the highest probability of belonging to the grid beam class in the plurality of mask maps.
It should be noted that steps S401 to S403 are the same as steps S201 to S203, and the related description is please refer to the corresponding contents above, which is not repeated herein.
And S404, denoising the target mask image by using a morphological method.
It will be appreciated that morphology is typically used in the pre-processing stage of images, and has the effect of removing noise points in the image, extracting image boundaries, extracting object skeletons, filling holes in the image, etc., simplifying the image data, preserving their basic shape characteristics, and removing irrelevant structures. Basic algorithms of morphology include dilation-erosion operation and the like.
In the embodiment of the present application, the mask pattern representing the grid beam obtained in step S403, i.e., the target mask pattern, is subjected to operations such as expansion etching to remove noise points and fill the picture, i.e., some small regions are used as noise points and are filled with surrounding pixels, so as to finally obtain a morphologically processed picture.
And S405, extracting the grid beam image block which meets the preset condition from the denoised target mask image.
Specifically, after removing the noise of the target mask image by using a morphological method, extracting the corresponding grid beam image block from the target mask image after removing the noise, where the extraction process is the same as that in step S204, and is not described herein again.
Compared with the grid beam image block extraction process corresponding to fig. 4 and the grid beam image block extraction process corresponding to fig. 2, the former process has better effect, and the extracted grid beam image block is more accurate.
In order to better describe the technical solutions provided by the embodiments of the present application, the following description will be made with reference to specific drawings.
First, a to-be-processed image as shown in fig. 5 is obtained, and the to-be-processed image is a picture including a coverage area of a to-be-detected grid beam, and the picture is taken by an unmanned aerial vehicle. Fig. 5 includes green vegetation and grid-beam covered areas, with some of the grid-beams in the middle area being bare, free of green vegetation, and some of the grid-beams in the middle area being green vegetation. Artificially setting that the middle coverage area of the grid beam is abnormal in exposure, namely detecting the exposed grid beam in the middle coverage area of the grid beam.
After the graph 5 is obtained, since the graph 5 mainly includes 3 types of green plants, the grid beam middle coverage area is exposed, the grid beam middle coverage area is not exposed, and the like, the number of the clustering center points is set to 3 manually. The K-means algorithm, the EM algorithm, and the gaussian mixture model are then used on fig. 5 to convert fig. 5 into the target picture. The target picture includes three color portions of red, green and blue, white is set to represent a red pixel portion, a green pixel portion and a blue pixel portion, and black represents other colors. Next, 3 mask maps as shown in fig. 6 are extracted from the target picture as many as the number of pixel portions.
Then, all 3 mask maps shown in fig. 6 are cut into image blocks, a pre-trained convolutional neural network model is used to predict the probability value of each image block belonging to the grid beam class, and then the total probability value of the 3 mask maps belonging to the grid beam class is calculated. And then, selecting a mask map with the maximum total probability value as a target mask map.
The target mask selected from fig. 6 is subjected to operations such as dilation-erosion to remove noise of the target mask, and a picture as shown in fig. 7 is obtained. Then, the area of the grid beam is set as the attention area in advance, the OpenCv findContours API is used to perform contour extraction on fig. 7 to mark (for example, circle with a rectangular frame) the grid beam in fig. 7, and then the marked grid beam area is extracted to obtain the grid beam image block.
After extracting the grid beam image blocks, inputting the grid beam image blocks and the abnormal grid beam image blocks into a pre-trained depth learning model so as to calculate the similarity between the grid beam image blocks and the abnormal grid beam image blocks; and marking the grid beam image blocks with the similarity exceeding the threshold value by adopting a manual mode, and then combining the image blocks to obtain the picture shown in the figure 8. In fig. 8, the grid beam regions considered as anomalies are marked with boxes and are marked with values at corresponding positions, which are similarity values (or distance values), e.g., 0.53, 0.57, etc. It can be seen from fig. 8 which grid beam region is abnormal and which grid beam region is normal.
Corresponding to the method for detecting an abnormality of a grid beam coverage area described in the foregoing embodiment, fig. 9 shows a block diagram of a structure of a device for detecting an abnormality of a grid beam coverage area provided in an embodiment of the present application.
Referring to fig. 9, the apparatus may include:
the image acquisition module 91 is used for acquiring an image comprising a coverage area of the grid beam to be detected;
the grid beam image block extraction module 92 is used for extracting a grid beam area meeting preset conditions in the image to obtain a grid beam image block;
the similarity calculation module 93 is configured to calculate similarities of the abnormal grid beam image block and each grid beam image block;
and the anomaly detection module 94 is used for obtaining an anomaly detection result of the coverage area of the grid beam to be detected according to the similarity.
In a possible implementation manner, the grid beam image block extraction module is specifically configured to:
converting the image into a target picture;
extracting a plurality of mask images from a target picture; the target picture comprises at least one pixel part, and the number of the mask pictures is equal to that of the pixel parts of the target picture;
selecting a target mask map from the plurality of mask maps, wherein the target mask map is the mask map with the highest probability of belonging to the grid beam category in the plurality of mask maps;
and extracting the grid beam image block meeting the preset condition from the target mask image.
In a possible implementation manner, the grid beam image block extraction module is specifically configured to: cutting each mask image to obtain mask image blocks;
inputting each mask image block into a pre-trained convolutional neural network model to obtain a probability value which is output by the convolutional neural network model and represents that each mask image block belongs to the grid beam category;
obtaining the probability value of each mask image block character belonging to the grid beam category according to the probability value of each mask image block;
and taking the mask graph with the maximum probability value as a target mask graph.
In a possible implementation manner, the grid beam image block extraction module is specifically configured to:
automatically labeling the grid beam region in the target mask image by a contour extraction method based on the contour of the attention region, wherein the contour of the attention region is the contour of the grid beam region;
and extracting the marked grid beam area to obtain a grid beam image block.
In a possible implementation manner, the grid beam image block extraction module is specifically configured to: according to the set quantity of the clustering centers, obtaining an initial point for the image by using a K-means algorithm;
and after initializing the cluster center parameters of the Gaussian mixture model by using the initial point as a cluster center, adjusting the cluster center parameters of the Gaussian mixture model by using a maximum expectation algorithm to obtain a clustering result, wherein the clustering result is a target picture.
In a possible implementation manner, the grid beam image block extraction module is further specifically configured to:
denoising the target mask image by using a morphological method;
the grid beam image block extraction module is specifically configured to: and extracting the grid beam image block which meets the preset condition from the denoised target mask image.
In a possible implementation manner, the similarity calculation module is specifically configured to:
and inputting each grid beam image block and each abnormal grid beam image block into a pre-trained deep learning model to obtain the similarity of the deep learning model output.
In a possible implementation manner, the anomaly detection module is specifically configured to: and after the grid beam image blocks with the similarity exceeding the preset threshold are labeled, combining the grid beam image blocks according to the positions during extraction to obtain a picture representing an abnormal detection result.
It should be noted that, for the information interaction, execution process, and other contents between the above-mentioned devices/units, the specific functions and technical effects thereof are based on the same concept as those of the embodiment of the method of the present application, and specific reference may be made to the part of the embodiment of the method, which is not described herein again.
Fig. 10 is a schematic structural diagram of an abnormality detection apparatus for a grid beam coverage area according to an embodiment of the present application. As shown in fig. 10, the grid beam covered area abnormality detection apparatus 10 of this embodiment includes: at least one processor 100, a memory 101, and a computer program 102 stored in the memory 101 and executable on the at least one processor 100, the processor 100 implementing the steps in any of the various satellite positioning signal strength prediction method embodiments described above when executing the computer program 102.
The device 10 for detecting an abnormality in a coverage area of a grid beam may be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server, and may be integrated with a device such as an unmanned vehicle or an unmanned aerial vehicle. Of course, the abnormal detection device for the coverage area of the grid beam can also be embodied as an unmanned aerial vehicle or other mobile equipment. The abnormal detection device for the grid beam coverage area may include, but is not limited to, a processor 100 and a memory 101. It will be understood by those skilled in the art that fig. 10 is only an example of the grid beam coverage area abnormality detection apparatus 10, and does not constitute a limitation to the grid beam coverage area abnormality detection apparatus 10, and may include more or less components than those shown in the drawings, or combine some components, or different components, for example, may further include an input/output device, a network access device, and the like.
The Processor 100 may be a Central Processing Unit (CPU), and the Processor 100 may be other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable Gate Array (FPGA) or other Programmable logic device, a discrete Gate or transistor logic device, a discrete hardware component, or the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The storage 101 may be an internal storage unit of the grid beam covered area abnormality detection apparatus 10 in some embodiments, for example, a hard disk or a memory of the grid beam covered area abnormality detection apparatus 10. In other embodiments, the memory 101 may also be an external storage device of the grid beam coverage area abnormality detection apparatus 10, for example, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), or the like provided on the grid beam coverage area abnormality detection apparatus 10. Further, the memory 101 may also include both an internal storage unit and an external storage device of the grid beam coverage area abnormality detection apparatus 10. The memory 101 is used for storing an operating system, an application program, a BootLoader (BootLoader), data, and other programs, such as program codes of the computer program. The memory 101 may also be used to temporarily store data that has been output or is to be output.
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-mentioned division of the functional units and modules is illustrated, and in practical applications, the above-mentioned function distribution may be performed by different functional units and modules according to needs, that is, the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-mentioned functions. Each functional unit and module in the embodiments may be integrated in one processing unit, or each unit may exist alone physically, or two or more units are integrated in one unit, and the integrated unit may be implemented in a form of hardware, or in a form of software functional unit. In addition, specific names of the functional units and modules are only used for distinguishing one functional unit from another, and are not used for limiting the protection scope of the present application. The specific working processes of the units and modules in the system may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
The embodiments of the present application further provide a computer-readable storage medium, where a computer program is stored, and when the computer program is executed by a processor, the computer program implements the steps in the above-mentioned method embodiments.
The embodiments of the present application provide a computer program product, which when running on a grid beam coverage area anomaly detection apparatus, enables the grid beam coverage area anomaly detection apparatus to implement the steps in the above method embodiments when executed.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, all or part of the processes in the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium and can implement the steps of the embodiments of the methods described above when the computer program is executed by a processor. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer-readable medium may include at least: any entity or device capable of carrying computer program code to a photographing apparatus/terminal apparatus, a recording medium, computer Memory, Read-Only Memory (ROM), Random Access Memory (RAM), an electrical carrier signal, a telecommunications signal, and a software distribution medium. Such as a usb-disk, a removable hard disk, a magnetic or optical disk, etc.
In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and reference may be made to the related descriptions of other embodiments for parts that are not described or illustrated in a certain embodiment.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
In the embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the modules or units is only one logical division, and there may be other divisions when actually implemented, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, indirect coupling or communication connection of devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present application and are intended to be included within the scope of the present application.
Claims (7)
1. A grid beam coverage area abnormity detection method is characterized by comprising the following steps:
acquiring an image including a coverage area of a grid beam to be detected;
converting the image into a target picture;
extracting a plurality of mask images from the target picture; wherein the target picture comprises at least one pixel part, and the number of the mask images is equal to that of the pixel parts of the target picture;
cutting each mask image to obtain mask image blocks;
inputting each mask image block into a pre-trained convolutional neural network model, and obtaining a probability value which is output by the convolutional neural network model and represents that each mask image block belongs to the grid beam category;
obtaining the probability value of each mask image block representing the grid beam category according to the probability value of each mask image block;
taking the mask graph with the maximum probability value as a target mask graph;
extracting grid beam image blocks meeting preset conditions from the target mask image;
respectively calculating the similarity of the abnormal grid beam image block and each grid beam image block;
obtaining an abnormal detection result of the coverage area of the grid beam to be detected according to the similarity;
wherein the converting the image into a target picture comprises:
according to the set quantity of the clustering centers, obtaining an initial point for the image by using a K-means algorithm;
and after initializing the cluster center parameters of the Gaussian mixture model by using the initial point as a cluster center, adjusting the cluster center parameters of the Gaussian mixture model by a maximum expectation algorithm to obtain a clustering result, wherein the clustering result is the target picture.
2. The method for detecting the abnormal grid beam coverage area according to claim 1, wherein the extracting the grid beam image block meeting the preset condition from the target mask image comprises:
labeling the grid beam region in the target mask map by a contour extraction method based on the contour of the concerned region, wherein the contour of the concerned region is the contour of the grid beam region;
and extracting the marked grid beam area to obtain the grid beam image block.
3. The method of detecting anomalies in a grid beam footprint of claim 1, the method further comprising:
denoising the target mask image by using a morphological method;
the extracting of the grid beam image block meeting the preset condition from the target mask image includes:
and extracting the grid beam image block which meets the preset condition from the denoised target mask image.
4. The method for detecting abnormality of a grid beam covered area according to any one of claims 1 to 3, wherein the separately calculating the similarity of the abnormal grid beam image patch and each of the grid beam image patches includes:
and inputting each grid beam image block and the abnormal grid beam image block to a pre-trained deep learning model to obtain the similarity output by the deep learning model.
5. The method for detecting the abnormality of the grid beam coverage area according to claim 4, wherein the obtaining of the abnormality detection result of the grid beam coverage area to be detected according to the similarity includes:
and after the grid beam image blocks with the similarity exceeding the preset threshold are labeled, combining the grid beam image blocks according to the positions during extraction to obtain a picture representing an abnormal detection result.
6. A grid beam coverage area anomaly detection apparatus comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the grid beam coverage area anomaly detection method according to any one of claims 1 to 5 when executing the computer program.
7. A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the grid beam coverage area anomaly detection method according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911025801.2A CN110827243B (en) | 2019-10-25 | 2019-10-25 | Method and device for detecting abnormity of coverage area of grid beam |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911025801.2A CN110827243B (en) | 2019-10-25 | 2019-10-25 | Method and device for detecting abnormity of coverage area of grid beam |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110827243A CN110827243A (en) | 2020-02-21 |
| CN110827243B true CN110827243B (en) | 2022-07-05 |
Family
ID=69550580
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911025801.2A Active CN110827243B (en) | 2019-10-25 | 2019-10-25 | Method and device for detecting abnormity of coverage area of grid beam |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110827243B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117522872B (en) * | 2024-01-04 | 2024-03-22 | 长峡数字能源科技(湖北)有限公司 | Real-time abnormality detection method and device for photovoltaic power station |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102903110A (en) * | 2012-09-29 | 2013-01-30 | 宁波大学 | Segmentation method for image with deep image information |
| CN103544693A (en) * | 2013-04-11 | 2014-01-29 | Tcl集团股份有限公司 | Method and system for extracting foreground object |
| CN105069779A (en) * | 2015-07-20 | 2015-11-18 | 童垸林 | Building ceramic surface pattern quality detection method |
| CN109242869A (en) * | 2018-09-21 | 2019-01-18 | 科大讯飞股份有限公司 | A kind of image instance dividing method, device, equipment and storage medium |
| CN109685788A (en) * | 2018-12-22 | 2019-04-26 | 北京工业大学 | A kind of flooring defect image automatic testing method based on morphological feature |
| CN110136170A (en) * | 2019-05-13 | 2019-08-16 | 武汉大学 | A kind of remote sensing image building change detecting method based on convolutional neural networks |
-
2019
- 2019-10-25 CN CN201911025801.2A patent/CN110827243B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102903110A (en) * | 2012-09-29 | 2013-01-30 | 宁波大学 | Segmentation method for image with deep image information |
| CN103544693A (en) * | 2013-04-11 | 2014-01-29 | Tcl集团股份有限公司 | Method and system for extracting foreground object |
| CN105069779A (en) * | 2015-07-20 | 2015-11-18 | 童垸林 | Building ceramic surface pattern quality detection method |
| CN109242869A (en) * | 2018-09-21 | 2019-01-18 | 科大讯飞股份有限公司 | A kind of image instance dividing method, device, equipment and storage medium |
| CN109685788A (en) * | 2018-12-22 | 2019-04-26 | 北京工业大学 | A kind of flooring defect image automatic testing method based on morphological feature |
| CN110136170A (en) * | 2019-05-13 | 2019-08-16 | 武汉大学 | A kind of remote sensing image building change detecting method based on convolutional neural networks |
Non-Patent Citations (3)
| Title |
|---|
| Automatic Segmentation and Ventricular Border Detection of 2D Echocardiographic Images Combining K-Means Clustering and Active Contour Model;S.Nandagopalan et al;《2010 Second International Conference on Computer and Network Technology》;20100601;第448-450页 * |
| Facial expression recognition based on binarized statistical image features;Wenjin Chu et al;《2013 Ninth International Conference on Natural Computation (ICNC)》;20140519;第328-330页 * |
| 基于阈值分割的黑龙江省森林类型遥感识别;袁玉娟等;《地理科学进展》;20160531;第35卷(第5期);第655-663页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110827243A (en) | 2020-02-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108416784B (en) | Method and device for rapidly extracting boundary of urban built-up area and terminal equipment | |
| CN109583345B (en) | Road recognition method, device, computer device and computer readable storage medium | |
| CN109409377B (en) | Method and device for detecting characters in image | |
| CN111080526A (en) | Method, device, equipment and medium for measuring and calculating farmland area of aerial image | |
| CN111192239B (en) | Remote sensing image change area detection method and device, storage medium and electronic equipment | |
| CN110992366B (en) | Image semantic segmentation method, device and storage medium | |
| CN110189341B (en) | Image segmentation model training method, image segmentation method and device | |
| CN112989995B (en) | Text detection method and device and electronic equipment | |
| CN111680704B (en) | Automatic and rapid extraction method and device for newly-increased human active plaque of ocean red line | |
| CN111539403A (en) | Agricultural greenhouse identification method and device and electronic equipment | |
| CN114241326B (en) | Progressive intelligent production method and system for ground feature elements of remote sensing images | |
| CN111353956B (en) | Image restoration method and device, computer equipment and storage medium | |
| CN112233077A (en) | Image analysis method, device, equipment and storage medium | |
| CN114722944A (en) | Point cloud precision determination method, electronic device and computer storage medium | |
| CN112651953A (en) | Image similarity calculation method and device, computer equipment and storage medium | |
| CN110827243B (en) | Method and device for detecting abnormity of coverage area of grid beam | |
| CN114880730A (en) | Method and device for determining target equipment and photovoltaic system | |
| CN111738310B (en) | Material classification method, device, electronic equipment and storage medium | |
| CN111860465A (en) | Remote sensing image extraction method, device, equipment and storage medium based on super pixels | |
| CN115345895B (en) | Image segmentation method and device for visual detection, computer equipment and medium | |
| CN111104965A (en) | Vehicle target identification method and device | |
| CN115512238A (en) | Method and device for determining damaged area, storage medium and electronic device | |
| CN115266719A (en) | Water quality monitoring method and device based on artificial intelligence | |
| CN115512258A (en) | Desensitization method and device for video image, terminal equipment and storage medium | |
| CN111062984B (en) | Method, device, equipment and storage medium for measuring area of video image area |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20220214 Address after: 518000 2515, building 2, Huilong business center, North Station community, Minzhi street, Longhua District, Shenzhen, Guangdong Province Applicant after: Shenzhen Zhongke Baotai Aerospace Technology Co.,Ltd. Address before: Room 1101-1102, building 1, Changfu Jinmao building, No.5, Shihua Road, free trade zone, Fubao street, Futian District, Shenzhen, Guangdong 518000 Applicant before: Shenzhen Zhongke Baotai Technology Co.,Ltd. |
|
| TA01 | Transfer of patent application right | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |