CN103714342B - Insulator chain automatic positioning method of taking photo by plane based on bianry image shape facility - Google Patents

Abstract

The invention discloses an insulator string automatic positioning method based on binary image shape features, which belongs to the field of image processing technology and power transmission and transformation equipment operating state maintenance. The specific steps of the method are: 1) performing morphological filtering on the binary image and Remove the small area processing; 2) Digitally describe the three shape features of the insulator string, and process the image based on this, remove the areas that do not meet the three shape features, and retain the areas that meet the three shape features; 3) Use The minimum circumscribed rectangle marks the region in the image that satisfies the shape characteristics of the insulator string to achieve positioning. The present invention makes full use of the shape features of insulator strings, and then adopts a method based on binary image shape feature description to solve the problem of difficult positioning of insulator strings in aerial images with complex backgrounds; the invention is practical and achieves better positioning effect, and the time required is short, without manual participation, which has certain reference significance for the design of related problems.

Description

Automatic positioning method of aerial insulator strings based on binary image shape features

technical field

The invention belongs to the field of image processing technology and operation state maintenance of power transmission and transformation equipment, and in particular relates to an automatic positioning method for insulator strings based on binary image shape features.

Background technique

Locating the target of interest in the image is a key technology in the fields of image processing, computer vision, etc. Its purpose is to extract the position information of some important targets in the image, so as to lay the foundation for the target recognition and subsequent processing. The automatic positioning of insulator strings by aerial photography is an important prerequisite for the realization of insulator condition monitoring and fault diagnosis, and has great practical significance.

There are generally three types of existing positioning methods: (1) methods based on segmentation, which divide the original image into multiple regions and mark the regions of interest; the problem with this method is that it is necessary to manually set the gray level of the target, Moreover, the positioning accuracy is low, and it is difficult to apply it to automatic monitoring. (2) The method based on edge detection finds the outline of the target of interest and realizes positioning; the problem of this method is that it is sensitive to noise, and the accuracy of aerial images with complex background, low resolution and high noise is low, and Due to the complex background and many edge changes, the running speed is slow. (3) The texture-based method analyzes the texture features of the target of interest, and extracts the target position based on this criterion; the biggest problem with this method is that it takes a lot of time to extract texture information, and for some texture features related to insulator strings Similar false targets are difficult to distinguish.

Existing methods do not consider the unique shape features of insulator strings in the segmented binary image. Objects such as insulator strings and towers and lines have obvious shape differences. After binarization, this shape feature is more prominent. If it can be used, it can improve positioning accuracy, reduce time-consuming, and improve automation performance.

Contents of the invention

In view of the problems existing in the above-mentioned prior art, the present invention proposes an automatic positioning method for insulator strings based on binary image shape features, which is characterized in that the specific steps of the method are as follows:

1) Perform morphological filtering and remove small areas on the binary image;

2) Digitally describe the three shape features of the insulator string, and process the image based on this, remove the areas that do not meet the three shape features, and retain the areas that meet the three shape features;

3) Use the minimum circumscribed rectangle to mark the region in the image that meets the shape characteristics of the insulator string to achieve positioning.

The concrete steps of described step 1) are:

11) First, perform morphological corrosion operations on the binarized aerial insulator image, separate objects at thin points, and remove extremely small area noise;

12) Then perform the morphological expansion operation to fill the internal cavity of the target and smooth the boundary of larger objects; among them, in order to better achieve the purpose of fully smoothing the target boundary, the shape parameter of the filter is selected as "circle";

13) Some small regions remain after two morphological operations, and the regions with an area smaller than m×n×1.5×10 ^-5 are removed, where m and n are the number of rows and columns of the binary image, respectively.

The 3 shape features of the insulator string in the step 2) are:

Shape feature 1: The insulator string contains several umbrella disks, and the thickness of each umbrella disk is equal;

Shape feature 2: The thickness of the edge of the umbrella plate is smaller than the thickness of the center, and the thickness gradually increases from the edge to the center;

Shape feature 3: The spacing distance between each umbrella disc is small and equal in value.

The concrete steps of described step 2) are:

21) Digitally describe the shape feature by using the image-to-digital conversion method, set the target value in the binary image to 1, and set the background value to 0 to extract the target information;

22) Extract each column of the image separately to obtain n groups of digital sequences containing only 0 and 1 {l ₁ , l ₂ ,, l ₃ ,..., l _j ,..., l _n }, n is the number of columns in the image ;

23) Describe the shape features of these digital sequences, remove non-insulator objects from the sequence, and then convert the sequence into a matrix to reconstruct the binary image.

The description and processing of the shape feature 1 in the step 23) are:

a1) Search the pixel point from 1 to 0 in the sequence l _j , and record the position of the pixel point with a value of 0, which is the edge point;

b1) Count the number of consecutive 1s before this edge point, record it as O, if there are s edge points, then record s O values {O ₁ , O ₂ , O ₃ ,..., O _i ,..., O _s } ;

In the formula: m is the number of rows of the binary image, t ₁ is the thickness coefficient of the umbrella disk, and N _min is the minimum value of the number of umbrella disks in the insulator string;

When the sequence satisfies the formula (1), it is considered that the edge point of this position belongs to the true edge point, and the point information is retained; if not satisfied, the edge point of this position is considered to be a false edge point, and the "removal" operation is performed;

c1) Based on the criterion in step b1), traverse all sequences {l ₁ , l ₂ , l ₃ , ..., l _j , ..., l _n }, and remove the regions that do not satisfy shape feature 1.

The description and processing of the shape feature 2 in the step 23) are:

a2) After being processed by shape feature 1, the image becomes a plurality of disconnected regions, assuming that there are u disconnected regions, and numbering all regions 1, 2, 3, ..., i, ..., u;

b2) For each area, count the number of 1 in each column and save it as a vector, recorded as G={G ₁ , G ₂ , G ₃ ,...,G _i ,...G _p }, the elements in the vector G _i is the length of each column in the area, and p is the number of non-zero columns in the area; calculate the difference between elements in G to obtain a vector H with a length of p-1, as shown in formula (2):

H _i =G _i+1 -G _i (2)

In formula (2), if the signs of elements H _i in H are all positive _, the thickness of the region gradually increases; The area belongs to the insulator, and the information of this area is reserved;

c2) According to the position where the first value of 1 appears in the sequence _lj , further remove the non-insulator target. For each area, count and record the position of the first value of 1 in each column, J={J ₁ , J ₂ , J ₃ ,...,J _i ,...,J _p }; Calculate the difference of elements in J to get a vector K with length p-1, as shown in formula (3):

K _i =J _i+1 -J _i (3)

In formulas (2) and (3), when the signs of elements H _i in H are both positive, K _i should be both negative; when the signs of elements H _i in H are both negative, K _i should be both positive; When the above conditions are met, it is judged that this area belongs to the insulator umbrella disk part, and this area is reserved;

d2) Based on the criteria in steps b2) and c2), traverse all u regions, and remove the regions that do not satisfy the shape feature 2.

The description and processing of the shape feature 3 in the step 23) are:

a3) Perform image-to-number conversion on the image processed by shape feature 2 to obtain n groups of 0, 1 digital sequences {f ₁ , f ₂ , f ₃ ,..., f _j ,..., f _n }, n is the column of the image number;

b3) In the sequence f _j , search for the pixel point that changes from 0 to 1, mark the position, count the number of consecutive 0s before these marked points, and record it as R; if there are w marked points, record w R value {R ₁ , R ₂ , R ₃ ,...,R _i ,...R _w };

When R satisfies the conditions in formula (4), it is considered that the position belongs to the gap of the insulator shed plate, and the position is set as 1; when w<V _min , it is considered that the position does not belong to the insulator,

In the formula: m is the number of binary image lines, t ₂ is the spacing coefficient of the umbrella disk, and V _min represents the minimum value of the number of gaps in the insulator string;

c3) Based on the criterion in step b3), traverse all sequences {f ₁ , f ₂ , f ₃ , ..., f _j , ..., f _n }, remove the regions that do not satisfy shape feature 3, and describe by shape feature 3 After processing, the insulator area is connected as a whole and grows into a large area in the whole graph.

The concrete steps of described step 3) are:

31) After the shape feature processing step in step 2), large-area connected areas appear in the position of the insulator string, while there are many non-insulator targets with small areas in other positions; design a suitable threshold value calculation method to obtain the threshold value, and delete those whose area is smaller than the threshold value region, a binary image containing only insulator string information is obtained;

32) Use the minimum circumscribed rectangle to mark the target frame, and the obtained rectangular frame is the position of the insulator;

33) Mark the rectangular frame in the original image to obtain the final automatic positioning result of the insulator string.

The calculation of the threshold in the step 31) adopts the maximum area correlation method, and the steps are as follows:

311) Calculate the area of each area, and record the values p ₁ , p ₂ , p ₃ , p ₄ of the four areas with the largest areas;

312) Calculate the mean of the above 4 areas

313) Calculate the threshold s according to formula (5);

In the formula: m is the number of rows of the binary image, n is the number of columns, and a is the area proportional coefficient.

Beneficial effects of the invention: Compared with the existing methods, the present invention makes full use of the shape features of the insulator strings, and then adopts a method based on binary image shape feature description to solve the problem of difficult positioning of the insulator strings in aerial images with complex backgrounds; The invention is practicable and achieves a good positioning effect, and the required time is short without manual participation, which has certain reference significance for the scheme design of related problems.

Description of drawings

Fig. 1 is the flow chart of the present invention's aerial photography insulator string automatic location method based on binary image shape feature;

Figure 2(a) is the original aerial image;

Figure 2(b) is the image after binarizing the original aerial image;

Fig. 2 (c) is the graph of morphological filtering processing result;

Figure 2(d) is the result of processing after removing the small area;

Fig. 3 (a) is step 2) in feature 1 description and processing result figure;

Fig. 3 (b) is step 2) in feature 2 description and processing result figure;

Fig. 3 (c) is a description and processing result figure of feature 3 in step 2);

Figure 4(a) is the image marked with the smallest circumscribed rectangle after removing the small area in step 3);

Figure 4(b) shows the positioning result image obtained by displaying the minimum circumscribed rectangle on the original grayscale image.

detailed description

In order to better understand the technical solution of the present invention, the present invention will be further described below in conjunction with the accompanying drawings.

The whole process of the automatic positioning method of aerial insulator strings based on binary image shape features can be shown in Figure 1. The specific steps of the method are as follows:

1) Perform morphological filtering and remove small areas on the binary image;

The specific steps are:

11) First, perform morphological corrosion operations on the binarized aerial insulator image, separate objects at thin points, and remove extremely small area noise;

12) Then perform the morphological expansion operation to fill the internal cavity of the target and smooth the boundary of larger objects; among them, in order to fully smooth the target boundary, combined with the shape of the insulator umbrella disk, the shape parameter of the filter is selected as "circle".

13) Some small regions remain after two morphological operations, and the regions with an area smaller than m×n×1.5×10 ^-5 are removed, where m and n are the number of rows and columns of the binary image, respectively.

2) Digitally describe the three shape features of the insulator string, and process the image based on this, remove the areas that do not meet the three shape features, and retain the areas that meet the three shape features;

Among them, the three shape features of the insulator string are:

Shape feature 1: The insulator string contains several umbrella disks, and the thickness of each umbrella disk is similar;

Shape feature 2: The thickness of the edge of the umbrella plate is smaller than the thickness of the center, and the thickness gradually increases from the edge to the center;

Shape feature 3: The distance between each umbrella disk is small and the value is similar.

The concrete steps of described step 2) are:

21) Digitally describe the shape feature by using the image-to-digital conversion method, set the target value in the binary image to 1, and set the background value to 0 to extract the target information;

22) Extract each column of the image separately to obtain n groups of digital sequences {l ₁ , l ₂ ,, l ₃ , ..., l _j , ..., l _n } that only contain 0 and 1;

23) Describe the shape features of these digital sequences, remove non-insulator objects from the sequence, and then convert the sequence into a matrix to reconstruct the binary image.

Among them, the description and processing of shape feature 1 are:

a1) Search the pixel point from 1 to 0 in the sequence l _j , and record the position of the pixel point with a value of 0, which is the edge point;

b1) Count the number of consecutive 1s before this edge point, record it as O, if there are s edge points, then record s O values {O ₁ , O ₂ , O ₃ ,..., O _i ,..., O _s } ;

In the formula: m is the number of rows of the binary image, t ₁ is the thickness coefficient of the umbrella disk, and N _min is the minimum value of the number of umbrella disks in the insulator string;

When the sequence satisfies the condition in formula (1), it is considered that the edge point of this position belongs to the true edge point, and the information of this point is retained; if not satisfied, the point s is considered to be a false edge point, and the "removal" operation is performed;

c1) Based on the criterion in step b1), traverse all sequences {l ₁ , l ₂ , l ₃ , ..., l _j , ..., l _n }, and remove the regions that do not satisfy shape feature 1.

The description and processing of shape feature 2 are:

a2) After being processed by shape feature 1, the image becomes a plurality of disconnected regions, assuming that there are u disconnected regions, and numbering all regions 1, 2, 3, ..., i, ..., u;

b2) For each area, count the number of 1 in each column and save it as a vector, recorded as G={G ₁ , G ₂ , G ₃ ,...,G _i ,...G _p }, the elements in the vector G _i is the length of each column in the area, and p is the number of non-zero columns in the area; calculate the difference between elements in G to obtain a vector H with a length of p-1, as shown in formula (2):

H _i =G _i+1 -G _i (2)

In formula (2), if the signs of H _i elements in H are both positive _, the thickness of the region gradually increases; The area belongs to the insulator, and the information of this area is reserved;

c2) According to the position where the first value of 1 appears in the sequence _lj , further remove the non-insulator target. For each area, count and record the position of the first value of 1 in each column, J={J ₁ , J ₂ , J ₃ ,...,J _i ,...,J _p }; Calculate the difference of elements in J to get a vector K with length p-1,

K _i =J _i+1 -J _i (3)

In formulas (2) and (3), when the signs of elements H _i in H are both positive, K should be both negative; when the signs of elements H _i in H are all negative, K should be both positive; area i satisfies the above When the conditions are met, it is judged that this area belongs to the part of the insulator umbrella plate, and this area is reserved;

d2) Based on the criteria in steps b2) and c2), traverse all u regions, and remove the regions that do not satisfy the shape feature 2.

The description and processing of shape feature 3 are as follows:

a3) Perform image-to-number conversion on the image processed by shape feature 2 to obtain n groups of 0, 1 digital sequences {f ₁ , f ₂ , f ₃ ,..., f _j ,..., f _n }, n is the column of the image number;

b3) In the sequence f _j , search for the pixel point that changes from 0 to 1, mark the position, count the number of consecutive 0s before these marked points, and record it as R; if there are w marked points, record w R value {R ₁ , R ₂ , R ₃ ,...,R _i ,...R _w };

When R satisfies the conditions in formula (4), it is considered that the position belongs to the gap of the insulator shed plate, and the position is set as 1; when w<V _min , it is considered that the position does not belong to the insulator,

In the formula: m is the number of binary image lines, t ₂ is the spacing coefficient of the umbrella disk, and V _min represents the minimum value of the number of gaps in the insulator string;

c3) Based on the criterion in step b3), traverse all sequences {f ₁ , f ₂ , f ₃ , ..., f _j , ..., f _n }, remove the regions that do not satisfy shape feature 3, and describe by shape feature 3 After processing, the insulator area is connected as a whole and grows into a large area in the whole graph.

3) Use the minimum circumscribed rectangle to mark the region in the image that meets the shape characteristics of the insulator string to achieve positioning.

The specific steps are:

31) After the shape feature processing step in step 2), large-area connected areas appear in the position of the insulator string, while there are many non-insulator targets with small areas in other positions; design a suitable threshold value calculation method to obtain the threshold value, and delete those whose area is smaller than the threshold value region, a binary image containing only insulator string information is obtained;

The threshold is calculated using the maximum area correlation method, and the steps are as follows:

311) Calculate the area of each area, and record the values p ₁ , p ₂ , p ₃ , p ₄ of the four areas with the largest area;

312) Calculate the mean of the above 4 areas

313) Calculate the threshold s according to formula (5);

In the formula: m is the number of rows of the binary image, n is the number of columns, and a is the area proportional coefficient.

32) Use the minimum circumscribed rectangle to mark the target frame, and the obtained rectangular frame is the position of the insulator;

33) Mark the rectangular frame in the original image to obtain the final automatic positioning result of the insulator string.

Example

The real aerial image is processed by the automatic positioning method of aerial insulator string based on the binary image shape feature. The original image is shown in Figure 2(a), with a size of 640×353, and the binarized image is shown in Figure 2(b); first, the morphological filtering method is used for filtering, and the result is shown in Figure 2(c), and the removed area is smaller than The area of the set value, the result is shown in Figure 2(d); then the insulator string shape features are digitally described, and the processing results of Feature 1, Feature 2 and Feature 3 are shown in Figure 3(a), Figure 3(b), and Figure 3 respectively (c); then use the maximum area correlation method to calculate the threshold, remove the area with an area smaller than the threshold, retain a large area, and mark it with the smallest circumscribed rectangle, the processing result is shown in Figure 4 (a); finally, display the rectangular positioning frame on In the original image, the automatic positioning of the insulator string is realized, and the result is shown in Figure 4(b).

The above is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed in the present invention can easily think of changes or Replacement should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be determined by the protection scope of the claims.

Claims (6)

1. The insulator string automatic location method based on binary image shape feature, it is characterized in that, the concrete steps of this method are as follows: 1) Perform morphological filtering and remove small areas on the binary image; 2) Digitally describe the three shape features of the insulator string, and process the image based on this, remove the areas that do not meet the three shape features, and retain the areas that meet the three shape features; 3) Use the minimum circumscribed rectangle to mark the region in the image that meets the shape characteristics of the insulator string to achieve positioning; The 3 shape features of the insulator string in the step 2) are: Shape feature 1: The insulator string contains several umbrella disks, and the thickness of each umbrella disk is equal; Shape feature 2: The thickness of the edge of the umbrella plate is smaller than the thickness of the center, and the thickness gradually increases from the edge to the center; Shape feature 3: The distance between each umbrella disc is small and the value is equal; The concrete steps of described step 2) are: 21) Digitally describe the shape feature by using the image-to-digital conversion method, set the target value in the binary image to 1, and set the background value to 0 to extract the target information; 22) Extract each column of the image separately, and obtain n groups of digital sequences {l ₁ , l ₂ , l ₃ ,..., l _j ,..., l _n } containing only 0 and 1, where n is the number of columns in the image; 23) Describe the shape features of these digital sequences, remove non-insulator objects from the sequences, convert the sequences into matrices, and reconstruct binary images; The description and processing of the shape feature 1 in the step 23) are: a1) Search the pixel point from 1 to 0 in the sequence l _j , and record the position of the pixel point with a value of 0, which is the edge point; b1) Count the number of consecutive 1s before this edge point, record it as O, if there are s edge points, then record s O values {O ₁ , O ₂ , O ₃ ,..., O _i ,..., O _s } ; In the formula: m is the number of rows of the binary image, t ₁ is the thickness coefficient of the umbrella disk, and N _min is the minimum value of the number of umbrella disks in the insulator string; When the sequence satisfies the formula (1), it is considered that the edge point of this position belongs to the true edge point, and the point information is retained; if not satisfied, the edge point of this position is considered to be a false edge point, and the "removal" operation is performed; c1) Based on the criterion in step b1), traverse all sequences {l ₁ , l ₂ , l ₃ , ..., l _j , ..., l _n }, and remove the regions that do not satisfy the shape feature 1. 2. the insulator string automatic positioning method based on the binary image shape feature according to claim 1, is characterized in that, the concrete steps of described step 1) are: 11) First, perform morphological corrosion operations on the binarized aerial insulator image, separate objects at thin points, and remove extremely small area noise; 12) Then perform the morphological expansion operation to fill the internal cavity of the target and smooth the boundary of larger objects; among them, in order to better achieve the purpose of fully smoothing the target boundary, the shape parameter of the filter is selected as "circle"; 13) Some small regions remain after two morphological operations, and the regions with an area smaller than m×n×1.5×10 ^-5 are removed, where m and n are the number of rows and columns of the binary image, respectively. 3. the insulator string automatic positioning method based on the binary image shape feature according to claim 1, is characterized in that, the description and processing to shape feature 2 in the described step 23) are: a2) After being processed by shape feature 1, the image becomes a plurality of disconnected regions, assuming that there are u disconnected regions, and numbering all regions 1, 2, 3, ..., i, ..., u; b2) For each area, count the number of 1 in each column and save it as a vector, recorded as G={G ₁ , G ₂ , G ₃ ,...,G _i ,...G _p }, the elements in the vector G _i is the length of each column in the area, and p is the number of non-zero columns in the area; calculate the difference between elements in G to obtain a vector H with a length of p-1, as shown in formula (2): H _i =G _i+1 -G _i (2) In formula (2), if the signs of H _i elements in H are both positive _, the thickness of the region gradually increases; The area belongs to the insulator, and the information of this area is reserved; c2) According to the position where the first value of 1 appears in the sequence _lj , further remove the non-insulator target. For each area, count and record the position of the first value of 1 in each column, J={J ₁ , J ₂ , J ₃ ,...,J _i ,...,J _p }; Calculate the difference of elements in J to get a vector K _i with length p-1, as shown in formula (3): K _i =J _i+1 -J _i (3) In formulas (2) and (3), when the signs of elements H _i in H are both positive, K _i should be both negative; when the signs of elements H _i in H are both negative, K _i should be both positive; When the above conditions are met, it is judged that this area belongs to the insulator umbrella disk part, and this area is reserved; d2) Based on the criteria in steps b2) and c2), traverse all u regions, and remove the regions that do not satisfy the shape feature 2. 4. the insulator string automatic positioning method based on the binary image shape feature according to claim 1, is characterized in that, in described step 23), description and processing to shape feature 3 are: a3) Perform image-to-number conversion on the image processed by shape feature 2 to obtain n groups of 0, 1 digital sequences {f ₁ , f ₂ , f ₃ ,..., f _j ,..., f _n }, n is the column of the image number; b3) In the sequence f _j , search for the pixel point that changes from 0 to 1, mark the position, count the number of consecutive 0s before these marked points, and record it as R; if there are w marked points, record w R value {R ₁ , R ₂ , R ₃ , ..., R _i , ..., R _w }; When R satisfies the conditions in formula (4), it is considered that the position belongs to the gap of the insulator shed plate, and the position is set as 1; when w<V _min , it is considered that the position does not belong to the insulator, In the formula: m is the number of binary image lines, t ₂ is the spacing coefficient of the umbrella disk, and V _min represents the minimum value of the number of gaps in the insulator string; c3) Based on the criterion in step b3), traverse all sequences {f ₁ , f ₂ , f ₃ , ..., f _j , ..., f _n }, remove the regions that do not satisfy shape feature 3, and describe by shape feature 3 After processing, the insulator area is connected as a whole and grows into a large area in the whole graph. 5. the insulator string automatic positioning method based on binary image shape feature according to claim 1, is characterized in that, described step 3) concrete steps are: 31) After the shape feature processing step in step 2), large-area connected areas appear in the position of the insulator string, while there are many non-insulator targets with smaller areas in other positions; use the threshold value calculation method to obtain the threshold value, and delete the area whose area is smaller than the threshold value. Obtain a binary image containing only insulator string information; 32) Use the minimum circumscribed rectangle to mark the target frame, and the obtained rectangular frame is the position of the insulator; 33) Mark the rectangular frame in the original image to obtain the final automatic positioning result of the insulator string. 6. the insulator string automatic positioning method based on the binary image shape feature according to claim 5, is characterized in that, the calculation of threshold value in described step 31) adopts the maximum area correlation method, and the steps are as follows: 311) Calculate the area of each area, and record the values p ₁ , p ₂ , p ₃ , p ₄ of the four areas with the largest area; 312) Calculate the mean of the above 4 areas 313) Calculate the threshold s according to formula (5); In the formula: m is the number of rows of the binary image, n is the number of columns, and a is the area proportional coefficient.