Potato browning detection method based on machine vision

Abstract

The present invention discloses a potato browning detection method based on machine vision, which belongs to the technical field of non-destructive testing of food and agricultural products, and can be applied to various potato processing industries that need to peel potatoes. This method uses a computer vision system to collect the characteristic conditions of the potato surface after peeling, simulates the recognition method of the human eye, and establishes the correlation between the color characteristics of the potato surface and the browning of the potato. In different production links, for different objects, the color information of the object being tested can be detected, and the browning classification determined by artificial senses or other production standards can be used to detect the browning of potatoes in real time. The method described in the present invention focuses on the practical problems existing in the potato processing industry, and based on advanced and efficient rapid non-destructive testing technology, proposes a new solution to the problem of browning detection of peeled potatoes in the processing link, which plays a positive role in improving the quality of potato processing and the level of intelligence.

Description

Technical Field

[0001] The invention relates to a potato browning detection method based on machine vision, mainly targeting the potato browning problem after peeling and cutting operations in the potato industry, and belongs to the technical field of non-destructive detection of agricultural products. Background Art

[0002] Potato is the world's fourth largest food crop, second only to rice, wheat and corn. Since 2015, my country has gradually promoted the development strategy of potato as a staple food, encouraging the potato processing industry to carry out technological innovation and product upgrading. As people's demand for potatoes and their processed products continues to increase, product quality is becoming more and more important, and defect-free potato products have received widespread attention in the market. In the processing link of the potato industry, the browning problem has always been the main constraint on new product development and process optimization. Therefore, how to weaken or even eliminate the impact of browning on potato products is a key technical problem in the development of the industry.

[0003] Potato browning is widely found in the peeling and cutting process of the fresh-cut potato industry and other industries. Physical processing methods such as sound, light, electricity, and heat, and chemical treatment methods such as nitrite and citric acid are mainly used to address the browning problem at home and abroad. Regardless of the treatment method used, the purpose is to inhibit browning and improve product quality. However, there are differences in the browning rate of untreated potatoes themselves, and treated potatoes will also undergo a certain degree of browning. These unknown factors are obviously not conducive to quality management. Therefore, real-time monitoring of potato browning is necessary.

[0004] Existing potato browning detection methods include colorimeter, UV spectrophotometer, high performance liquid chromatography, etc. These traditional methods are cumbersome, expensive, time-consuming, and are mostly used in laboratories. On-site industrial inspections mostly use manual visual inspections, which to a certain extent solve the production control problem. However, labor is bound to become more and more precious in industrial development, and replacing manpower with more advanced and intelligent technologies is an inevitable trend in the development of the industry.

[0005] In recent years, machine vision technology has been widely used in various industries. Using a camera to capture images in real time, and then performing qualitative or quantitative analysis on the images to evaluate the object being tested is a fast and non-destructive detection method. At present, machine vision technology has been applied to potato detection, but most of them are aimed at detecting individual size, defects, etc., and have not yet been applied to browning detection. This patent application applies machine vision technology to potato browning detection, which has the advantages of convenience, low cost, and real-time collection and analysis, which helps to improve the quality of potato processing products. Summary of the invention

[0006] In order to overcome the shortcomings of the prior art, the present invention discloses a potato browning detection method based on machine vision, which can be widely used in the potato processing industry. The method uses an image acquisition device to obtain a potato image, and then uses an image processing program to analyze the obtained image to obtain the specific degree of potato browning. It is fast, non-destructive, simple to operate, and does not require other expensive professional instruments or reagents.

[0007] The present invention provides a method for detecting potato browning based on machine vision, which is carried out according to the following steps:

[0008] (1) Build a machine vision inspection device and prepare experimental samples.

[0009] (2) The potato sample image is collected in a machine vision detection device, and the image is captured by a CCD (charge coupled device image sensor) camera and saved in a computer.

[0010] (3) Preprocess the image, convert the RGB (red, green and blue) image into a grayscale image, and use a filtering algorithm to remove noise.

[0011] (4) Background segmentation: Otsu’s threshold segmentation method is used to obtain the target area. A small amount of noise interference and target missing are further processed using erosion and dilation operators.

[0012] (5) Feature extraction: The extracted target image is converted into three color spaces for expression, namely RGB, HSV (hue, saturation, brightness), and Lab (brightness, red-green, yellow-blue), and they are segmented into single-channel images. The grayscale mean and standard deviation of each channel are extracted from the nine single-channel images in the above three color spaces, and the colors are analyzed from different angles to form image information describing a sample.

[0013] (6) Establish a browning degree discrimination model and perform qualitative or quantitative analysis on each sample.

[0014] Furthermore, the device for machine vision inspection in step (1) is an integrated device and a detachable device. The device is designed and assembled according to the specific inspection environment and inspection object, and can be disassembled and replaced at any time to meet image acquisition requirements, such as in different experimental sites, production lines, or for inspecting multiple objects at a time.

[0015] Furthermore, the machine vision detection integrated device described in step (1) includes a hardware device and a software system, wherein the hardware device includes a housing of the machine vision detection integrated device, a ring light, a CCD camera, and a computer, and the software part consists of an electrical control module and an image processing module. A CCD camera is installed on the top of the housing of the machine vision detection integrated device through a bayonet, and the CCD camera is connected to the computer through a data cable, and a ring light is also installed on the housing wall of the machine vision detection integrated device.

[0016] Furthermore, the sample described in step (1) is a fresh potato, the variety is Dutch potato, and the flesh is yellow. When selecting samples, try to select samples with good quality and the same appearance characteristics (size, color and shape). The purchased samples are stored in a cool and dry environment, and the room temperature is controlled at 13°C (±2) for one week. Samples with no obvious changes in appearance are selected for the experiment. Before the experiment begins, the selected potato samples are cleaned with tap water and the skin is peeled. The cleaned potatoes are cut into uniform 3 mm thick slices, and the equipment used is a mechanical slicer (NEMCO-55200AN, USA). Then use a uniform template to cut the slices into squares with a length and width of 3.5 cm, and put the prepared samples into a refrigerator at 4°C for standby use. Potato samples are tested at 0, 2, 4, 6, and 8 days, and 15 slices are randomly selected from the spare samples each time.

[0017] Furthermore, when collecting visual information of the sample in step (2), the form, intensity, angle of illumination, height and focal length of the camera are all adjustable to ensure comprehensive and clear acquisition of color and texture information on the potato surface.

[0018] Furthermore, the filtering algorithm in step (3) can be adjusted according to different images obtained under different test conditions, and median filtering is preferred.

[0019] Furthermore, the background segmentation in step (4) is preferably performed under a black background. If the background is more complex, it is processed accurately according to the specific situation.

[0020] Furthermore, the 18 eigenvalues ​​in step (5) are part of the sample color information, and other eigenvalues ​​such as maximum value, minimum value, etc. can be added as needed to supplement the sample information to increase accuracy.

[0021] Furthermore, the browning degree discrimination model in step (6) uses the browning degrees of potatoes at 0, 2, 4, 6, and 8 days prepared by us as 5 levels and models them with the machine vision detection results, so as to perform qualitative discrimination on other samples with the trained model.

[0022] Furthermore, if a more accurate browning degree is required for the browning degree discrimination model in step (6), the samples are scored manually or other instrumental methods are used to obtain the browning score, and then the scoring results are regressed with the machine vision detection results to obtain a browning quantitative scoring model.

[0023] Compared with the prior art, the potato browning detection method designed by the present invention replaces the human eye with a simulated visual sense, has higher accuracy than the traditional technology, is non-contact and non-destructive to the sample, and can achieve rapid detection. If the present invention is applied to potato browning detection, production monitoring and product grading can be better carried out, the automation and intelligence level of the potato industry can be increased, productivity and product quality can be improved, and economic benefits can be expanded. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 The device diagram of the potato browning detection method based on machine vision, wherein 1 is the sample to be detected, 2 is the housing of the machine vision detection integrated device, 3 is the ring light, 4 is the CCD camera, and 5 is the computer.

[0025] Figure 2 The graphs are of experimental samples, showing the browning degree of fresh-cut potatoes at 0, 2, 4, 6 and 8 days, with 15 potatoes per day.

[0026] Figure 3 Qualitative analysis model results of machine vision for 5 levels of experimental samples, including A-LDA (Linear Discriminant Analysis) model classification results and B-SVM (Support Vector Machine) model classification results.

[0027] Figure 4 The results of the quantitative analysis model for expert scoring of 5 levels of experimental samples for machine vision, where A is the training set result and B is the prediction set result. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] See also Figure 1, an embodiment of the present invention. A machine vision detection integrated device includes a hardware device and a software system, wherein the hardware device includes 1-a sample to be detected, 2-a housing of the machine vision detection integrated device, 3-a ring light, 4-a CCD camera, and 5-a computer. The software part consists of an electrical control module and an image processing module. A 4-CCD camera is installed on the top of the housing of the machine vision detection integrated device through a bayonet, and the 4-CCD camera is connected to the 5-computer through a data cable, and a 3-ring light is also installed on the housing wall of the machine vision detection integrated device.

[0030] The present invention is universal for potato browning detection. Potatoes have various shapes during processing. The machine vision detection integrated device in the present invention lists the specific structure and size of fresh-cut potato slices as the object. The specific implementation steps are described as follows:

[0031] (1) Preparation of experimental samples. The samples were fresh potatoes of the Dutch variety with yellow flesh. When selecting samples, try to select samples with good quality and the same appearance characteristics (size, color and shape). The purchased samples were stored in a cool and dry environment with the room temperature controlled at 13°C (±2) for one week. Samples with no obvious changes in appearance were selected for the experiment. Before the experiment began, the selected potato samples were cleaned with tap water and the skin was peeled. The cleaned potatoes were cut into uniform 3 mm thick slices using a mechanical slicer (NEMCO-55200AN, USA). The slices were then cut into squares with a length and width of 3.5 cm using a uniform template. The prepared samples were placed in a refrigerator at 4°C for later use. Potato samples stored on days 0, 2, 4, 6, and 8 were selected for testing, and 15 slices were randomly selected from the spare samples each time.

[0032] (2) The sample is placed in an integrated machine vision inspection device, and the image is captured by a CCD camera and saved in a computer.

[0033] (3) The sample image obtained in step (2) is preprocessed, background segmented and feature value extracted. First, the image is preprocessed, the RGB image is converted into a grayscale image, and the noise is removed using a filtering algorithm. Then the background is segmented, and the target area is obtained using the Otsu threshold segmentation method. A small amount of noise interference and target loss are further accurately processed using corrosion and dilation operators. Then the feature value is extracted, and the extracted target image is converted into RGB, HSV, and Lab color spaces for expression, and they are segmented into single-channel images. The grayscale mean and standard deviation of each channel are extracted from the above three color spaces and 9 single-channel images, and the colors are analyzed from different angles to form image information describing a sample.

[0034] (4) Qualitative judgment of potato browning degree. The browning degree of potatoes at 0, 2, 4, 6, and 8 days prepared by us was used as 5 levels to model the results of machine vision detection, and the support vector machine classification method was used to construct a nonlinear decision system for the degree of potato browning, so as to use the trained model to make qualitative judgments on other samples.

[0035] (5) Quantitative determination of the degree of browning of potatoes. The experimental samples were scored by experts, and the expert scores of the samples were used as training output values. The regression model was constructed using the partial least squares method (PLS). After the model was trained, the feature values ​​obtained by machine vision were input to obtain the expert scoring results.

[0036] It should be noted that, for the purpose of clarity, representation and description of components or processes that are not related to the present invention and are known to those of ordinary skill in the art are omitted in the drawings and the description.

[0037] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A potato browning detection method based on machine vision, characterized in that Follow the steps below: (1) Build a machine vision inspection device and prepare experimental samples; (2) Capturing potato sample images in a machine vision inspection device, taking images with a CCD camera and saving them in a computer; (3) Preprocess the image, convert the RGB image into a grayscale image, and use a filtering algorithm to remove noise; (4) Background segmentation: Use Otsu's threshold segmentation method to obtain the target area. A small amount of noise interference and target missing are further processed using corrosion and expansion operators; (5) Feature extraction: convert the extracted target image into three color spaces for expression, namely RGB, HSV, and Lab, and segment them into single-channel images; extract the grayscale mean and standard deviation of each channel from the nine single-channel images in the above three color spaces, analyze the color from different angles, and together form the image information describing a sample; (6) Establish a browning degree discrimination model and conduct qualitative or quantitative analysis on each sample; In step (6), a browning degree discrimination model is established, and the browning degrees of the prepared potatoes at 0, 2, 4, 6, and 8 days are used as 5 levels to model the machine vision detection results, so as to perform qualitative discrimination on other samples with the obtained model; In step (6), the browning degree discrimination model is established. If a more accurate browning degree is required, the samples are scored manually or other instrumental methods are used to obtain the browning score, and then the scoring results are regressed with the machine vision detection results to obtain a browning quantitative scoring model.

2. A potato browning detection method based on machine vision according to claim 1, characterized in that The device for machine vision inspection described in step (1) is an integrated device and a detachable device; the device is designed and assembled according to the specific inspection environment and inspection object, and can be disassembled and replaced at any time to meet image acquisition requirements in different experimental sites, production lines or when inspecting multiple objects at a time.

3. A potato browning detection method based on machine vision according to claim 1, characterized in that The integrated machine vision detection device described in step (1) comprises a hardware device and a software system. The hardware device comprises a housing of the integrated machine vision detection device, a ring light, a CCD camera, and a computer. The software part consists of an electrical control module and an image processing module. A CCD camera is installed on the top of the housing of the integrated machine vision detection device through a bayonet, and the CCD camera is connected to the computer through a data cable. A ring light is also installed on the housing wall of the integrated machine vision detection device.

4. A potato browning detection method based on machine vision according to claim 1, characterized in that The sample described in step (1) is a fresh potato, the variety is Dutch potato, and the flesh is yellow; when selecting samples, try to select samples with good quality and the same appearance characteristics, the appearance characteristics are size, color and shape; the purchased samples are stored in a cool and dry environment, the room temperature is controlled at 13±2°C for one week, and the samples with no obvious change in appearance are selected for the experiment; before the experiment begins, the selected potato samples are cleaned with tap water and the skin is peeled off; the cleaned potatoes are cut into uniform 3 mm thick slices, the equipment used is a mechanical slicer, model NEMCO-55200AN; then a uniform template is used to cut the slices into squares with a length and width of 3.5 cm, and the prepared samples are placed in a refrigerator at 4°C for standby use; the potato samples are tested at 0, 2, 4, 6, and 8 days, and 15 slices are randomly selected from the standby samples each time.

5. The method for detecting potato browning based on machine vision according to claim 1, characterized in that In step (2), when collecting potato sample images in the machine vision detection device, the form, intensity, angle of illumination, height and focal length of the camera are all adjustable to ensure that the color and texture information of the potato surface is fully and clearly obtained.

6. The method for detecting potato browning based on machine vision according to claim 1, characterized in that The filtering algorithm in step (3) is median filtering.

7. The method for detecting potato browning based on machine vision according to claim 1, characterized in that The background segmentation described in step (4) is to perform detection under a black background. If the background is more complex, it will be accurately processed according to the specific situation.

8. The method for detecting potato browning based on machine vision according to claim 1, characterized in that The grayscale mean and standard deviation of each channel are extracted from the nine single-channel images in the three color spaces described in step (5) as part of the sample color information, and the maximum or minimum value is added as needed to supplement the sample information to increase accuracy.

Images