RU2777718C1 - Method for non-destructive optical and visual testing of items by means of computer vision - Google Patents

Abstract

FIELD: testing.

SUBSTANCE: invention relates to methods for non-destructive testing in the fields of mechanical engineering, production of home appliances and military machinery, building, metallurgy. Claimed method consists in forming a depth map for the surface of metal with reflective properties by means of a visible spectrum IP video camera and laser illumination for the purpose of detecting non-flatness defects. Method for non-destructive testing of sheet metal consists in the IP camera positioned at an angle of 15° to the surface of the tested item continuously filming the target with the reflection of the laser illumination lines from the surface of the product, directed at an angle of 30° to the surface of the tested item. The item is moved toward the target. If non-flatness defects are present in the item, the line on the screen of the target at the point corresponding to the location of the defect on the sheet deviates relative to the conditional "zero" line on the target, depending on the direction of strain. The reflected beam is recorded by the IP camera at such a set frame per second rate that allows filming with the required resolution and preprocessing the obtained data. As a result, a height map of the target is formed for each conventional unit surface of the item, determined by the speed of movement of the item.

EFFECT: simplicity of implementation and high accuracy of the results of non-destructive testing of sheet metal.

1 cl, 24 dwg, 1 tbl

Description

The invention relates to methods for non-destructive testing of sheet metal with a reflective surface in the fields of mechanical engineering, production of household appliances and military equipment, construction, metallurgy.

Usage: for non-destructive testing of products.

From the existing prior art, contact and non-contact methods for determining the non-flatness of a surface are known. The main disadvantages of the mentioned well-known methods of non-destructive testing, in the case of metal sheets up to 1 mm thick, are insufficient sensitivity (and, as a result, measurement accuracy) to defects whose linear dimensions do not exceed tens of microns, as well as a high probability of damage to the sheet surface during testing. control.

A device for laser control of workpieces of machine parts is known (utility model patent RU 118737, G01B 11/00, 27.07.2012). The device includes a semiconductor laser, an image analyzer, a control device and a microprocessor unit for synthesizing a vector-raster model. A semiconductor laser illuminates the product under test in various planes with the help of a control device, at the same time, the image analyzer registers the image of the product in a fixed plane, which then enters the microprocessor-based synthesis unit of the vector-raster model, where the information is converted into a three-dimensional digital image of the product, taking into account surface defects. , which can then be exported to any CAD/CAM/CAE system for further analysis of surface defects.

The disadvantage of this method is the technical complexity of its implementation. In the claimed method, high-resolution residential IP cameras and software that analyzes frames received from IP cameras perform the function of an image analyzer and a microprocessor block for synthesizing a vector-raster model. Also, in the proposed method, instead of three-dimensional modeling, a panoramic image is used (the so-called "vertex map").

A series of laser 3D scanners LPX is known. LPX is a series of non-contact laser scanners designed for digitizing 3D products. LPX-60/600/1200 scans an object with a laser beam. The reflected beam is received by the receiver (CCD matrix). In the matrix, the laser beam is converted into an electrical signal. Depending on the size of the beam, the value of the distance to the object is determined.

The disadvantage of LPX laser scanners is the impossibility of determining external surface defects of parts, such as shrinkage cavities, nicks, ledges.

A device for detecting and measuring surface defects is known (patent for invention RU 2335734, IPC G01B 11/24, 10.10.2008), which makes it possible to simultaneously measure surface defects in two mutually perpendicular sections and obtain digital images of defects for computer processing, documentation and storage of results measurements.

The disadvantage of the device for detecting and measuring surface defects is the inability to perform digitization of three-dimensional products and the inability to indicate the detected defects within a short period of time after receiving frames from the IP camera.

A device for non-contact measurement of the surface roughness of parts of complex shape (utility model patent RU 104697, IPC G01B 11/30, 2011.05.20) is known.

Common to the known and claimed methods is an optical-visual method for obtaining data on the current geometry of the controlled product and the state of its surfaces.

The disadvantage of this method is the need to mount the housing with a lens and an illuminator on a rod rigidly connected to a stand mounted on the base, while a rotation drive is placed on the rod, which is kinematically connected to the housing. The advantage of the proposed method lies in the absence of the need to move the camera lens and the linear laser in the vicinity of the controlled product and, as a result, a smaller amount of work to modernize the production equipment.

The general advantage of the proposed method in comparison with known methods is the relatively less complexity of the technical implementation of the method. The equipment used in the method (target, line lasers and IP cameras) has a fairly wide range of applications and is more common on the market of similar devices, while together providing high accuracy of optical-visual control due to in-depth software analysis of received frames with an image surface of the controlled product.

The main common difference of the proposed method from the described well-known methods is the use of the reflective properties of the surface of the controlled product, the method for obtaining a scan of the image of its surface, the set of equipment used, the algorithmic analysis of the images of the controlled product and the flexible setting of the filtering parameters for detected defects according to tolerance.

A device for flaw detection and laser control for CNC machines is known (utility model patent RU 162587, IPC G01B 11/30, 06/20/2016). The method implemented by the specified device is selected as the closest analogue. Common to the known and claimed methods is an optical-visual method for obtaining data on the current geometry of the controlled product and the state of its surfaces. However, the implementation of this method is associated with significant, in comparison with the proposed method, the complexity of technical operation and the impossibility of indicating the detected defects within a short period of time after receiving frames from the IP camera.

The claimed method uses a residential IP camera instead of image analyzers that are used in the known method. Image analyzers are more expensive and more difficult to operate than IP cameras. The IP camera transmits images of the reflected lines of the laser beam on the target directly to software installed and running on a separate computer. The function of the analyzer is performed by software on a computer built on the basis of classical components.

The frames received from the IP cameras collectively form a vertex map, which is processed by the software. This is not a full-fledged 3D modeling, i.e. the operation is less resource-intensive.

Performing non-destructive testing of products of complex shape using the described complex is impossible due to the appearance of shadows. However, the possibility of organizing the control of geometrically complex products exists, for which the following requirements must be met:

a) the surface of the product must have reflective properties;

b) it must be technically possible to install a line laser and an IP camera at the required angle to the controlled surface;

c) it is possible to describe the surface of the controlled product with a complex function and perform an approximation if necessary.

The similarity of the proposed method and known from RU 162587 U1 is the possibility of non-destructive testing without removing the controlled product from the machine and the use of lasers of various types designed to identify various defects.

The objective of the claimed invention is to create an easy-to-implement and reliable method that is sensitive to individual microdefects and areas with any density of microdefects, providing high-precision results for optical-visual quality control of raw materials and finished products, and no need for immobility of the controlled product.

EFFECT: ensuring ease of implementation, high accuracy of the results of non-destructive testing of sheet metal.

The method consists in forming a depth map for a metal surface with reflective properties using an IP video camera of the visible spectrum and laser illumination in order to detect non-flatness defects.

A method for non-destructive testing of sheet metal is claimed, which consists in the fact that an IP camera, placed at an angle of 15° to the surface of the controlled product, continuously captures the target with reflection of laser illumination lines from the surface of the product, directed at an angle of 30° to the surface of the controlled product. The product moves towards the target. If there are defects in the non-flatness of the product, the line on the target screen at the point corresponding to the location of the defect on the sheet deviates relative to the conditional "zero" line on the target, depending on the direction of deformation. The reflected beam is recorded by the IP camera at such a set frame rate per second, which allows you to shoot with the required resolution and pre-process the received data. As a result, for each conventional unit of the surface of the product, determined by the speed of movement of the product, a map of heights on the target is formed.

The final image resolution depends on:

• number of horizontal pixels in the camera matrix (X-axis, product width);

• number of vertical pixels in the camera matrix (Z-axis, deviation depth);

• product movement speed and the number of frames captured by the camera per second (Y-axis, product length).

The sequence for detecting defects is as follows:

1. Pre-processing of images containing a height map is performed, during which the coordinates of surface points are established. Based on the coordinates, the image of the depth map is constructed according to the following rules:

• the position of the point along the X axis on the image received from the camera corresponds to the position of the point on the depth map;

• position of a point along the Y-axis on the image received from the camera corresponds to the pixel value on the depth map. The basis and offset for adjustment after the installation of the equipment are also set here;

• The position of the point along the Y axis on the depth map corresponds to the order of the images received from the camera.

2. The sheet contour is determined, for which a threshold algorithm is used, which detects contrast transitions from the “background” to the controlled product. The result is an array describing the position of the product in the frame, corresponding to the region of interest.

3. Points are marked, potentially corresponding to non-flatness defects, as a result of building a surface height map and processing it with software.

4. Sorting and forming segments with points according to the specified filtering rules. The segments are classified based on their geometric characteristics.

The method is illustrated with illustrations.

Figure 1 shows a schematic representation of the implementation of the method, where 1 is a target with control lines, 2 is a control sample, 3 is an IP camera,

figure 2 shows the frame received from the IP camera,

figure 3 - zone shooting the surface of products,

figure 4 - sample 1, zone 2, the original image without illumination,

figure 5 - sample 1, zone 2, image with backlight,

figure 6 - sample 1, zone 2, image with illumination, enlarged at one point,

figure 7 - sample 1, zone 6, backlit image,

figure 8 - sample 2, zone 1, the original image without illumination,

figure 9 - sample 2, zone 1, image with backlight,

figure 10 - sample 2, zone 2, the original image without illumination,

figure 11 - sample 2, zone 2, image with backlight,

figure 12 - sample 2, zone 3, the original image without illumination,

in Fig.13 - sample 2, zone 3, image with backlight,

in Fig.14 - sample 2, zone 4, the original image without illumination,

in Fig.15 - sample 2, zone 4, image with backlight,

in Fig.16 - sample 2, zone 5, the original image without illumination,

in Fig.17 - sample 2, zone 5, image with backlight,

Fig.18 - sample 2, zone 6, the original image without illumination,

Fig.19 - sample 2, zone 6, backlit image,

figure 20 - sample 3, the original image without illumination,

figure 21 - sample 3, image with backlight,

Fig.22 - sample 4, the original image without illumination,

Fig.23 - sample 4, image with backlight,

Fig.24 shows three-dimensional images of the surfaces of sheets with various defects. The software, depending on the sensitivity settings, detects any local deviations and traces - wrinkling, fracture, corrugation, pressure.

The method is carried out as follows.

According to FIG. 1 target with control lines 1 is fixed perpendicular to the reflective surface of the sample. Control sample 2 is placed on a flat horizontal surface and illuminated by a line laser at an angle of 30° to the sample surface. Residential IP camera 3 with a resolution of 1920×1080 pixels. fixed on the side of the line laser.

The sample moves towards the target. Control lines are projected onto the surface of the sample. In the case of bulges and other non-flatness defects, the reflected control lines are distorted, which is recorded by the IP camera. In FIG. 2 shows a frame received from an IP camera. The resulting frame stream is pre-processed by the software with the formation of a depth map, after which the result is displayed in the working window of the Internet browser.

To indicate deviations from flatness, the software highlights the points of the target lines, the trend lines of the curves actually reflected by the product, and local deviations on the trend lines from the target lines, the so-called “break points” of the trend line.

To indicate surface irregularities while maintaining the flatness of the product, the software selects, filters, segments and aggregates lines and groups of lines according to specified properties, provided that the defects are sufficiently contrasting. To detect defects with minimal contrast, aggregation is disabled.

The tests were carried out using metal sheets 210 × 297 mm in size from rolled products of grade 23Kh15N5AM3-Sh (VNS9-Sh, h = 0.3 mm).

Table 1 shows the non-flatness defects of the sheet - an imprint from the sleeve, located perpendicular to the direction of rolling, and longitudinal stripes with branches in the direction of rolling.

Table 1 - Non-flatness defects considered Sheet flatness defects Samples 1, 2 Sleeve imprint located perpendicular to the rolling direction Samples 3, 4 Longitudinal strips with branches in the direction of rolling

To detect non-flatness defects, the following solution was applied: the controlled sample is located on a flat horizontal surface. The line laser is installed and directed at an angle of 30° to the surface of the product, the IP camera and the target are placed in accordance with Fig. 1. Target lines are reflected by the surface of the product. Reflection is distorted at non-flatness points of the sheet.

In accordance with FIG. 2, the reflected lines do not have local deviations or breaks from the trend line for flat areas of the sheet, while breaks are visible in areas of non-flatness, taking into account the viewing angle.

Areas for shooting the surface of products - in accordance with Fig. 3.

The points of the target lines are yellow lines, the trend lines of the reflection curves of the target are pink lines, local deviations on the trend lines of the curves reflected from the target are red circles.

The software highlighted areas of local deviations above the tolerance on the trend lines of the detected curves from the target. The actual defect lines run horizontally on the sample along the centers of the red circles. The detected break points of the lines correspond to the boundaries of the section on the sample indicated by the marker.

For sample 2, the method for detecting defects is similar to the method used in the control of sample 1.

Sample 3 defects are clearly expressed in the conditional zone bounded by black straight lines.

Defects in sample 4 were detected by the method used in cases of control of samples 1 and 2. The defect is less pronounced.

The result of the work is the confirmation of the effectiveness of the described invention and the classification of defects, indicating the methods for their detection and recognition methods.

The invention was implemented in a hardware and software system that provides the following functionality:

• automatic sheet feeding to the inspection unit and return for visual inspection;

• laser illumination;

• detection of nonflatness defects;

• visualization (location, part and class) of detected defects;

• the ability to view the inspection history (both the current sheet and the previously passed inspection);

• automatic decision-making based on the tolerances of defects set by the operator and informing the operator about the suitability of products for release;

• generation of reports on detected defects for a specified period;

• automatic photographic recording (saving images with or without highlighting defects, together with calculated data);

• data transfer to any information system;

• Possibility of test control of table drive and line laser.

Thus, the claimed method:

a) simple in technical implementation, because common equipment is used and its scope is wider than in known methods;

b) allows to obtain high accuracy of the results of non-destructive testing of sheet metal in a short period of time;

c) provides low labor intensity of reading and analyzing the results of the software operation.

Claims (1)

A method for non-destructive testing of sheet metal, which consists in the fact that by means of an IP camera placed at an angle of 15° to the surface of the controlled product, a continuous shooting of the target is carried out with the reflection of laser illumination lines from the surface of the product directed at an angle of 30° to the surface of the controlled product, with In this case, the product is moved towards the target, in the presence of defects in the non-flatness of the product, the line on the target screen at the point corresponding to the location of the defect on the sheet deviates relative to the conditional "zero" line on the target, depending on the direction of deformation, the reflected beam is fixed by an IP camera with such a set frame rate per second, which allows you to perform shooting with the required resolution and pre-process the received data, as a result, for each conventional unit of the surface of the product, determined by the speed of the product, a height map is formed on the target.

Images