CN103529053B - Bottle mouth defect detection method - Google Patents

Abstract

The invention discloses a bottle mouth defect detection method, comprising: S1. Dividing the inspection area of the bottle mouth into a first inner ring, a second inner ring, an evaluation ring, a sealing ring and an inner sealing ring area, and Detect each area; S2. Summarize the detection results and output the comprehensive detection results for the bottle mouth. If all the detection items are normal, output the bottle mouth detection is normal, otherwise output the bottle mouth detection abnormality. The bottle mouth defect detection method proposed by the invention has fast detection speed, high precision, stable performance and convenient debugging, and is suitable for long-term real-time uninterrupted detection of bottle mouth defects on a high-speed automatic assembly line.

Description

Bottle mouth defect detection method

technical field

The invention relates to the technical field of industrial automatic detection, in particular to a method for detecting a defect in a bottle mouth.

Background technique

At present, my country's food safety management system is advancing rapidly, and people are paying more and more attention to food safety. The European HACCP standard and my country's GB4927-91 standard both put forward strict requirements for the detection of empty beer bottles. In the existing production process of beer, beverages and pharmaceuticals, the filling containers are required to meet the corresponding quality standards. Every step of production needs to be inspected. When unqualified bottles enter the market, not only consumers may be harmed, but also for the products. The manufacturer's economy and reputation will also be damaged. At present, the detection of empty beer bottles mainly relies on manual detection, but manual detection has the following disadvantages: (1) The detection speed is slow and the efficiency is low. In recent years, the output of glass drinks represented by beer in my country has increased rapidly. Beer output has been ranking first in the world since 2002. The speed of beer production lines in China has also increased rapidly to 240,000 to 40,000 bottles per hour, the fastest abroad. The speed exceeds 70,000 bottles per hour. It is difficult to rely on manual inspection at such a speed; (2) With the rapid increase of labor costs in recent years, manual inspection has become more and more expensive; (3) Human inspection requires a large number of personnel, but the work is boring and intensive. There are fewer and fewer young people who are willing to engage in this work, which makes it difficult for enterprises to recruit workers; (4) Due to the influence of personnel fatigue and emotions, the quality and quality consistency of wine bottles after human inspection are poor, and it is difficult to meet the increasing quality requirements. Therefore, the glass bottle empty bottle inspection machine that uses automatic methods to detect glass bottle empty bottles is being rapidly promoted and used at home and abroad.

There have been some successful cases of foreign glass bottle inspection machines for empty bottles, and their products have a wide customer base in Europe and the United States. However, imported bottle inspection machines have the following problems when used in domestic beer production enterprises: (1) The price is expensive and the supply The delivery cycle is long. my country's annual beer production and number of production lines rank first in the world, but the high price and maintenance costs are not affordable for most companies, and the general pre-supply and after-sales maintenance cycles are long; (2) standard Different, most of the imported bottle inspection machines design algorithms and parameters based on European testing standards, which are inconsistent with my country's national standards, which brings troubles to beer manufacturers; (3) The bottle sources are different, and more than 80% of my country's beer bottles For recycled bottles, 20% are new bottles, while in Europe the ratio is reversed. The production of new bottles in my country is generally produced by multiple glass bottle factories at the same time, and the quality and shape of new bottles are also different. Recycled bottles are mostly used for many times, and the bottle body is seriously damaged, scratched and cracked. In this case, if the imported bottle inspection machine is used, it will lose the flexibility of domestic bottle sources and the quality requirements of various manufacturers, and often cannot meet the needs of domestic beer manufacturers, so that the imported bottle inspection machine will appear "acclimatized". . Therefore, the research and vigorous development of self-owned brand bottle inspection machine equipment with completely independent intellectual property rights is of great significance for improving the control of my country's own core technology and technological innovation capabilities, and has practical value for the improvement of hygiene and safety in my country's food and beverage industry.

As an important part of glass bottle empty bottle detection, bottle mouth inspection is highly valued by beer manufacturers. When the bottle mouth is seriously damaged, it may cause harm to drinkers, and even a small defect will cause the wine to be damaged during transportation due to air leakage. Rapid deterioration during storage, so bottle mouth detection requires high precision. On the other hand, since bottles with defective mouths are generally discarded bottles, most beer manufacturers directly reject and crush bottles that fail the inspection of the mouth of the bottle, and wash back the bottles that fail the inspection of the side wall and bottom of the bottle. After the bottle machine is cleaned again, it returns to the chain channel and then performs empty bottle detection. Therefore, it is required that the false rejection rate of bottle mouth detection must be very low, otherwise a large number of qualified bottles will be mistakenly rejected and crushed, which will cause considerable economic losses for beer production enterprises.

At present, research work on bottle mouth detection has been carried out in some universities. However, due to many practical requirements such as high speed, stability, high precision, strong adaptability, and easy debugging by engineers, the bottle inspection machine needs to weigh the above factors, especially speed and stability, in the algorithm design process of glass bottle mouth detection. In order to meet the requirements of the field, develop a glass bottle empty bottle mouth detection algorithm suitable for industrialization.

Contents of the invention

(1) Technical problems to be solved

The technical problem to be solved by the present invention is: how to provide a high-speed and high-precision bottle mouth defect detection method, which is used to solve the problems of slow speed, low precision and problems of the existing high-speed glass bottle filling automatic production line empty bottle bottle mouth defect detection method. The problem of poor stability.

(2) Technical solution

In order to solve the above problems, the present invention provides a method for detecting defects in the bottle mouth, including: S1. Divide the detection area of the bottle mouth into a first inner ring, a second inner ring, an evaluation ring, a sealing ring and an inner sealing ring Ring area, and detect each area; S2. Summarize the test results and output the comprehensive test results for the bottle mouth. If the results of all the test items are normal, then output the bottle mouth detection is normal, otherwise output the bottle mouth detection abnormal.

Preferably, in the step S1, a parallel computing mechanism is used to simultaneously detect the first inner ring, the second inner ring, the evaluation ring, the sealing ring and the inner sealing ring area, and each inspection is completed in parallel.

Preferably, detecting the first inner ring area includes: S111, performing a binarization operation on the first inner ring area after filtering, the normal area is black, and the abnormal area is white; S112, the white area Carry out connected domain analysis, when the maximum connected domain area is greater than a given threshold, return the first inner ring detection abnormality, otherwise return the first inner ring detection normal.

Preferably, detecting the evaluation ring area includes: S131, performing binarization after filtering the evaluation ring area, making the normal area black and the abnormal area white, and then performing polar coordinate expansion; S132, along the Observe the number of white pixels in each column of pixels in the radial direction of the evaluation ring. If the number of white pixels in a certain column exceeds a given range, mark the column as unqualified; S133. If a certain column is detected to be unqualified, the error value increases. On the contrary, if it is qualified, the error value will decrease. If the peak value of the error value exceeds the given threshold, the output is unqualified. If the output is unqualified, the evaluation loop will return to detect abnormality, otherwise the evaluation loop will be returned to detect normal.

Preferably, detecting the seal ring area includes: S141, filtering the seal ring area and performing binarization processing, making the normal area black and the abnormal area white, and then performing polar coordinate expansion; S142, along the seal ring Observe the number of white pixels in each column of pixels in the radial direction, if the number of white pixels in a certain column is lower than a given value, then mark this column as unqualified; S143, if a certain column is detected as unqualified, the error value increases, otherwise If it is qualified, the error value will decrease. If the peak value of the error value exceeds the given threshold, the output will be unqualified. If the output is unqualified, it will return the seal ring detection exception, otherwise it will return the seal ring detection normal.

Preferably, detecting the inner sealing ring area includes: S151. After filtering the inner sealing ring area, segment the image of the area using a grayscale threshold, set the grayscale to black for the area smaller than the threshold value, and use it as a background area. The grayscale of the threshold area remains unchanged, and is used as the foreground area; S152, after averaging the foreground area and adding a gray scale offset, it is used as a threshold, and the foreground area is binarized. Set the white area to white; S153, open the white area to remove the influence of small stripes and small false areas; S154, calculate the connected domain information of the white area, and set the upper and lower limits of the connected domain area, width and height and other detection items , when there is at least one connected domain and all the information is within the limited range at the same time, it will return the inner sealing ring exception, otherwise it will return the inner sealing ring normal.

(3) Beneficial effects

The bottle mouth defect detection method proposed by the invention has fast detection speed, high precision, stable performance and convenient debugging, and is suitable for long-term real-time uninterrupted detection of bottle mouth defects on a high-speed automatic assembly line.

Description of drawings

Fig. 1 is a schematic flow chart of a method for detecting defects at a bottle finish according to an embodiment of the present invention;

Fig. 2 is a schematic diagram of a bottle mouth detection image according to an embodiment of the present invention;

Fig. 3 is a schematic diagram of an image acquisition system for bottle finish detection according to an embodiment of the present invention.

Detailed ways

The specific implementation manners of the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

The invention provides a method for detecting a bottle mouth defect, comprising:

S1. Divide the detection area of the bottle mouth, and use the parallel computer mechanism to simultaneously detect the first inner ring, the second inner ring, the evaluation ring, the sealing ring and the inner sealing ring;

S2. Summarize the test results, and finally output the comprehensive test result for the bottle mouth. If there is no less than one test item with abnormal output, the test result is abnormal, otherwise the test result is normal.

Wherein, in the step S1, the first inner ring, the second inner ring, the evaluation ring, the sealing ring and the inner sealing ring etc. are tested at the same time by adopting a parallel computer mechanism, and each detection item is completed in parallel, and each item is not detected during the detection process. Coupling relationship, the specific process is:

S11, the first inner ring detection, first filter and binarize the area, the normal area is black, the abnormal area is white, and then the connected domain analysis is performed on the white area, if the maximum area exceeds the threshold, return to the first inner ring detection Abnormal, otherwise return the first inner loop detection is normal;

S12. The second inner ring detection, the detection process of which is the same as that of the first inner ring detection;

S13, evaluation ring detection, first perform region filtering and binarization operations, normal regions are black, abnormal regions are white, and then the number of white pixels in each column of pixels is counted along the radial direction of the evaluation ring, when a column of white pixels When the number of pixels is outside the normal range, mark the column as unqualified. If the weighted unqualified column exceeds a given threshold, return the evaluation ring to detect abnormality, otherwise return to the evaluation ring to detect normal;

S14. Sealing ring detection, first perform region filtering and binarization operations, the normal region is black, and the abnormal region is white, and then the number of white pixels in each column of pixels is counted along the radial direction of the sealing ring, if a column of white pixels If the number is less than the given value, mark the column as unqualified. If the weighted unqualified column exceeds the given threshold, return the seal ring detection exception, otherwise return the seal ring detection normal;

S15 , the detection of the inner sealing ring is first divided into a foreground area with a constant gray value and a black background area. Then carry out binarization processing, the ones smaller than the threshold are set as black, that is, the background area, and the ones larger than the threshold are set as white, the third is to open the white area, and the fourth is to calculate the connected domain information of the white area, and set the connected domain area, For the upper and lower limits of detection items such as width and height, if there is at least one connected domain and all information is within the limit at the same time, it will return the inner seal ring as abnormal, otherwise it will return the inner seal ring as normal.

The process flow of the bottle mouth defect detection method proposed in this embodiment is shown in FIG. 1 . The captured images and the marks of each detection area are shown in Figure 2. Wherein the outer circular bright ring is the sealing ring, the inner circular bright ring is the inner sealing ring, the first inner ring is between the sealing ring and the inner sealing ring, the second inner ring is inside the inner sealing ring, and the sealing ring extends outward and outward for a certain amount. The area after the width is the evaluation ring. The acquisition system of the bottle mouth image is shown in Figure 3, in which the ring-shaped LED light source 2 obliquely illuminates the bottle mouth in the bottle body 3, and the light enters the CCD camera 1 after being reflected by the surface of the bottle mouth. When the surface of the bottle mouth is smooth and complete, the image is in the shape of a complete and standardized bright ring, while when the surface of the bottle mouth is defective, the bright ring of the image appears to be disconnected or accompanied by large white spots.

The main steps of detection include:

S11. First inner loop detection. When there is a large gap or grinding in the mouth of the bottle, a large area of white spots will appear in the first inner ring area, and the detection of the first inner ring area is mainly for the detection of the above defects. Its detection method is:

S111. Perform a binarization operation after filtering the first inner ring area, so that the normal area is black and the abnormal area is white;

S112. Perform connected domain analysis on the white area, and when the largest connected domain area is greater than a given threshold, return the first inner ring detection abnormality, otherwise return the first inner ring detection normal;

S12. The second inner ring detection. When there is a large foreign matter blocking the bottle mouth, a large area of white spots may appear in the second inner ring area. The detection of the second inner ring area is mainly to detect the above-mentioned defects. The detection method is the same as that of the first inner ring detection, and will not be repeated here.

S13. Evaluation ring detection. When there is a defect in the bottle mouth, the sealing ring will be broken. If there is a grinding mouth, there will be a large white spot around the sealing ring. The detection of the evaluation ring area is mainly aimed at the detection of the above-mentioned defects. Its detection method is:

S131. Perform binarization after filtering the evaluation ring area, and then perform polar coordinate expansion;

S132. Observe the number of white pixels in each row of pixels along the radial direction of the evaluation ring. If the number of white pixels in a row exceeds a given range, mark the row as unqualified;

S133. If it is detected that a certain column is unqualified, the error value is increased; otherwise, if it is qualified, the error value is decreased. If the error value peak exceeds the given threshold, the output fails.

If the output is unqualified, return the evaluation loop detection exception, otherwise return the evaluation loop detection normal.

S14. Sealing ring detection. When there is a defect in the bottle mouth, the sealing ring will be broken. The sealing ring detection is to detect this kind of defect, and it is also the most important item among the various detection items. The detection method is:

S141. Perform binarization processing after filtering the sealing ring area, and then perform polar coordinate expansion;

S142. Observing the number of white pixels in each row of pixels along the radial direction of the sealing ring, if the number of white pixels in a row is lower than a given value, mark the row as unqualified. ;

S143. If a column is detected to be unqualified, the error value is increased; otherwise, if it is detected to be qualified, the error value is decreased; if the peak value of the error value exceeds a given threshold, the output is unqualified.

If the output is unqualified, it will return that the seal ring detection is abnormal, otherwise it will return that the seal ring detection is normal.

S15. Inner sealing ring detection. When the inner edge of the bottle mouth is damaged, there will be white spots on the inner sealing ring, but due to the different shapes of the inner edge of the bottle mouth, the imaging of the inner sealing ring is different, and its brightness, width and texture are different, which increases the detection. Difficulty, the detection method of the inner sealing ring is:

S151. After filtering the area of the inner sealing ring, use the gray threshold to segment the image of the area, set the gray level to black in the area smaller than the threshold, and use it as the background area; the gray level in the area greater than the threshold remains unchanged, and use it as the foreground area;

S152. Taking the average value of the foreground area and adding the gray scale offset as the threshold value, and performing binarization processing on the foreground area, setting the foreground area as black, and setting the value greater than the threshold as white;

S153. Perform an open operation on the white area to remove the influence of small stripes and small false areas;

S154. Calculate the connected domain information of the white area, set the upper and lower limits of the connected domain area, width and height and other detection items. If there is at least one connected domain and all the information is within the limited range at the same time, return the inner seal ring exception, otherwise return the inner seal ring normal.

S2. Summary of test results. After the detection results of each detection item are synthesized, the final result is output. If the results of all positioning items and detection items are normal, the output of the bottle mouth detection is normal, otherwise, the output of the bottle mouth detection is abnormal.

In summary, the present invention has the following advantages:

1. Fast detection speed. At the algorithm level, each detection item is completely independent, so multi-thread parallel computing can be used in programming. When the collected image is 640×480 pixels, and the calculation is performed on the I5CPU computer, the calculation time of each image is within 14ms, which can meet the needs of the fastest beer production line in the world.

2. High detection accuracy and low false detection rate. The method can accurately detect the 5mm ² defect of the bottle mouth. At the same time, it can effectively resist interference, and the false detection rate is not higher than 0.1%.

3. Stable performance. The detection program generated by this algorithm has small memory usage and throughput during operation, and will not cause operation errors when the parameter settings are unreasonable or non-bottleneck image input. Therefore, it is difficult to cause abnormal operation such as crash and memory exhaustion during long-term operation.

4. It is convenient for engineers to debug, and the requirements for engineers are relatively low. The detection results are not sensitive to parameters other than the accuracy threshold; at the same time, the parameters to be set are intuitive and easy to understand, and the parameter setting can be adjusted by simply analyzing the image. Therefore, engineers with low theoretical level and experience can also be competent for on-site debugging.

The above is only a preferred embodiment of the present invention, it should be pointed out that for those of ordinary skill in the art, without departing from the technical principle of the present invention, some improvements and replacements can also be made, these improvements and replacements It should also be regarded as the protection scope of the present invention.

Claims (5)

1. A bottleneck detection method, characterized in that, comprising: S1. Divide the detection area of the bottle mouth into the first inner ring, the second inner ring, the evaluation ring, the sealing ring and the inner sealing ring area, and detect each area; Detecting the area of the inner sealing ring includes: S151. After filtering the area of the inner sealing ring, use the gray threshold to segment the image of the area, set the gray level to black in the area smaller than the threshold, and use it as the background area, and keep the gray level in the area greater than the threshold, and use it as the foreground area; S152. Taking the average value of the foreground area and adding the gray scale offset as the threshold value, and performing binarization processing on the foreground area, setting the foreground area as black, and setting the value greater than the threshold as white; S153. Perform an open operation on the white area to remove the influence of small stripes and small false areas; S154. Calculate the connected domain information of the white area, set the upper and lower limits of the connected domain area, width and height and other detection items. If there is at least one connected domain and all the information is within the limited range at the same time, return the inner seal ring exception, otherwise return the inner seal ring normal; S2. Summarize the test results and output the comprehensive test results for the bottle mouth. If all the test items are normal, then output that the bottle mouth is detected as normal, otherwise output that the bottle mouth is detected as abnormal. 2. The method according to claim 1, characterized in that, in the step S1, a parallel computing mechanism is used to simultaneously perform an operation on the first inner ring, the second inner ring, the evaluation ring, the sealing ring and the inner sealing ring area All tests are performed in parallel. 3. The method according to claim 1 or 2, wherein detecting the first inner ring area comprises: S111. Perform a binarization operation on the first inner ring area after filtering, the normal area is black, and the abnormal area is white; S112. Perform connected domain analysis on the white area, and when the largest connected domain area is greater than a given threshold, return the first inner ring detection abnormality, otherwise return the first inner ring detection normal. 4. The method according to claim 1 or 2, wherein detecting the evaluation ring region comprises: S131. Perform binarization after filtering the area of the evaluation ring, make the normal area black and the abnormal area white, and then perform polar coordinate expansion; S132. Observing the number of white pixels in each row of pixels along the radial direction of the evaluation ring, if the number of white pixels in a row exceeds a given range, mark the row as unqualified; S133. If it is detected that a certain column is unqualified, the error value will increase; otherwise, if it is qualified, the error value will decrease; if the peak value of the error value exceeds a given threshold, the output will be unqualified; if the output is unqualified, return to the evaluation loop to detect abnormality, Otherwise returns the evaluation loop to detect OK. 5. The method according to claim 1 or 2, wherein detecting the sealing ring area comprises: S141. Perform binarization processing after filtering the sealing ring area. The normal area is white, and the abnormal area is black, and then polar coordinate expansion is performed; S142. Observing the number of white pixels in each row of pixels along the radial direction of the sealing ring, if the number of white pixels in a row is lower than a given value, mark the row as unqualified; S143. If it is detected that a certain column is unqualified, the error value will increase, otherwise, if it is qualified, the error value will decrease. If the peak value of the error value exceeds a given threshold, the output will be unqualified. If the output is unqualified, return the seal ring detection exception. Otherwise return to the seal ring detection is normal.