CN103279763B - A kind of wheel hub type automatic identifying method of structure based feature - Google Patents

Abstract

The invention discloses a method for automatically identifying wheel hub types based on structural features. (1) For a wheel hub of an unknown model, it is necessary to obtain all the characteristic parameters of the wheel hub through learning and store them in a database; Image, extract the structural features of this type of hub, and use these parameters to realize automatic identification of different types of hubs. (2) During identification, compare the parameters of the wheel hub type stored in the database within the allowable range of error. When the characteristic parameters are all consistent, the wheel hub model of the identification object is obtained. If any parameter is inconsistent, it indicates that the wheel hub has not been recognized. Database does not exist in . The invention adopts the most stable feature of the wheel hub as the identification parameter, which can effectively avoid various misjudgments during the identification of the blank wheel hub, and improve the automation degree of the wheel hub production.

Description

A method for automatic identification of wheel hub type based on structural features

technical field

The invention relates to image segmentation, feature extraction and pattern recognition technologies belonging to the field of machine vision, in particular to a method for automatic recognition of wheel hub types based on structural features.

Background technique

Machine vision obtains external images through optical imaging and sensor technology, analyzes and processes image information, and replaces human eyes and brains with machines to describe and recognize objects in the objective world. Machine vision technology is a highly comprehensive high-tech, involving image processing, pattern recognition, artificial intelligence, automatic control and other fields. It has the characteristics of fast speed, strong stability, and multiple functions. It is widely used in modern industrial automation in production.

Wheel hub, also called rim, is an important driving and force-bearing part of motor vehicles such as automobiles and trains. At present, it is mainly produced by casting. The production of the wheel hub is firstly the melting and refining of raw materials, and then the rough product is obtained by low-pressure casting in the designed mold. After the quality is checked by the X-ray non-destructive testing equipment, heat treatment, high-precision machining and air-tightness inspection are carried out. Finally, the Coating process. In the automatic production line of the wheel hub, multi-variety wheels need to be produced in a mixed flow. The first problem to be solved is the identification of the wheel type, because the X-ray non-destructive testing of the rough wheel hub in the early stage, the machining of the semi-finished wheel hub in the middle stage, and the final coating treatment are different. Different models of hubs require different procedures. In the past, this problem has been solved manually by writing the model on the wheel hub with a pen, and then on the automatic production line, the staff will select the corresponding processing route according to the written model. This is inefficient and expensive, and it is far from meeting the needs of modern production. It can be seen that the automatic recognition technology of wheel hub models based on machine vision is of great significance to the automatic production of wheel hubs.

Domestic research on automatic wheel recognition technology has only started in recent years. The application publication date is November 28, 2007, and the invention patent with the publication number CN101079107 "Automatic Identification Method of Wheel Hub Model" discloses a technical scheme for wheel hub model identification for automatic detection of non-destructive testing equipment, and only describes the general outline of wheel hub identification. The train of thought does not involve specific identification steps and methods. People in the same industry have also conducted research on the automatic identification of wheel hub models, and proposed a system for wheel hub identification. The image processing algorithm and identification results used for identification are given, without consideration of the wheel hub. In the actual production process on the assembly line, the recognition parameters used are unstable, and the recognition methods used have a large amount of calculation, complex operations, poor robustness, and do not have real-time and versatility.

Contents of the invention

The purpose of the present invention is to provide a method for automatic recognition of wheel hub types based on structural features. The algorithm has a small amount of calculation, strong robustness, and fast execution speed, and can perform the same type of wheel hub from the initial blank product to the intermediate one. Semi-finished products and final finished products can be accurately identified, and are widely used in all links of the wheel hub automation production line. Tongjin, which meets the real-time requirements of the system, helps to realize the intelligentization of wheel hub production and management.

The technical scheme adopted in the present invention is:

A method for automatic identification of wheel hub types based on structural features, the implementation steps are:

(1), the establishment of the database of the structural characteristic parameters corresponding to the hub;

Step 1. Store the characteristic parameters such as the diameter of the hub, the number of spokes, the number of bolt holes, the appearance shape of the area where the bolt holes are located, and the area of the center hole into the database as the identification basis for this type of hub;

(2), identify the wheel hub on the production line, and get the conclusion that there is no such wheel type in the model of the wheel hub or in the database:

Step 2, collect the visible light image of the hub to be identified, and separate the appearance image of the hub;

Step 3. Find the circumscribed circle and its center of the separated hub, obtain the diameter data of the hub, and compare it with all the diameter data of the hub stored in the database. If there is the same wheel type, continue to identify, if not, end The identification process leads to the conclusion that this wheel type does not exist in the database;

Step 4. Take the center of the circumscribed circle of the hub as the center, draw a small circle with half the size of the hub as the diameter, divide the hub area into two parts, the inner circle and the ring, and calculate the spokes on the ring area between the circumscribed circle and the small circle The number of spokes, search for the wheel type with the same number of spokes as the current number of spokes within the database range limited in step 3, if there is, continue to identify, if not, end the identification process, and conclude that there is no such wheel type in the database in conclusion;

Step 5. Divide the hub image in the small circle to obtain the number of hub bolt holes, and search for the wheel type with the same number of bolt holes within the database range limited in step 4. If there is, continue to recognize, and if not, end the recognition process, and draw the conclusion that this wheel type does not exist in the database;

Step 6. Segment the wheel image in the small circle to obtain the appearance shape data of the area where the bolt holes of the wheel hub are located. Search for the wheel type with the same appearance shape of the bolt holes within the database range limited in step 5. If there is, continue to identify, if not Then end the identification process and draw the conclusion that there is no such wheel type in the database;

Step 7. Segment the hub image in the small circle to obtain the area of the center hole area of the hub. Search for a wheel type with the same area of the center hole area within the database range defined in step 6. If there is, obtain the final model of the hub , if not, draw the conclusion that there is no such wheel type in the database, and end the identification process.

Further, the database of the structural feature parameters corresponding to the hub is established by learning the hub to be identified, and the corresponding structural feature parameters are obtained and stored in the database:

Step 1. Collect the visible light image of the hub, and accurately separate the appearance image of the hub from it;

Step 2, process the separated hub image, find out the circumcircle of the hub and its center, and obtain the diameter size data of the hub;

Step 3. Take the center of the circumscribed circle of the hub as the center, draw a small circle with half the size of the hub as the diameter, and divide the hub area into two parts: the inner circle and the ring;

Step 4, counting the number of holes on the hub image on the above-mentioned ring area to obtain the data of the number of spokes;

Step 5. Segment the hub image in the small circle to obtain the number of hub bolt holes, the appearance shape of the bolt hole area and the area of the center hole area, and store all the data obtained in the above steps into the database as the final product of this type of hub. Identify parameters.

Compared with prior art, the present invention has following advantage:

1. Select the geometric appearance features of the hub as all identification parameters. When the wheel hub is cast in the mold, its geometric shape has been basically determined. Even after the subsequent precision machining and painting processes, these parameters remain basically unchanged. Using them as the basis for identification can ensure that the wheel hub in each production process. All links can be accurately identified, which effectively avoids the disadvantages of other technologies when using the grayscale features of the wheel image as identification parameters, such as specific requirements for lighting conditions, and can only be applied before a specific link in wheel hub production.

2. The identification of the five characteristic parameters is carried out in sequence. First, the approximate range of the hub to be identified is determined by the size of the hub, and secondly, the identification area is further narrowed by means of the number of spokes. After the two are determined, the bolt hole is examined. The number of bolt holes, the appearance shape of the area where the bolt holes are located, and the area of the central hole area. As long as one of the above parameters does not match in the database, the identification process will end, and the conclusion that the hub cannot be identified can be concluded only when they are all the same. Determine the model of the hub. The above identification steps effectively draw on the idea of manually identifying the wheel hub model, first compare the most obvious and stable features, and gradually narrow the matching range layer by layer, which not only reduces the amount of calculation in the identification process, but also ensures the accuracy of the identification results , which is significantly better than conventional pattern classification and recognition methods.

3. When extracting the characteristic parameters of the wheel hub through the image segmentation technology, the geometric characteristics of the wheel hub structure are fully considered, and this prior knowledge is integrated into the image segmentation step, which effectively eliminates the differences in the appearance of the rough wheel hub, semi-finished and finished wheels. The possible misjudgment improves the recognition accuracy of the present invention and expands the scope of application of the present invention.

Description of drawings

The specific implementation manners of the present invention will be described in further detail below in conjunction with the accompanying drawings.

Fig. 1 is the system flow diagram of the learning and training hub of the present invention;

Fig. 2 is a flow chart of the system for identifying the hub model of the present invention.

detailed description

Below in conjunction with accompanying drawing and example the present invention is further described.

A method for automatic identification of wheel hub types based on structural features disclosed in the present invention includes two major steps of learning and training of wheel hubs and automatic identification, as shown in FIG. 1 . Before identifying a new wheel hub, it is first necessary to obtain all the characteristic parameters of this type of wheel hub through learning and training, including the following steps:

Step 1. Collect the visible light image of the wheel hub, and accurately separate the appearance image of the wheel hub from it. Specifically, for a pair of wheel hub images acquired by a CCD camera, the wheel hub must first be extracted from the entire image. Under the illumination of visible light sources, the gray level of the image of the entire wheel hub is significantly higher than that of the background, which is a very stable feature when collecting wheel hub images. In view of this feature, the method of setting a fixed threshold is used to achieve the extraction of the hub: the value greater than this threshold is 1, corresponding to the extracted hub, and the value smaller than this threshold is 0, corresponding to the background. If the segmentation result when the threshold is set to 60, the entire wheel hub is completely extracted, and the threshold value in specific applications can be determined in combination with the brightness of the visible light source in the image acquisition environment;

Step 2. Carry out circle fitting processing on the binarization result of the separated hub image to obtain the circumcircle of the hub and its center. The diameter of the circumscribed circle is the diameter of the hub, which is stored as the first parameter obtained in the learning process. into the database;

Step 3. Take the center of the above circle as the center and draw a small circle with half the radius of the circumscribed circle as the radius to get the second circle on the hub image. The small circle divides the hub area into two parts, the inner circle and the ring. According to the geometric structure of the hub , the inner circle must completely contain the bolt hole and center hole of the hub, and the annular area between the outer circle and the inner circle is the spoke part of the hub;

Step 4. Directly count the holes in the image in the ring area. By virtue of the fact that the areas of the holes are equal and the positions are distributed at equal angles, the interference caused by the burrs during the casting of the wheel hub is removed. The number of holes finally obtained is the number of holes in the wheel hub. The number of spokes, from which the second parameter of the learning process is obtained;

Step 5. After obtaining the two characteristic parameters of the wheel hub diameter and the number of spokes, start to analyze the image area determined in the small circle (the minimum image area including the small circle), in which the gray level of the bolt hole is brighter than that of the surrounding area, and the center The hole area is in the geometric center of the image, corresponding to the gate part of the hub casting, and the gray scale is darker than the surrounding area. Segment the image for the first time, the purpose is to find the bolt hole area, which is characterized by high gray level, equiangular distribution in space, and each area is independent of each other, and the area is approximately equal; based on this, the image is segmented, and the target is segmented The number of regions is the number of bolt holes, and the perimeter area ratio of the target region is used as the identification standard for the shape of the bolt hole region. The second segmentation of the image is to find the center hole area of the wheel hub, and record the area of the segmentation result because it is located in the geometric center and the gray value is lower than that of the surrounding area, that is, the last parameter of learning and training, the area of the center hole area.

The characteristic parameters including the diameter of the hub, the number of spokes, the number of bolt holes, the appearance shape of the area where the bolt holes are located and the area of the central hole are stored in the database as the identification basis for this type of hub; the same type of hub only needs to Learn and train once, just call the parameters in the database in the specific recognition.

Figure 2 is the identification of the hub on the production line. The specific identification process of the hub still follows the same sequence. First, the hub is extracted from the image taken by the CCD camera and its diameter data is obtained, which is consistent with the diameter of all the wheel types stored in the database. Compare, if there is the same continuation identification within the allowable range of error, if not, then end the identification process, and draw the conclusion that the wheel hub cannot be identified. On the basis of extracting the image of the hub, analyze the number of spokes in this area, compare with the spoke data in the database that was reduced in the previous step, if there is the same, continue to identify, if not, end the identification process, and conclude that the hub is not identified conclusions. On the premise that the diameter of the hub and the number of spokes are consistent, the number of bolt holes and the shape of the bolt hole area can be obtained from the image in the continuous small circle, and compared with the corresponding data in the reduced-range database, if the number of bolt holes If the numbers are consistent, and the shape of the bolt hole area is also consistent with some wheel types in the database within the allowable range of error, continue to analyze the center hole area, otherwise end the identification process, and draw the conclusion that the wheel hub has not been identified. Finally, the area of the central hole area is consistent with a certain wheel type in the database within the allowable range of error, and the final recognition result of the wheel hub is obtained. The recognition conclusion ends the recognition process.

The specific steps of identification are as follows:

Step 1, collect the visible light image of the hub to be identified, and separate the appearance image of the hub;

Step 2. Find the circumscribed circle and its center of the separated hub, get the diameter data of the hub, and compare it with all the diameter data of the hub stored in the database. If there is the same wheel type, continue to identify, if not, end The identification process leads to the conclusion that this wheel type does not exist in the database;

Step 3. Take the center of the circumscribed circle of the hub as the center, draw a small circle with half the size of the hub as the diameter, divide the hub area into two parts, the inner circle and the ring, and calculate the spokes on the ring area between the circumscribed circle and the small circle The number of spokes, search for the wheel type with the same number of spokes as the current number of spokes in the database range limited in step 2, if there is, continue to identify, if not, end the identification process, and conclude that there is no such wheel type in the database in conclusion;

Step 4. Divide the wheel image in the small circle to obtain the number of bolt holes in the wheel hub. Search for the wheel type with the same number of bolt holes in the database range limited in step 3. If there is, continue to recognize, and if not, end the recognition process, and draw the conclusion that this wheel type does not exist in the database;

Step 5. Segment the wheel image in the small circle to obtain the appearance shape data of the area where the bolt holes of the wheel hub are located. Search for the wheel type with the same appearance shape of the bolt holes within the database range limited in step 4. If there is, continue to identify, if not Then end the identification process and draw the conclusion that there is no such wheel type in the database;

Step 6. Divide the hub image in the small circle to obtain the area of the center hole area of the hub. Search for a wheel type with the same area of the center hole area within the range of the database defined in step 5. If there is, obtain the final model of the hub , if not, draw the conclusion that there is no such wheel type in the database, and end the identification process.

The above identification process can be shortened according to specific needs. If there are fewer types of hubs to be identified on the production line, or certain features are enough to distinguish them, certain parameters can be skipped during identification, such as the area of the center hole area, or the number of spokes in the middle wheel. Wait, speed up the recognition.

Claims (2)

1. A method for automatic identification of wheel hub types based on structural features, characterized in that: the realization steps are: (1), the establishment of the database of the structural characteristic parameters corresponding to the hub; Step 1. Store the characteristic parameters such as the diameter of the hub, the number of spokes, the number of bolt holes, the appearance shape of the area where the bolt holes are located, and the area of the center hole into the database as the identification basis for this type of hub; (2), identify the wheel hub on the production line, and get the conclusion that there is no such wheel type in the model of the wheel hub or in the database: Step 2, collect the visible light image of the hub to be identified, and separate the appearance image of the hub; Step 3. Find the circumscribed circle and its center of the separated hub, obtain the size data of the hub diameter, compare it with the size data of all hub diameters stored in the database, if there is the same wheel type, continue to identify, if not, end the identification process, and draw the conclusion that this wheel type does not exist in the database; Step 4. Take the center of the circumscribed circle of the hub as the center, draw a small circle with half the size of the hub as the diameter, divide the hub area into two parts, the inner circle and the ring, and calculate the spokes on the ring area between the circumscribed circle and the small circle The number of spokes, search for the wheel type with the same number of spokes as the current number of spokes within the database range limited in step 3, if there is, continue to identify, if not, end the identification process, and conclude that there is no such wheel type in the database in conclusion; Step 5. Divide the hub image in the small circle to obtain the number of hub bolt holes, and search for the wheel type with the same number of bolt holes within the database range limited in step 4. If there is, continue to recognize, and if not, end the recognition process, and draw the conclusion that this wheel type does not exist in the database; Step 6. Segment the wheel image in the small circle to obtain the appearance shape data of the area where the bolt holes of the wheel hub are located. Search for the wheel type with the same appearance shape of the bolt holes within the database range limited in step 5. If there is, continue to identify, if not Then end the identification process and draw the conclusion that there is no such wheel type in the database; Step 7. Segment the hub image in the small circle to obtain the area of the center hole area of the hub. Search for a wheel type with the same area of the center hole area within the database range defined in step 6. If there is, obtain the final model of the hub , if not, draw the conclusion that there is no such wheel type in the database, and end the identification process. 2. A method for automatically identifying wheel hub types based on structural features as claimed in claim 1, wherein: the database of structural feature parameters corresponding to the wheel hub is established to obtain the corresponding structural feature parameters by learning the wheel hub to be identified and stored in the database, the specific steps are: Step 1. Collect the visible light image of the hub, and accurately separate the appearance image of the hub from it; Step 2, process the separated hub image, find out the circumcircle of the hub and its center, and obtain the diameter size data of the hub; Step 3. Take the center of the circumscribed circle of the hub as the center, draw a small circle with half the size of the hub as the diameter, and divide the hub area into two parts: the inner circle and the ring; Step 4, counting the number of holes on the hub image on the above-mentioned ring area to obtain the data of the number of spokes; Step 5. Segment the hub image in the small circle to obtain the number of hub bolt holes, the appearance shape of the bolt hole area and the area of the center hole area, and store all the data obtained in the above steps into the database as the final product of this type of hub. Identify parameters.