A foreign object detection method for the bottom of a packaging hose

Abstract

The present invention discloses a method for detecting foreign objects at the bottom of a packaging hose, belonging to the field of detection technology, and comprising the following steps: accurately segmenting the area at the bottom of the hose; automatically positioning the region of interest of the image to be detected; and identifying foreign objects by means of Fourier transform and inverse Fourier transform. The present invention adopts an annular and circular area segmentation method, a dynamic positioning method for the region of interest, and a Fourier foreign object recognition algorithm to achieve accurate and rapid detection; to solve the deficiencies existing in the existing detection methods, ensure that each hose undergoes accurate foreign object detection before filling, reduce false detections and missed detections, and avoid to the greatest extent possible the products with foreign objects from flowing into the market. While ensuring the product quality and protecting the enterprise and its brand image, this method also improves the stability of the detection system and the production efficiency of the production line, reduces losses to a certain extent, lowers the production cost, and enhances the competitiveness of the enterprise, and has a wide range of application spaces and market values.

Description

Technical Field

[0001] The present invention belongs to the technical field of detection, and particularly relates to a method for detecting foreign objects at the bottom of a packaging hose. Background Art

[0002] Packaging hoses are produced by professional packaging enterprises and then transported to customer enterprises. During the production, transportation, and filling of packaging hoses, it is inevitable that foreign objects will fall into the tubes. To reduce the possibility of foreign objects in the packaging hoses, dust collectors are generally installed on filling machines to suck away foreign objects entering the tubes. However, due to reasons such as insufficient air pressure, blocked air pipes, and other mechanical failures, the foreign objects cannot be sucked away, which will affect the product quality to a lesser extent and may cause major health and safety liability accidents in severe cases, not only damaging the interests of consumers but also affecting the reputation of the enterprise.

[0003] Before filling paste into hose-packaged products, in order to ensure product quality, a foreign object detection device needs to be installed on the filling machine to detect whether there are foreign objects in the tube. With the continuous maturity of machine vision technology, foreign object detection based on machine vision technology has been recognized by many manufacturers. First, in a foreign object detection device based on machine vision technology, it is necessary to segment the detection area at the bottom of the hose. Currently, the commonly used method is to divide the bottom into several square areas. However, due to the annular distribution of the bottom brightness, this area segmentation method has a high false detection rate. Second, in actual detection, due to the installation position of the tube cup and the position difference of the hose in the tube cup, the image positions will be inconsistent, and the fixed detection area cannot meet the requirements. At the same time, in the detection process of hose foreign objects, each hose is a new sample, that is, there is a certain randomness, and general image detection algorithms are difficult to meet the detection requirements. Therefore, there is an urgent need for a new foreign object detection method that can accurately and quickly achieve area segmentation, area positioning, and foreign object recognition. Summary of the Invention

[0004] In view of the above technical problems existing in the prior art, the present invention proposes a method for detecting foreign objects at the bottom of a packaging hose, with reasonable design, overcoming the deficiencies of the prior art and having good effects.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A method for detecting foreign objects at the bottom of a packaging hose, comprising the following steps:

[0007] Step 1: Adopt an annular and circular area segmentation method to accurately segment the area at the bottom of the hose;

[0008] Step 2: Adopt an interest area dynamic positioning method to automatically position the interest area of the image to be detected;

[0009] Step 3: Use Fourier transform and inverse Fourier transform for foreign object recognition.

[0010] Preferably, in Step 1, the bottom of the hose is segmented according to brightness, and areas with the same and uniform brightness are divided into the same detection area. In the foreign object detection of the hose, a square region of interest is used, which can span multiple gray value regions.

[0011] Preferably, in Step 2, through the bright and dark features in the image, the center of the bottom of the hose is found twice, and the difference between the coordinates of the two centers is calculated; if the deviation is within the threshold range, it is considered that the center coordinates are found, and the positions of all the regions of interest at the bottom change dynamically with the change of the center coordinates to achieve dynamic tracking; if the difference between the coordinates of the centers found twice is large, it is directly determined that there is a foreign object at the bottom of the hose, and there is no need to perform the next foreign object detection.

[0012] Preferably, in Step 3, first use Fourier transform to convert the time-domain image into a frequency-domain image, then perform filtering and inverse Fourier transform, and finally detect foreign objects.

[0013] The beneficial technical effects brought by the present invention:

[0014] The present invention innovatively adopts an annular and circular region segmentation method, a dynamic positioning method for regions of interest, and a Fourier foreign object recognition algorithm to achieve accurate and rapid detection; to solve the deficiencies of existing detection methods, ensure that each hose undergoes accurate foreign object detection before filling, reduce false detections and missed detections, and avoid products with foreign objects from entering the market to the greatest extent. This method improves the stability of the detection system and the production efficiency of the production line while ensuring product quality and protecting the enterprise and its brand image. To a certain extent, it also reduces losses, lowers production costs, and improves the competitiveness of the enterprise, and has broad application space and market value. Description of the Drawings

[0015] Figure 1 It is an internal view of an empty tube of a packaging hose without foreign objects.

[0016] Figure 2 It is a segmentation diagram of the region of interest inside the hose.

[0017] Figure 3 It is a flow chart of time-domain and frequency-domain conversion. Detailed Embodiments

[0018] The present invention will be further described in detail below in conjunction with the drawings and specific embodiments:

[0019] Figure 1 It is an internal view of an empty tube of a packaging hose without foreign objects. From Figure 1As can be seen, the overall image brightness inside the hose is not uniform or uneven. If only one region of interest (ROI) covering the entire bottom is framed in the figure for detection, all the parameters in this region are the same, and the false detection and missed detection rates will be very high at this time. If the detection region is simply divided into multiple regular square regions of interest and parameters are set for each of the subdivided regions of interest, this method will reduce the false detection rate and missed detection rate to a certain extent. However, due to the large number of parameters that need to be set, not only is the workload large, but the false detection rate will also be very high if the position of the hose fluctuates slightly. From Figure 1 a feature is found: in the figure, it can be divided into several circles or rings according to brightness, and the brightness in the same circle or ring region is almost the same and relatively uniform. If the same ring or circle is set as a region of interest, the false detection will be very small.

[0020] Figure 2 is the segmentation diagram of the region of interest inside the hose. In Figure 2 , the circles or rings with the same brightness are set as a region of interest, and the center coordinates are found within circle 2 (the thick-line circle) based on the gray values of the inner and outer contours of ring 1. The positions of all detection regions change dynamically with the center of the circle, achieving precise positioning of the detection regions. For example: the innermost circle in the figure is used as a region of interest, the ring between the innermost circle and the second circle is used as a region of interest, the outermost circle and the second circle are used as a region of interest, and so on.

[0021] In actual detection, due to the installation position of the tube cup and the position difference of the hose in the tube cup, the image positions will be inconsistent, affecting the final detection effect. To solve the above problems, an automatic positioning method is adopted to accurately locate the detection region and eliminate the influence of the installation position of the tube cup and the hose position difference on the detection result. In Figure 1 it can be seen that under normal circumstances, the circles or rings with different brightnesses inside the hose are concentric. If the center of the circle is found and used as a reference point, all regions of interest move with the center of the circle, and they maintain a constant position relative to the center of the circle. Even if the position of the hose changes, the absolute position of the region of interest inside the hose will not change, which greatly reduces the false detection caused by the change in the hose position. In Figure 1 , the gray value of the gray circle at position 1 is 100, the gray value of the gray circle at position 2 is about 90, and the gray value of the bright circle at position 2 is 220. When performing positioning, a circle is drawn at position 1 with a gray value of 100 and the center coordinates are calculated. A circle is drawn at position 2 with a gray value of 220 and the center coordinates are calculated. If the two centers of the circles coincide or are close, it is considered that the center of the circle is found, and all regions of interest are positioned based on this center of the circle. If the difference between the two center coordinates is large, it is considered that the search for the center of the circle fails due to the presence of foreign objects.

[0022] Figure 3 is the time-domain and frequency-domain conversion flow chart.Figure 3 First, the time-domain image is converted into a frequency-domain image, then filtered and inverse Fourier-transformed, and finally foreign objects are detected.

[0023] For a periodic signal f(t) with a period of T 1 , its fundamental angular frequency is ω 1 = 2π / T 1 . Under the condition of satisfying the Dirichlet conditions, the signal can be expanded as:

[0024]

[0025] The above formula is called the Fourier series in trigonometric form, where:

[0026]

[0027]

[0028]

[0029] An image is equivalent to a two-dimensional signal. For an image, the frequency represents the degree of change in the information contained in the pixels of the image. For an image of a white house, there is not too much intense change in the color distribution among its pixels, and the corresponding frequency is relatively low at this time. For an image of a square, since it contains more content, the corresponding frequency is relatively high. For the Fourier transform in an image, its essence is to convert the gray-level distribution function of the image into the gray-level gradient distribution function of the image, and its inverse transform is to convert the gray-level gradient distribution function of the image into the gray-level distribution function. For an image with M*N pixels, its discrete Fourier transform is obtained by the following equation:

[0030]

[0031] where m = 0, 1, 2 ……, m - 1 and v = 0, 1, 2 ……, n - 1. The variables u and v are used to determine their frequencies, and it can be obtained that the spectral components of the spectral system at the four corners (0, 0), (0, N - 1), (N - 1, 0), (N - 1, N - 1) of the spectrogram are all 0 along the u and v directions.

[0032] The discrete inverse Fourier transform is given by the following formula:

[0033]

[0034] The frequency spectrum diagram obtained by Fourier transform characterizes the gray-scale changes in the time-domain image. Generally, the high-frequency part in the frequency-domain diagram represents the parts with obvious mutations in the image, which may represent noise in some cases. The low-frequency part, on the other hand, represents the parts with gentle changes in the image, that is, the contour information of the image. In addition, when there are periodic interference signals in the image, some filters need to be designed to filter out the influence of these noise interferences. For the frequency spectrum diagram, if there are fewer bright spots and more dark spots on the diagram, it means that the distribution of the time-domain image is relatively uniform. On the contrary, if there are more bright spots in the frequency-domain image, then the distribution of the time-domain image is relatively sharp, with obvious boundaries and a large difference in the gray-scale distribution at both ends of the boundary.

[0035] The present invention relates to a method for detecting foreign objects at the bottom of a packaging hose. By innovatively using the annular and circular area segmentation method, the area dynamic positioning method, and the Fourier foreign object recognition algorithm, accurate and rapid detection is achieved, ensuring that each packaging hose undergoes accurate foreign object screening and detection. This greatly reduces the possibility of hose products containing foreign objects flowing into the market. While improving product quality and production efficiency, it also reduces the production costs of enterprises and enhances the competitiveness of enterprises, having certain application value and application space.

[0036] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions, or substitutions made by those skilled in the art within the essence of the present invention should also fall within the protection scope of the present invention.

Claims

1. A method for detecting foreign objects at the bottom of a packaging hose, characterized in that: It includes the following steps: Step 1: Use the annular and circular area segmentation method to accurately segment the area at the bottom of the hose; Step 2: Use the dynamic positioning method of the region of interest to automatically locate the region of interest of the image to be detected; Through the bright and dark features in the image, find the center of the circle at the bottom of the hose twice, and calculate the difference between the coordinates of the two centers of the circle; if the deviation is within the threshold range, it is considered that the center coordinates are found, and the positions of all the regions of interest at the bottom change dynamically with the change of the center coordinates to achieve dynamic tracking; if the difference between the coordinates of the two centers of the circle found twice is large, it is directly judged that there are foreign objects at the bottom of the hose, and there is no need to perform the next step of foreign object detection; Step 3: Use Fourier transform and inverse Fourier transform for foreign object recognition.

2. The method for detecting foreign objects at the bottom of a packaging hose according to claim 1, characterized in that: In step 1, the bottom of the hose is segmented according to brightness, and the areas with the same and uniform brightness are divided into the same detection area. In the detection of foreign objects in the hose, a square region of interest is used, which can span multiple gray value regions.

3. The method for detecting foreign objects at the bottom of a packaging hose according to claim 1, characterized in that: In step 3, first use Fourier transform to convert the time-domain image into a frequency-domain image, then perform filtering and inverse Fourier transform, and finally detect foreign objects.

Images