CN110084779A - A kind of extraction of aircraft thickness covering end surface features point and denoising method based on laser scanning - Google Patents

Abstract

The invention discloses a method for extracting and denoising feature points on the end face of thick skin of an aircraft based on laser scanning. The Euclidean distance to the boundary feature point is used to calculate the neighborhood of the boundary feature point, and then the sharp features and the boundary feature points at the short side of the end face are denoised, and finally the missing feature points at the sharp features are re-extracted. The characteristics of the present invention are: 1) Compared with the method of extracting feature points by a single scanning line, the feature points extracted by this method are more stable and have higher precision; 2) For skin end faces of different shapes, this method can Get better feature points and provide accurate data for end face fitting; 3) Easy to use and high efficiency.

Description

A Method of Extracting and Denoising Feature Points of Aircraft Thick Skin End Face Based on Laser Scanning

technical field

The invention relates to an aircraft manufacturing technology, in particular to an aircraft skin manufacturing technology, in particular to a method for extracting and denoising feature points of an aircraft thick skin end face based on laser scanning.

Background technique

As we all know, in the process of aircraft skin manufacturing, especially in the process of zigzag skin processing, in order to plan the processing trajectory reasonably, it is necessary to sort the end face scan line data and reconstruct the end face. Use the end face to extract the repair amount, and use the end face to intersect with the upper surface of another skin for subsequent processing trajectory extraction. When extracting point cloud feature points, first extract the boundary feature points of the skin end face. At present, many scholars at home and abroad have studied the boundary feature recognition of measurement data, mainly using the curvature extreme value method: Milroy M.J and Yang use the local coordinate system The quadratic polynomial surface in the point cloud data is used to estimate the curvature value of the point cloud data, and the extreme point of curvature is obtained, and the boundary points are extracted from it; Hu Xin et al. use the gradient solution method in image processing to estimate the normal vector of each point in the point cloud and curvature, the candidate boundary points are obtained through the threshold. The main problems of these methods are the poor stability of the feature points, low precision and low efficiency. The scan line point cloud is orderly, and the boundary feature points can be extracted according to the point-to-plane distance, which is an effective method.

SUMMARY OF THE INVENTION

The purpose of the present invention is to solve the problems of poor stability, low precision and low efficiency in the existing skin end face feature point extraction method, and combine laser scanning technology to invent a laser scanning based aircraft thick skin end face feature point extraction and denoising method.

Technical scheme of the present invention is:

A method for extracting and denoising feature points of aircraft thick skin end faces based on laser scanning, characterized in that it includes the following steps

1) Scan the side of the thick skin, and extract the boundary feature points through the point-to-plane distance;

2) Calculate the boundary feature point neighborhood according to the Euclidean distance from other points in the point cloud to the boundary feature point, and then denoise the sharp features and the boundary feature points at the short side of the end face;

3) Re-extract the missing feature points at the sharp features.

Calculate the Euclidean distance from other points in the point cloud to the boundary feature points, and then arrange these points in ascending order of Euclidean distance, use the radius r to intercept the nearest point as a neighborhood, and pass the angle between the boundary feature point and the neighboring point projection point connection line Denoise sharp features and boundary feature points on end faces.

The described boundary feature point extraction method is:

The scanning results of the skin are divided into blocks, and M scanning lines are extracted each time for processing. The value of M is between 20 and 50. When obtaining the boundary feature points on one side, first extract the point P _i,j on the scanning line ( 0≤i≤29,4≤j≤24) to fit a plane whose equation is Ax+By+Cz=D; n points on a single scan line (n on each scan line may not necessarily be equal), set d _j0 , d _j1 , d _j2 are the distances from three adjacent points P _j , P _j+1 , P _j+2 to the plane respectively, and set a threshold as δ; then use the formula (1) to calculate the distance from the point to the plane, P _j , P _j+1 , P _j+2 to the plane are respectively d _j0 , d _j1 , d _j2 ; at this time, j gradually increases from 0, and when d _j0 , d _j1 , d _j2 are all less than the threshold for the first time, P _j is the boundary feature point, set the boundary feature point as the j'th point on a single scanning line at this time; when calculating the boundary feature point on the other side, gradually reduce j from n-2, when d _j0 , d _j1 , d When _j2 is less than the threshold for the first time, P _j+2 is the boundary feature point, and the boundary feature point at this time is the j'+k point on a single scanning line; the boundary feature points on both sides are P _j' and P _{j'+ k} , extract points P _j' , P _j'+[k/3] , P _j'+[2*k/3] , P _j'+k on each scan line as feature points;

The described boundary feature point denoising method is: first calculate the Euclidean distance from other points in the point cloud to the boundary feature point, then arrange these points in ascending order of Euclidean distance, intercept the nearest point with radius r as the N _r neighborhood, and pass The angle between the boundary feature point and the projection point of the neighborhood point is used to determine whether the boundary feature point is removed; due to the uncertainty of the number of points in the neighborhood, the linked list structure is used to store data, and the N _r neighborhood points of the boundary feature point P have The sequence is stored in the linked list _nrlist(P,r _s ), and the neighborhood radius r _s is set to 3 to 5 times the point spacing; then it is judged whether the boundary feature points should be removed, the specific steps are as follows:

Step 1 The points in the neighborhood fit the plane, and the neighborhood points are projected into the plane;

Step 2 sorts the projection points;

Step 3: connect the boundary feature points and projection points, and find the angle θ between adjacent straight lines;

Step 4 Analyze the included angle θ to find the maximum included angle θ _max . When θ _max < 120°, the boundary feature points are located inside the point cloud. At this time, the boundary feature points and the feature points on the same line as the boundary feature points should be removed ; Otherwise, the boundary feature points are located at the boundary of the point cloud and should be retained.

The re-extraction of feature points refers to the lack of sharp feature points after the boundary feature points are denoised, and needs to be extracted from the removed boundary feature points again; a threshold δ is set to find the distance D between the removed boundary feature points and the feature point P _{j j} , if D _j < δ, extract the boundary feature point; then calculate the distance between the boundary feature point and the feature point P _j+k/3 , P _j+2*k/3 , P _j+k and extract the corresponding boundary feature point as feature points.

Beneficial effects of the present invention:

1) Compared with the method of extracting feature points from a single scan line, the feature points extracted by this method are more stable and have higher precision;

2) For skin end faces of different shapes, this method can obtain better feature points and provide accurate data for end face fitting;

3) Easy to use and high efficiency.

Description of drawings

Figure 1 is a schematic diagram of boundary feature point extraction.

Figure 2 shows the method for identifying boundary feature points.

Figure 3 shows the boundary feature points at short sides and sharp features.

Figure 4 shows the denoising of boundary feature points.

Figure 5 shows missing feature points missing at sharp features.

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

As shown in Figure 1-5.

A method for extracting and denoising feature points of aircraft thick skin end faces based on laser scanning, characterized in that it includes the following steps

1) Scan the side of the thick skin and extract feature points through the point-to-plane distance;

2) Calculate the boundary feature point neighborhood according to the Euclidean distance from other points in the point cloud to the boundary feature point, and then denoise the sharp features and the boundary feature points at the short side of the end face;

3) Extract the missing feature points at the sharp features.

The specific calculation algorithm is as follows:

1) Feature point extraction:

The scanning result of skin A is divided into blocks, and 30 scan lines (or any number between 20 and 50) are extracted for processing each time. When obtaining the boundary feature points on one side, first extract the point P on the scan line _{i, j} (0≤i≤29, 4≤j≤24) fit a plane equation Ax+By+Cz=D. The planes A, B, C, and D are the constants calculated by the fitted plane, such as Figure 1(a); n points on a single scanning line (n on each scanning line is not necessarily equal), as shown in Figure 1(b), let d _j0 , d _j1 , and d _j2 be three adjacent points P _j , For the distances from P _j+1 , P _j+2 to the plane, set a threshold as δ; then use the formula (1) to calculate the distance from the point to the plane, and the distances from P _j , P _j+1 , P _j+2 to the plane are respectively d _j0 , d _j1 , d _j2 ; at this time j gradually increases from 0, when d _j0 , d _j1 , d _j2 are all less than the threshold for the first time, P _j is the boundary feature point, and the boundary feature point at this time is set as a single scanning line The j'th point; when finding the boundary feature point on the other side, gradually reduce j from n-2, and when d _j0 , d _j1 , and d _j2 are all smaller than the threshold for the first time, P _j+2 is the boundary feature point, and set At this time, the boundary feature point is the j'+k point on a single scan line; the boundary feature points on both sides are P _j' and P _j'+k , and the points P _j ' and P _j ' on each scan line are extracted _+[k/3] , P _j'+[2*k/3] , P _j'+k as feature points;

2) Boundary feature point denoising:

When scanning the skin butt joint surface, the scanning situation shown in Figure 3 will appear at the short side and sharp features of the butt joint surface. At this time, the extracted boundary feature points are located inside the side surface or on the short side of the butt joint surface, and need to be denoised. Before denoising, the point cloud is deleted, and the points between the boundary feature points and the boundary feature points are taken on each scan line and the points on the side are extracted. In order to denoise, the side of the skin is scanned separately after extracting the boundary feature points so that the boundary feature points on the short side of the docking surface are also located inside the point cloud. At this time, the two cases are regarded as one case.

First calculate the Euclidean distance from other points in the point cloud to the boundary feature points, then arrange these points in ascending order of Euclidean distance, use the radius r to intercept the nearest point as the N _r neighborhood, and connect the boundary feature points with the neighbor point projection points The angle between them is used to determine whether the boundary feature point is removed. Due to the uncertainty of the number of points in the neighborhood, a linked list structure is used to store data, and the N _r neighborhood points of the boundary feature point P are stored in the linked list _nrlist(P,r _s ), and the neighborhood radius r _s is usually set to 3 to 5 times the dot pitch. Then judge whether the boundary feature points should be removed, the specific steps are as follows:

Step 1: The points in the neighborhood fit the plane, and the neighborhood points are projected into the plane.

Step 2 sorts the projected points.

Step 3: connect the boundary feature points and projection points, and find the angle θ between adjacent straight lines.

Step 4 Analyze the included angle θ to find the maximum included angle θ _max . When θ _max < 120°, the boundary feature points are located inside the point cloud. At this time, the boundary feature points and the feature points on the same line as the boundary feature points should be removed ; Otherwise, the boundary feature points are located at the boundary of the point cloud and should be retained, as shown in Figure 4.

3) Feature point re-extraction

As shown in Figure 5, after the boundary feature points are denoised, the sharp feature points are missing, and it needs to be extracted from the removed boundary feature points again. Set a threshold δ, find the distance D _j between the feature point of the removed boundary and the feature point P _j , if D _j < δ, then extract the boundary feature point. Then calculate the distance between the removed boundary feature point and the feature point P _j+k/3 , P _j+2*k/3 , P _j+k and extract the corresponding boundary feature point as the feature point.

The parts not involved in the present invention are the same as the prior art or can be realized by adopting the prior art.

Claims (6)

1. a kind of aircraft thickness covering end surface features point based on laser scanning extracts and denoising method, it is characterised in that it includes such as Lower step

1) edge feature point is extracted to plan range by point in the side for scanning thick covering；

2) boundary characteristic vertex neighborhood is found out according to the Euclidean distance of other point to edge feature points in cloud, then to sharp features And edge feature point carries out denoising at the short side of end face；

3) characteristic point lacked at sharp features is extracted again.

2. the aircraft thickness covering end surface features point according to claim 1 based on laser scanning extracts and denoising method, It is characterized in that, calculates the Euclidean distance that other points in point cloud arrive edge feature point, these are then pressed into Euclidean distance ascending order row Column, use radius r to intercept nearest point as neighborhood, by the angle between edge feature point and neighborhood point subpoint line to sharp The denoising of the edge feature point of feature and end face.

3. the aircraft thickness covering end surface features point according to claim 2 based on laser scanning extracts and denoising method, It is characterized in that, characteristic point is extracted again at sharp features.

4. the aircraft thickness covering end surface features point according to claim 1 based on laser scanning extracts and denoising method, Being characterized in that the boundary characteristic point extracting method is:

Piecemeal processing is carried out to the scanning result of covering, M scan line is extracted every time and is handled, M value between 20-50, Point P in scan line is first extracted when seeking side edge feature point_i,j(0≤i≤29,4≤j≤24) fit a plane equation For the plane of Ax+By+Cz=D；N point of single scan line (n in every scan line is not necessarily equal), if d_j0、d_j1、d_j2Point It Wei not adjacent 3 points P_j、P_j+1、P_j+2To the distance of plane, a fixed threshold value is δ；It recycles formula (1) to calculate point and arrives plane separation From P_j、P_j+1、P_j+2Distance to plane is respectively d_j0、d_j1、d_j2；J is gradually increased by 0 at this time, works as d_j0、d_j1、d_j2It is small for the first time P when threshold value_jFor edge feature point, if edge feature point is jth in single scan line ' a point at this time；Seek another lateral boundaries When characteristic point, j is gradually reduced by n-2, works as d_j0、d_j1、d_j2P when being respectively less than threshold value for the first time_j+2For edge feature point, if side at this time Boundary's characteristic point is jth '+k points in single scan line；Finding out two sides edge feature point is P_j'And P_j'+k, extract every scan line Upper P_j'、P_j'+[k/3]、P_j'+[2*k/3]、P_j'+kAs characteristic point；

5. the aircraft thickness covering end surface features point according to claim 1 based on laser scanning extracts and denoising method, Being characterized in that the edge feature point denoising method is: calculate first other points in point cloud to edge feature point it is European away from From, then by these press Euclidean distance ascending order arrangement, use radius r to intercept nearest point as N_rNeighborhood passes through boundary characteristic Angle between point and neighborhood point subpoint line judges whether the edge feature point removes；It is uncertain due to what is counted in neighborhood Property, using list structure storing data, by the N of edge feature point P_rNeighborhood point is orderly stored in chained list _ nrlist (P, r_s) in, Radius of neighbourhood r_sIt is set as 3~5 times of a spacing；Then judge whether edge feature point should remove, the specific steps are as follows:

Point fit Plane in step 1 neighborhood, neighborhood point project in plane；

Step 2 is ranked up subpoint；

Step 3 fillet characteristic point and subpoint seek the angle theta between adjacent straight line；

Step 4 analyzes angle theta, finds out maximum angle θ_max, work as θ_maxEdge feature point is located inside point cloud at 120 ° of <, It edge feature point and should be removed with the characteristic point on the same straight line of edge feature point at this time；Otherwise edge feature point is located at point cloud Boundary, it should retain.

6. the aircraft thickness covering end surface features point according to claim 1 based on laser scanning extracts and denoising method, Be characterized in that the characteristic point extract again refer to edge feature point denoising after point missing at sharp features, need again from removal Edge feature point in extract；One threshold δ is set, removal edge feature point and characteristic point P are asked_jDistance D_jIf D_j< δ is then mentioned Take the edge feature point；Then removal edge feature point and characteristic point P are asked_j+k/3、P_j+2*k/3、P_j+kDistance simultaneously extracts corresponding boundary Characteristic point is characterized a little.