WO2014198227A1

WO2014198227A1 - Line laser ranging method used for self-moving robot

Info

Publication number: WO2014198227A1
Application number: PCT/CN2014/079742
Authority: WO
Inventors: 汤进举
Original assignee: 科沃斯机器人有限公司
Priority date: 2013-06-14
Filing date: 2014-06-12
Publication date: 2014-12-18
Also published as: CN104236521A

Abstract

A line laser ranging method used for a self-moving robot, comprising: step 100: arranging a line laser (A) on a self-moving robot as an emission source, the emitted line laser being projected onto a target object (C), and an image sensor (B) shooting an image containing a laser stripe, which is projected onto the target object (C); step 200: determining a local image processing region on the image containing the line laser stripe; step 300: conducting distortion correction on the local image processing region, and determining the true position of the line laser stripe on the image; step 400: searching for an energy centre line in the line laser stripe after the distortion correction, and determining the position of the energy centre line on the image; and step 500: according to a pixel coordinate of the energy centre line on the image, obtaining the real distance (q) from the self-moving robot to the target object (C). The method only processes a local region of the image, which reduces the amount of data which needs to be processed, thereby improving the processing efficiency and saving working time on the premise that the accuracy rate is not affected.

Description

Line laser ranging method for self-mobile robots

The invention relates to a line laser ranging method, in particular to a line laser ranging method for a self-mobile robot, belonging to the technical field of small household appliance manufacturing. Background technique

The core technology of the mobile robot is to perceive the surrounding environment, then plan the walking route, effectively traverse the various areas, and complete the operations in each area. The most basic and critical part of this technology is the "sensing" module, the laser ranging sensor.

One of the existing sweeping robots is that the distance measuring sensor is not used, the obstacle is isolated by the collision plate, and then the ground is randomly cleaned. This will result in multiple invalid cleanings, such as: some corners are not cleaned or some areas are cleaned multiple times, the purpose of the work is blind and the level of intelligence is low, time-consuming and laborious. Some similar applications now use laser ranging sensors as their intelligent front-ends, but in these applications, the principle is to use a single-point laser as the light source, and the receiving device to form a transmitting-receiving system, by time, space, etc. Calculate the distance of a space object. Usually, when measuring the section distance, based on single-point laser ranging, combined with the rotating device, a two-dimensional section distance measuring device is constructed. The drawbacks of this type of distance measuring device and method are: the rotating tourb machine introduces noise, the system is not stable enough, the service life is short and the cost is high.

Another type of existing sweeping robot uses a distance detector to detect the distance between the robot and the obstacle. The sweeping robot avoids the ground obstacle in real time according to the distance detected by the detector to avoid collision. Alternatively, the sweeping robot can also position itself according to the distance information, and can also use the distance information to establish an indoor map, which facilitates the planning and setting of the subsequent walking path.

At present, most of the sweeping robots on the market use ultrasonic sensors or infrared sensors to detect the obstacle distance, but the ultrasonic signal or infrared signal is susceptible to external interference and the detection distance is small and the precision is not high. In order to further improve the range and accuracy of distance measurement, existing sweeping robots have begun to use laser ranging sensors to detect the distance of obstacles. For example, Chinese patent CN101809461 discloses a sweeping robot, which uses a laser ranging sensor, including a point laser source and an image sensor. The laser ranging sensor calculates the distance information of the obstacle according to the triangulation method. In order to obtain the distance information of the entire indoor obstacle, the laser detector on the sweeping robot is also provided with a rotatable base, so that the laser detector can scan the obstacles in the room at a sustainable 360 degrees. However, the rotating laser ranging sensor's rotating gyro opportunity introduces noise, the system is not stable enough, and the service life is short. And the cost is higher. In addition, Chinese patent CN101210800 also discloses a line laser ranging sensor comprising a line laser source and a camera. The line laser ranging sensor calculates the distance of the obstacle according to the pixel position of the laser stripe in the image, and the ranging method is to binarize the whole image to obtain laser stripe data, and then perform subsequent processing on the triangle after the data is processed. Obstacle distance. However, due to the need to process the complete image, the amount of information is large and the processing speed is slow. Summary of the invention

The technical problem to be solved by the present invention is to provide a line laser ranging method for a self-moving robot, which only partially images the captured line laser stripe image projected on the target object, in view of the deficiencies of the prior art. The processing of the area greatly reduces the data that needs to be processed, and improves the system efficiency without affecting the accuracy rate, so that the entire system runs at a double speed, saves working time and is convenient and reliable.

The technical problem to be solved by the present invention is achieved by the following technical solutions:

A line laser ranging method for a self-mobile robot, the method comprising the following steps:

Step 100: using a line laser disposed on the self-moving robot as a light source, the emission line laser is projected on the target object, and the image sensor captures an image of the line-containing laser stripe projected on the target object;

Step 200: Determine an image processing local area on the image of the line laser stripe;

Step 300: Perform distortion correction on the image processing local area to determine the true position of the line laser stripe on the image;

Step 400: Find an energy center line in the line laser stripe after distortion correction, and determine a position of the energy center line on the image;

Step 500: According to the pixel coordinates of the energy center line on the image, the actual distance from the mobile robot to the target object is obtained.

The step 200 specifically includes:

Step 201: traverse each line in the captured image with line laser stripe to find the point at which the laser gradation value is the largest in the line;

Step 202: N1 pixels are respectively taken on both sides of the point where the gray value found in each line is the largest, and the image processing local area is formed, where 10 N1 30.

The step 400 uses a centroid algorithm to find an energy center line, which specifically includes:

Step 401: In the image of the line laser stripe after the distortion correction, traverse each line to find the point at which the laser gradation value is the largest in the line;

Step 402: N2 pixels are respectively taken on both sides of the point where the gray value of the laser light having the largest value is found, wherein 5^N2^20;

Step 403: Performing a centroid method on the pixels to obtain an energy center line of the image of the line laser stripe;

The centroid algorithm formula is:

Yi i=∑ (f (χ, yi) *yi) / ∑ f ( , yi)

Where: yi i is the pixel column coordinate position after the centroid algorithm, (x, yi) is the original pixel coordinate (coordinate obtained after distortion correction), f (x, yi) is the gray at the (x, yi) position Degree value; yi i obtained in the above formula is the position of the energy center point of the line laser stripe image of each line; the energy center points of all the lines constitute the energy center line.

The curve fitting method can also be used in the step 400 to find the energy center line of the line laser stripe.

The step 500 specifically includes: establishing a triangular relationship formula according to the principle of triangulation, and obtaining the actual distance 9 from the moving object to the target object.

The step 500 specifically includes:

Step 501: Measure the energy center column coordinate yi i of the image of the line laser stripe corresponding to the _n actual distances _q , and establish a comparison table between the actual distance q and the energy center column coordinates;

Step 502: According to the energy center column coordinate yi i of the image of the actual line laser stripe, the actual distance q is obtained by an interpolation algorithm.

The n in the step 501 is greater than or equal to 20.

In summary, the present invention only performs local area processing on the captured image of the line laser stripe projected on the target object, so that the data to be processed is greatly reduced, and the system efficiency is improved without affecting the accuracy. It doubles the operating speed of the entire system and saves working time.

The technical solutions of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. DRAWINGS

Figure 1 is a schematic diagram of a similar triangular principle of the present invention. detailed description

Figure 1 is a schematic diagram of a similar triangular principle of the present invention. As shown in FIG. 1, the light emitted from the light source A is irradiated onto the target object C, and an image of the line-containing laser stripe projected on the target object C is captured by the image sensor B. In the present invention, the light source A is a line laser, and therefore, the spot shape projected on the target object C is stripe-like. A triangle AABC is formed by the light source A, the image sensor B, and the target object C. When the photographing focal length of the image sensor B is f, since the relative position between the light source A and the image sensor B is fixed, the graph The starting position a on the sensor film is determined. At this time, the emitted light emitted by the light source A is reflected back from the target object C, and the falling point on the film is b, and another triangular AabB is formed between the striped image and the image sensor. Obviously, AABC and AabB are similar triangles with a certain proportional relationship with each other. Since the light source A and the image sensor B are both disposed on the body of the mobile robot, the vertical distance between the target object C and the line connecting the two can be regarded as the actual distance q between the target object C and the self-mobile robot. In AABC, the value of the distance s between the light source A and the image sensor B is determined. In AabB, the value of f is determined, and the length X of the line stripe image can be fixed by value, using AABC. And AabB is a similar triangle, with a certain proportional relationship between each other, the actual distance q between the target object C and the self-mobile robot can be obtained.

Specifically, to the setting position of the light source A and the image sensor B on the self-mobile robot, the image sensor B can be disposed directly in front of the mobile robot, so that it is connected with the motherboard of the mobile robot, and provides the current location information of the motherboard. For the board to handle and make decisions. The light source A is disposed at the top of the front end of the mobile robot, substantially in the same vertical plane as the image sensor B.

Based on the above basic principle, the present invention provides a line laser ranging method for a self-mobile robot, the method comprising the following steps:

The step 200 specifically includes:

Step 201: traverse each line in the image of the captured line laser stripe to find a point at which the laser gradation value is the largest in the line;

Step 401: traverse each line in the image of the line laser stripe after the distortion correction, and find a point at which the laser gradation value is the largest in the line; Step 402: N2 pixels are respectively taken on both sides of the point where the gray value of the laser is found to be the largest, wherein 5^N2^20;

The centroid algorithm formula is:

Yi i=∑ (f (χ, yi) *yi) / ∑ f ( , yi)

The step 500 specifically includes:

The n in the step 501 is greater than or equal to 20.

Specifically, as shown in FIG. 1, the light source A in the present invention uses a line laser, and the emission line laser is projected on the target object C (shown in FIG. 1 is a single-point laser, line laser) It is an extension of the laser in the direction perpendicular to the paper.) A 640 X 480 CMOS image sensor is used to capture images with line laser stripes. Scanning each line of data containing the line laser stripe image by line scan, and discarding the unwanted background image, thereby forming line laser stripe (such as the maximum point of gray value and 30 points around) in the column direction of the image sensor. The line laser stripe processing is then performed to obtain the laser fringe energy centerline in the column direction. The values of Nl, N2, and n can be appropriately selected or adjusted according to the processing accuracy and speed.

In the above step 400, in addition to the centroid method for finding the energy center line of the line laser stripe, other methods such as a curve fitting algorithm may be used. The curve fitting method utilizes the principle that the cross-sectional intensity of the laser stripe approximates the Gaussian distribution. A sub-pixel-level optical strip center coordinate is obtained by curve fitting the pixel point coordinates and the gray value in the vicinity of the peak point of the light intensity distribution in the cross section. The specific steps are as follows: (1) Search each line sequentially for the processed image, and find the peak intensity point, that is, the maximum point of the gray value in the line, and set the point to M. (x., y.), the gray value is g. ; (2) Select M. The left neighbor M- ₂ (x., y), IVU (x., y- and the right neighbor Mi ^ y , M ₂ (x., y ₂ ), and their gray values are g- ₂ , g- _{1 ; gl} , g ₂ . Using the above 5 points, The quadratic curve fitting is performed according to the principle of least squares; (3) the coordinates of the maximal point in the direction are obtained by curve fitting. Then (x., which is the center point of the light stripe sought. The fitted curve equation can be a parabolic or Gaussian curve. This method is suitable for straight strips of light in the image where the normal direction does not change much.

Usually, due to the structural defects inherent in the camera inside the image sensor, the captured image will be deformed to a certain extent, especially in the direction of the edge of the image, the more severe the deformation. Therefore, the deformed image must be corrected by the process of distortion correction. Since the line laser fills a certain direction of the image, it is indispensable to correct the deformation to obtain a real image to obtain more accurate ranging accuracy. Through the distortion correction process, the deformed laser stripes in the image, such as the curved shapes on both sides, can be corrected to a true straight shape. However, if the entire image captured by the image sensor is subjected to distortion correction processing as in the prior art, it is necessary to traverse each pixel of the entire image, which requires a large amount of data and information processing. In the present invention, in step 200, the image processing local area is first determined on the image containing the line laser stripe, and then, in step 300, the image processing local area is corrected for distortion to determine the true position of the line laser stripe on the image. In this way, the distortion correction algorithm is used only for the obtained partial image, which can greatly speed up the calculation and speed up the image processing speed.

Local image distortion correction, including:

First, a laser stripe grayscale image is captured, and each line is traversed in the image to find a point where the laser gradation value is the largest;

Secondly, each pixel takes 30 pixels on each side of the maximum point of the laser gradation value, and performs local image distortion correction processing. Of course, the number of selected pixels is not limited to 30, and according to the processing precision or speed, Adjust the number of selected ones by yourself;

The distortion correction algorithm formula is:

1 1 Γ '+ΙίζΓ ⁴ 1⁄2 ₃ Γ +2*Pl (ΐ +2*χ )

( l+kir ² +k ₂ r ⁴ +k ₃ r ⁶ ) +2*p ₂ *xy+Pi* (r ² +2

Where: (x, is the original position of the original image distortion point captured on the image sensor, J is the corrected new position;

, , and respectively are the distortion coefficients of the camera, obtained by camera calibration; where r ² = f.

In the above step 400 of the present invention, the centroid algorithm is used to find the energy center line of the laser stripe, specifically, only the real laser stripe image is obtained in the previous step, and the data of the laser stripe in each line occupies several pixels. Location, no accurate calculations are possible. In order to calculate the corresponding spatial distance, it is necessary to extract the energy center line of the laser stripe, and the laser spot on the energy center line has a high pixel level on the image sensor, usually zero. A few pixel levels, which can more accurately correspond to the spatial distance.

After the corresponding data and information are obtained through the processing of the above respective steps, the actual distance q between the self-mobile robot and the target object C can be calculated. The calculation can be performed in the following two ways.

Specifically, the first method is to obtain the spatial distance corresponding to each center point of the laser stripe in the column direction by using the triangulation principle after obtaining the coordinates of the stripe energy center, thereby obtaining the laser line corresponding to Section distance. The method of obtaining the distance is also based on the principle of triangulation in the above figure, and a triangular relation formula is established, in which the variable is also yi i, and then calculated and obtained distance from other known parameters.

Another method is to calculate the distance by using the look-up table method, that is, each coordinate center coordinate corresponds to a spatial distance, and on this basis, a set of fixed tables consisting of the distance Z and the stripe center column coordinate yi i is established, that is, it needs to be advanced Measuring the data of n distances Z and the data of the column coordinates yi i corresponding thereto, thereby obtaining a linear relationship of the distance Z of the column coordinates yii, thereby establishing a table for reference calculation, in actual use, n according to the measurement The distance range is determined. For example, the measurement range within 5 meters needs to measure at least 20 sets of correspondence data, that is, the value of n is greater than or equal to 20, which is more reliable. After pre-testing to obtain this reference table, in actual Only the actual centerline coordinates of the stripe are obtained during the measurement, and then an interpolation algorithm is used to correspond to a spatial distance.

After obtaining all the spatial distances, the mobile robot can control the path planning and walking mode according to the relevant information.

In summary, the line laser ranging method for the self-mobile robot provided by the present invention solves the defect of the non-intelligent of the existing self-mobile robot through the line laser ranging sensor, and overcomes the existing ranging. In the technology, the single point laser and the rotating device bring about instability, high cost and short life. The distance measuring sensor of the present invention has high efficiency and cost performance in a short-range ranging application of 200 mm to 5000 mm indoors.

The laser ranging sensor of the present invention uses line structured light as a light source, which is easier to calibrate and adjust than a spot laser. The line structure light cooperates with the area array CMOS sensor to extract the depth distance information of the measurement target by the two-dimensional gray scale information in the CMOS image. In the line structure optical ranging, the camera needs to be calibrated to obtain the internal and external parameters of the camera. These parameters are used to correct the deformation of the image to obtain a true laser fringe image. Since the laser line segment of the line structure light modulated by the target object is a laser stripe having a certain width, the laser stripe obeys a Gaussian distribution in the cross section, so when calculating the spatial distance, the energy center of the line laser stripe needs to be extracted in the image. Line to calculate. The line structure light sensor is used as the "eye" of the robot to obtain the two-dimensional distance information within the sensor setting range, and provides the robot host with the indoor situation seen by the "eye". The host uses this information to locate, plan the route and establish an indoor map. The robot will make its own route according to the map to clean the ground, which makes the cleaning robot's work absolutely intelligent, and the efficiency is greatly improved.

Claims

claims

1. A line laser ranging method for self-moving robots. The method includes the following steps:

Step 100: Use the line laser set on the self-moving robot as the light source, emit the line laser and project it on the target object, and capture the image containing the line laser stripes projected on the target object through the image sensor on the self-moving robot;

Step 200: Determine the image processing local area on the image containing line laser stripes;

Step 300: Perform distortion correction on the local area of image processing to determine the true position of the line laser stripe on the image;

Step 400: Find the energy center line in the distortion-corrected line laser stripe, and determine the position of the energy center line on the image;

Step 500: According to the pixel coordinates of the energy center line on the image, obtain the actual distance (q) from the mobile robot to the target object.

2. The line laser ranging method for self-moving robots according to claim 1, characterized in that the step 200 specifically includes:

Step 201: Traverse each line in the captured image containing line laser stripes, and find the point with the largest laser grayscale value in the line;

Step 202: Take N1 pixels on both sides of the point with the largest grayscale value found in each row to form an image processing local area, of which 10 N1 30.

3. The line laser ranging method for self-moving robots as claimed in claim 1, characterized in that, in step 400, a centroid algorithm is used to find the energy center line of the line laser stripe, which specifically includes:

Step 401: In the distortion-corrected line laser stripe image, traverse each line and find the point with the largest laser grayscale value in the line;

Step 402: Take N2 pixels on both sides of the found point with the largest laser grayscale value, where,

5^N2^20;

Step 403: Perform the centroid method on these pixels to obtain the energy center line of the line laser stripe image;

The formula of the centroid algorithm is:

yi i=∑ (f (χ, yi) *yi) / ∑ f ( , yi)

Among them: yi i is the pixel column coordinate position after the centroid algorithm, (x, yi) is the original pixel coordinate (the coordinate obtained after distortion correction), f (x, yi) is the gray value at the (x, yi) position degree value; yi i obtained in the above formula That is, the location of the energy center point of the line laser stripe image of each row; the energy center points of all rows form the energy center line.

4. The line laser ranging sensing method for a self-moving robot as claimed in claim 1, characterized in that, in the step 400, a curve fitting method is used to find the energy center line of the line laser stripe.

5. The line laser ranging method for a self-moving robot as claimed in claim 1, characterized in that the step 500 specifically includes: establishing a triangular relationship formula according to the principle of triangular ranging, and obtaining the distance to the target from the self-moving robot. The actual distance of the object (q).

6. The line laser ranging method for self-moving robots according to claim 1, wherein the step 500 specifically includes:

Step 501: Measure the energy center column coordinates yi i of the line laser stripe images corresponding to n actual distances (q), and establish a comparison table between the actual distances (q) and the energy center column coordinates;

Step 502: According to the energy center column coordinates yi i of the actual line laser stripe image, obtain the actual distance (q) through the interpolation algorithm.

7. The line laser ranging method for self-moving robots according to claim 6, wherein n in step 501 is greater than or equal to 20.