A Hough transform parameter space peak extraction method for power line detection in complex environments

Abstract

The present invention discloses a Hough transform parameter space peak extraction method for power line detection in complex environments. The method comprises generating a candidate accumulation unit position matrix and recording accumulation units that meet threshold conditions; analyzing the candidate accumulation unit position matrix and grouping the candidate accumulation units, grouping candidate accumulation units belonging to the same line; extracting peak coordinates of the candidate accumulation units belonging to the same line based on an image centroid algorithm, finding parameter information of the line, updating the candidate accumulation unit position matrix, and clearing information of the detected line; repeating the above operations on the updated candidate accumulation unit matrix for analysis, and finally completing line detection. The method of the present invention improves the detection capability of multiple power lines.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a Hough transform parameter space peak extraction method for power line detection in complex environments. Background Art

[0002] Power lines are difficult to detect with the naked eye due to their small size in conditions of poor visibility and obstructions ahead, making them the most dangerous obstacle for aircraft flying at low altitudes.

[0003] Millimeter-wave radar, with its compact size, high resolution, and all-weather capability, has become a preferred choice for low-altitude collision avoidance systems developed worldwide in recent years. Powerline detection using images generated from millimeter-wave radar data holds significant practical significance and broad application prospects for detecting threats in low-altitude environments. However, powerline targets are small and reflect weakly, and multiple power lines often appear simultaneously and in close proximity, making detection difficult. This can sometimes lead to the detection of only one power line, or misjudgment of the power line's direction.

[0004] The Hough transform is a commonly used method for line detection. After the Hough transform, the typical processing method is to extract peaks in the parameter space, which serve as feature points in the data space to identify the target line. A threshold is often used to determine the estimated parameters of candidate lines. A fast and accurate estimation method is needed to extract the feature points of multiple target lines based on these estimated parameters.

[0005] In its patent application, "A Method for Airborne Millimeter-Wave Radar Power Line Detection" (Patent Application No.: 201710747304.8, Publication No.: 107561509A), the University of Electronic Science and Technology of China uses power line pylons to locate signal gates and uses this information to determine the approximate direction of the power lines. However, for short-range airborne millimeter-wave radar, if the power line pylons at either end of the power line are out of range, the direction of the power line cannot be determined without this information, resulting in certain application limitations.

[0006] In her dissertation, "Research on Radar Weak Target Detection Technology" (Shaanxi: Xidian University, 2012), Wang Juan employed the image centroid algorithm to extract peaks in Hough parameter space. This method can only be used for single-target straight lines, but cannot handle multiple power lines in complex environments.

[0007] In their journal article "Peak Extraction in Line Detection Based on Hough Transform" (Journal of Taiyuan University of Science and Technology, 2006, 27(4)), Chen Gaohua et al. published a peak extraction method that extracts actual estimated parameters from candidate accumulation matrices by comparing the variance of accumulation unit samples. However, this method is not effective in classifying candidate accumulation units parallel to the target line, resulting in the parameters obtained in subsequent operations being unsuitable for use as line parameters. Summary of the Invention

[0008] In view of the shortcomings of the existing technology, the present invention aims to provide a Hough transform parameter space peak extraction method for power line detection in complex environments, and improve the detection capability of multiple power lines through the method of the present invention.

[0009] In order to achieve the above object, the present invention adopts the following technical solutions:

[0010] A method for extracting peak values ​​from Hough transform parameter space for power line detection in a complex environment, the method comprising the following steps:

[0011] S1 generates a candidate accumulation unit position matrix and records the accumulation units that meet the threshold conditions;

[0012] S2 analyzes the candidate accumulation unit position matrix, groups the candidate accumulation units, and groups the candidate accumulation units belonging to the same straight line into one group;

[0013] S3 extracts the peak coordinates of the candidate accumulation units belonging to the same straight line based on the image centroid algorithm and finds the parameter information of the straight line;

[0014] S4 updates the candidate accumulation unit position matrix and clears the information of the detected straight line;

[0015] S5 repeats the above operation to analyze the updated candidate accumulation unit matrix, and finally completes the line detection.

[0016] It should be noted that, after setting a threshold, the accumulation units that meet the threshold condition are extracted and their position coordinates in the entire accumulation matrix H are recorded and stored in a two-dimensional array, which is called the candidate accumulation unit position matrix C. The first column element of this matrix represents the row number of the accumulation unit that meets the threshold condition, and the second column element represents the column number of the accumulation unit that meets the threshold condition. The accumulation units that meet the threshold condition are called candidate accumulation units.

[0017]

[0018] It should be noted that step S2 includes:

[0019] S2.1 defines a one-dimensional array rowfinal to store the row numbers of the candidate accumulation unit position matrix in which the position coordinates of the accumulation units located on the same straight line are recorded.

[0020] S2.2 extracts the latest row number of the rowfinal record (the initial record is 1), recorded as k.

[0021] S2.3 Subtract the element in the kth row from the element in the k+1th row from the element in the first column of the candidate accumulation unit position matrix according to the following formula.

[0022] Δ1=|c k+1,1 -c k,1 |

[0023] If the absolute value of the difference Δ1 is less than T1, the element in the kth row is subtracted from the element in the k+1th row of the second column of the candidate accumulation unit position matrix according to the following formula.

[0024] Δ2=|c k+1,2 -c k,2 |

[0025] If the absolute value of the difference Δ2 is less than T2, the row number of the element is recorded in rowfinal. The parameters determined by the two accumulation units are considered to correspond to the same straight line.

[0026] If the elements in the k+1th row of the alternative accumulation unit position matrix do not meet the above two conditions, the above operations are performed on the elements in the k+2th, k+3th,... rows of the alternative accumulation unit position matrix according to the following formula until the above conditions are met or the operations on the elements in the pth row of the alternative accumulation unit position matrix are completed.

[0027] Δ1=|c i,1 -c k,1 |

[0028] Δ2=|c i,2 -c k,2 |

[0029] i=k+2,k+3,…,p.

[0030] It should be noted that step S3 includes:

[0031] S3.1 One-dimensional array rowfinal=[r1,r2,…,t q ] T r i The accumulation units corresponding to the elements on the rows of the candidate accumulation unit position matrix recorded by (i=1, 2, ..., q) are the accumulation units on the same straight line.

[0032] The S3.2 centroid algorithm is to extract the centroid coordinates of the image; the (m+n) order of the pixel f(x,y) at the coordinate (x,y) is defined as:

[0033]

[0034] l 00 is the total grayscale value of f(x,y). 01 and l 10 l 00 Standardization gives the center of gravity coordinates G(x G ,y G );

[0035] The peak extraction method based on the image center of gravity uses the position coordinates of the candidate accumulation units of the same straight line in the accumulation matrix H to weight the corresponding accumulation values, and sums the weighted accumulation values ​​of the same straight line. Then, the position coordinates of the accumulated mean of the candidate accumulation units of the same straight line are calculated, which is the so-called image center of gravity.

[0036] According to the following formula, the element h of the accumulation unit of the same straight line extracted from the accumulation matrix H is i,j (i, j are the row and column numbers of the element in the matrix H respectively) to calculate:

[0037]

[0038]

[0039]

[0040]

[0041]

[0042] I G ,J G After rounding, this is the peak coordinate to be extracted from this straight line.

[0043] It should be noted that step S4 includes:

[0044] S4.1 clears the corresponding elements on the row number of the candidate accumulation unit position matrix C recorded in the one-dimensional array rowfinal to zero.

[0045] S4.2 clears rowfinal, and after clearing, records the first non-zero row number in the first column of the candidate accumulation unit position matrix C in rowfinal.

[0046] The beneficial effects of the present invention are:

[0047] 1. In order to suppress the interference of other power lines in complex environments, the present invention groups the alternative accumulation units by utilizing the position characteristics of the alternative accumulation units of the same target line in the entire accumulation matrix, and groups the alternative accumulation units belonging to the same line into one group, effectively eliminating the influence between closely parallel power lines.

[0048] 2. To address the problem of high computational complexity when using Hough transform for line detection, which makes it difficult to meet real-time requirements, the present invention uses the image centroid algorithm to weight the corresponding accumulation values ​​using the position information of the candidate accumulation units, thereby reducing the computational complexity and improving the real-time performance of the processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic flow diagram of the present invention;

[0050] Figure 2 It is the mapping diagram of the target straight line in the simulation of the present invention after Hough transformation in the parameter space;

[0051] Figure 3 It is a data space diagram in which the peak points obtained after peak extraction in the simulation of the present invention are inversely mapped to the data space. DETAILED DESCRIPTION

[0052] The present invention will be further described below. It should be noted that the following examples are based on the present technical solution and provide detailed implementation methods and specific operating processes, but the protection scope of the present invention is not limited to these examples.

[0053] Example

[0054] Refer to the attached Figure 1 , the specific implementation steps of the present invention are further described in detail.

[0055] Step 1: Generate a candidate accumulation unit position matrix.

[0056] According to the following formula, a threshold is set and the accumulation units that meet the threshold condition are extracted. Their position coordinates in the entire accumulation matrix H are recorded and stored in a two-dimensional array called the candidate accumulation unit position matrix C. The first column of this matrix represents the row number of the accumulation unit that meets the threshold condition, and the second column represents the column number of the accumulation unit that meets the threshold condition. The accumulation units that meet the threshold condition are called candidate accumulation units.

[0057]

[0058] Step 2: Analyze the candidate accumulation unit position matrix and group the candidate accumulation units.

[0059] (1) Define a one-dimensional array rowfinal to store the position coordinates of the accumulation units located on the same straight line and record the row number of the candidate accumulation unit position matrix.

[0060] (2) Extract the latest row number recorded in rowfinal (the initial record is 1), recorded as k.

[0061] (3) According to the following formula, the element in the kth row is subtracted from the element in the k+1th row of the element in the first column of the candidate accumulation unit position matrix.

[0062] Δ1=|c k+1,1 -c k,1 |

[0063] If the absolute value of the difference Δ1 is less than T1, the element in the kth row is subtracted from the element in the k+1th row of the second column of the candidate accumulation unit position matrix according to the following formula.

[0064] Δ2=|c k+1,2 -c k,2 |

[0065] If the absolute value of the difference Δ2 is less than T2, the row number of the element is recorded in rowfinal. The parameters determined by the two accumulation units are considered to correspond to the same straight line.

[0066] If the elements in the k+1th row of the alternative accumulation unit position matrix do not meet the above two conditions, the above operations are performed on the elements in the k+2th, k+3th,... rows of the alternative accumulation unit position matrix according to the following formula until the above conditions are met or the operations on the elements in the pth row of the alternative accumulation unit position matrix are completed.

[0067] Δ1=|c i,1 -c k,1 |

[0068] Δ2=|c i,2 -c k,2 |

[0069] i=k+2,k+3,…,p

[0070] Step 3: Image centroid algorithm.

[0071] (1) One-dimensional array rowfinal=[r1,r2,…,r q ] T r i The accumulation units corresponding to the elements on the rows of the candidate accumulation unit position matrix recorded by (i=1, 2, ..., q) are the accumulation units on the same straight line.

[0072] (2) The centroid algorithm is to extract the centroid coordinates of the image. The (m+n) order of the pixel f(x,y) at the coordinate (x,y) is defined as:

[0073]

[0074] l 00 is the total grayscale value of f(x,y). 01 and l 10 l 00 Standardization gives the center of gravity coordinates G(x G ,y G ).

[0075] The peak extraction method based on the image center of gravity uses the position coordinates of the alternative accumulation units of the same straight line in the accumulation matrix H to weight the corresponding accumulation values, and sums the weighted accumulation values ​​of the same straight line. Then, the position coordinates of the accumulated mean of the accumulation values ​​of the alternative accumulation units of the same straight line are calculated, which is the so-called image center of gravity.

[0076] According to the following formula, the element h of the accumulation unit of the same straight line extracted from the accumulation matrix H is i,j (i, j are the row and column numbers of the element in the matrix H respectively) to calculate:

[0077]

[0078]

[0079]

[0080]

[0081]

[0082] I G ,J G After rounding, this is the peak coordinate to be extracted from this straight line.

[0083] Step 4: Update the candidate accumulation unit position matrix C.

[0084] (1) Clear the corresponding elements on the row number of the candidate accumulation unit position matrix C recorded in the one-dimensional array rowfinal to zero.

[0085] (2) Clear rowfinal, and after clearing, record the first non-zero row number in the first column of the candidate accumulation unit position matrix C in rowfinal.

[0086] Step 5: Perform the operations of steps 2, 3, and 4 until all elements in the candidate accumulation unit position matrix C are zero. The values ​​of T1 and T2 can be determined according to actual needs.

[0087] Simulation test

[0088] 1. Simulation parameters:

[0089] Table 1 Simulation parameters list

[0090] Target line number Tilt angle (°) Distance (m) 1 175 70 2 175 80 3 8 80 4 5 90 5 5 80 6 13 60 7 165 100

[0091] 2. Simulation content and result analysis:

[0092] As shown in Table 1 above, 7 straight lines are assumed and Hough transform is performed to map from data space to parameter space, such as Figure 2 As shown. Binary accumulation is performed in the parameter space to obtain the parameter space transformation matrix, and a threshold is set. Then, the peak value is extracted in the parameter space using the technology of the present invention. After the peak value is extracted, it is inversely mapped according to the corresponding relationship between the data space and the parameter space to obtain a straight line in the data space. The simulation results are shown as follows: Figure 3 shown.

[0093] Figure 3 The white dashed line in the middle represents the target line in data space before the Hough transform, and the red line represents the line obtained after the Hough transform, peak extraction, and inverse mapping of these points to the data space. Target line 2 is parallel to target line 1 and intersects with target line 3. It can be seen that the inverse-mapped line and the white dashed line have a good overlap effect. This shows that the present invention can perform peak extraction in Hough parameter space, achieve detection of multiple target lines, and can be applied to the detection of multiple power lines in complex environments.

[0094] Those skilled in the art can make various corresponding changes and modifications based on the above technical solutions and concepts, and all of these changes and modifications should be included in the scope of protection of the claims of the present invention.

Claims

1. A Hough transform parameter space peak extraction method for power line detection in complex environments, characterized by: The method comprises the following steps: S1 generates a candidate accumulation unit position matrix and records the accumulation units that meet the threshold conditions; S2 analyzes the candidate accumulation unit position matrix, groups the candidate accumulation units, and groups the candidate accumulation units belonging to the same straight line into one group; S3 extracts the peak coordinates of the candidate accumulation units belonging to the same straight line based on the image centroid algorithm and finds the parameter information of the straight line; S4 updates the candidate accumulation unit position matrix and clears the information of the detected straight line; S5 repeats the above operation to analyze the updated candidate accumulation unit matrix, and finally completes the line detection; The step S3 includes: S3.1 One-dimensional array rowfinal=[r1,r2,…,r q ] T r i , where i = 1, 2, ..., q, and the accumulation units corresponding to the elements on the rows of the candidate accumulation unit position matrix are the accumulation units on the same straight line; The S3.2 centroid algorithm is to extract the centroid coordinates of the image; the (m+n) order of the pixel f(x,y) at the coordinate (x,y) is defined as: ； l 00 is the sum of the gray levels of f(x,y), for the first-order distance l 01 and l 10 l 00 Standardization gives the center of gravity coordinates G(x G ,y G ); The peak extraction method based on the image center of gravity uses the position coordinates of the candidate accumulation units of the same straight line in the accumulation matrix H to weight the corresponding accumulation values, and sums the weighted accumulation values ​​of the same straight line. Then, the position coordinates of the accumulated mean of the candidate accumulation units of the same straight line are calculated, which is the so-called image center of gravity. According to the following formula, the element h of the accumulation unit of the same straight line extracted from the accumulation matrix H is i,j Calculate, where i and j are the row and column numbers of the element in the matrix H respectively: ； ； ； ； ； I G , J G After rounding, this is the peak coordinate to be extracted from this straight line.

2. The Hough transform parameter space peak extraction method for power line detection in complex environments according to claim 1 is characterized in that: Set a threshold, extract the accumulation units that meet the threshold conditions, record their position coordinates in the entire accumulation matrix H, and store them in a two-dimensional array, called the candidate accumulation unit position matrix C; the first column element of this matrix represents the row number of the accumulation unit that meets the threshold conditions, and the second column element represents the column number of the accumulation unit that meets the threshold conditions; the accumulation units that meet the threshold conditions are called candidate accumulation units; 。 3. The Hough transform parameter space peak extraction method for power line detection in complex environments according to claim 1 is characterized in that: The step S2 includes: S2.1 defines a one-dimensional array rowfinal to store the row number of the candidate accumulation unit position matrix in which the position coordinates of the accumulation units located on the same line are recorded; S2.2 extracts the latest row number of the rowfinal record, denoted as k, and the initial record is 1; S2.3 Subtract the element in the kth row from the element in the first column of the candidate accumulation unit position matrix according to the following formula; ； If the absolute value of the difference If it is less than T1, the element in the kth row is subtracted from the element in the kth row in the second column of the candidate accumulation unit position matrix according to the following formula; ； If the absolute value of the difference If it is less than T2, the row number of the element is recorded in rowfinal, and the parameters determined by the two accumulation units are considered to correspond to the same straight line; If the elements in the k+1th row of the candidate accumulation unit position matrix do not satisfy the above two conditions, then the above operations are performed on the elements in the k+2th, k+3th, ... rows of the candidate accumulation unit position matrix according to the following formula until the above conditions are satisfied or the operations on the elements in the pth row of the candidate accumulation unit position matrix are completed; ； ； 。 4. The Hough transform parameter space peak extraction method for power line detection in complex environments according to claim 1 is characterized in that: The step S4 includes: S4.1 clear the corresponding elements of the row number of the candidate accumulation unit position matrix C recorded in the one-dimensional array rowfinal to zero; S4.2 clears rowfinal, and after clearing, records the first non-zero row number in the first column of the candidate accumulation unit position matrix C in rowfinal.

Images