CN108896564A - A kind of two-sided damage testing method of corn seed based on machine vision - Google Patents

Abstract

The invention relates to the technical field of machine vision detection, in particular to a machine vision-based double-sided damage detection method for corn seeds. The process mainly includes: collection of double-sided images of corn seeds, image preprocessing, feature value extraction, and using the trained support vector machine (SVM) to judge whether the corn seeds are qualified. Its features are as follows: a machine vision-based double-sided damage detection method for corn seeds is used to detect the double-sided images of corn seeds, which can comprehensively and accurately detect damaged corn seeds; Carry out damage classification, and when the test results are all 'qualified', it is judged that the corn seeds are qualified, otherwise they are unqualified.

Description

A double-sided damage detection method for corn seeds based on machine vision

technical field

The invention relates to the technical field of machine vision detection, in particular to a method for detecting double-sided damage of corn seeds based on machine vision.

Background technique

Corn is an important food crop in my country. Most of the pre-sowing seed detection work is done manually, which is greatly affected by individual subjective factors and has low efficiency, which cannot meet market demand. With the gradual maturity of machine vision inspection technology, it is more and more widely used in the field of agricultural product quality inspection. It is characterized by accuracy, high efficiency, and non-destructive. However, in the field of appearance quality inspection of corn seeds, large-scale commercial double-sided inspection has not yet appeared. There are few related double-sided damage detection methods. Therefore, a machine vision-based double-sided damage detection method for corn seeds proposed by the present invention has certain application value.

Contents of the invention

The present invention aims to provide a method for detecting double-sided damage of corn seeds based on machine vision. The method is accurate and efficient, can realize the recognition and detection of double-sided images of corn seeds, and accurately judge whether the corn seeds are damaged.

To achieve the above object, the solution of the present invention is as follows:

A method for detecting double-sided damage of corn seeds based on machine vision mainly includes the following steps:

Step 1: Use an image acquisition device to collect double-sided images of corn seeds, wherein CCD1 collects images of the front side of corn seeds, and CCD2 collects images of the back side of corn seeds.

Step 2: The frontal image of corn seeds collected by CCD1 is transmitted to the computer through the serial interface, and the computer preprocesses the image, removes noise and impurity points in the image, and performs binary segmentation on the image for subsequent image classification .

Step 3: Extract the eigenvalue parameters of the preprocessed corn seed front image. The eigenvalues include: perimeter, area, perimeter-area ratio, major axis length, minor axis length, aspect ratio, circularity, and rectangularity , compactness and 7 Hu invariant moments.

Step 4: Input the feature value of the frontal image as the input value into the trained SVM classifier 1. The SVM classifier has two outputs. When the output is 1, it means 'qualified', and when the output is 0, it means 'unqualified'.

Step 5: CCD2 will collect the image of the reverse side of the corn seed and transmit the image to the computer through the serial interface. The computer will preprocess the image, remove noise and impurity points in the image, and perform binary segmentation on the image.

Step 6: Extract the feature parameters of the reverse image of corn seeds.

Step 7: Input the feature value of the negative image as the input value into the trained SVM classifier 2, when the output is 1, it means 'qualified', and when the output is 0, it means 'unqualified'.

Step 8: When the output values of SVM1 and SVM2 are both 1, the corn seed is finally judged to be 'qualified'; when the output of SVM1 is 1, the output value of SVM2 is 0, the output value of SVM1 is 0, the output value of SVM2 is 1, and When the output values of SVM2 are all 0, the corn seeds are finally judged as 'unqualified'.

Description of drawings

Fig. 1 is a schematic diagram of an image acquisition device of the present invention.

Fig. 2 is a flow chart of double-sided damage detection of corn seeds in the present invention.

Fig. 3 is a flow chart of SVM image judgment in the present invention.

Detailed ways

The present invention will be further described in detail below in conjunction with various embodiments shown in the accompanying drawings.

With reference to Fig. 1, image acquisition device of the present invention mainly comprises: CCD1 and light shield [1] thereof, CCD2 and light shield [2] thereof, image acquisition area [3], support [4].

With reference to Fig. 2, the double-sided damage detection flow chart of corn seed of the present invention, CCD1 transmits the front image of corn seed that collects to computer, and computer carries out preprocessing, feature extraction work to front image successively, with characteristic value output SVM1, classifier output There are two quantities, when the output result is 1, it is 'qualified', and when the output result is 0, it is 'unqualified'; CCD2 transmits the reverse image of corn seeds collected to the computer, and the computer performs preprocessing and feature extraction on the reverse image in turn , output the eigenvalues to SVM2, there are two classifier outputs, when the output result is 1, it is 'qualified', and when the output result is 0, it is 'unqualified', the main lines of the two workflows are to perform frontal image detection first, then Reverse image detection. Image preprocessing includes several basic steps of grayscale, noise reduction, binarization segmentation, and removal of impurities.

With reference to Fig. 3, SVM image judging flow chart of the present invention, SVM1 first judges corn seed front image, when output value is 1, start SVM2 and carry out qualification judgment to the back image of corn seed, when output value is not 1, terminate The program judges the corn seeds as 'unqualified'; when the output value of SVM2 is 1, then judges the corn seeds as 'qualified'; when the output value of SVM2 is not 1, then judges the corn seeds as unqualified'.

Claims (4)

1. a kind of two-sided damage testing method of corn seed based on machine vision, process mainly include：Utilize Image Acquisition The dual-side image of device acquisition corn seed；Image is sent to computer by serial ports；Computer to image carry out gray processing, The base images pretreatments such as noise reduction, binarization segmentation, removal impure point；It is special to being carried out by pretreated corn seed image Value indicative is extracted；The characteristic value of extraction is inputted in trained SVM classifier, the input value of SVM classifier is characterized parameter, has Two output valves export and represent ' qualification ' when being 1, represent ' unqualified ' when output be 0, to realize that the damaged of corn seed is examined It surveys.

2. a kind of two-sided damage testing method of corn seed based on machine vision according to claim 1, feature exist It works in the damage testing for being able to achieve corn seed dual-side image.

3. a kind of two-sided damage testing method of corn seed based on machine vision according to claim 1, feature exist In using two SVM classifiers respectively to the progress damage testing classification of corn seed front and back sides image, job order SVM1 First front corn seed image is detected, SVM2 again detects reverse side corn seed image.

4. a kind of two-sided damage testing method of corn seed based on machine vision according to claim 1, feature exist In after SVM image judges that process starts, SVM1 first determines whether corn seed direct picture, when output valve is 1, starts SVM2 Qualification judgement is carried out to the verso images of corn seed, when output valve is not 1, terminator judges corn seed for ' no It is qualified '；When the output valve of SVM2 is 1, then judge that corn seed for ' qualification ', when the output valve of SVM2 is not 1, then judges Corn seed is not ' unqualified '.