Method for detecting surface orange by analyzing intensity images of shadows formed on surface of object by light source

[0032] The present invention will be further described in detail below with reference to the drawings and embodiments.

[0033] The specific steps of the present invention are:

[0034] The first step: image acquisition:

[0035] Fix the light source and the image acquisition component in a black box to form a set of image acquisition equipment. There are two rectangular slits on the surface of the black box. Between the slits is an image capture hole, and image capture elements such as CMOS/CDT lenses or sensors are placed inside. The black box has two built-in light strips composed of several standard LED light sources, and a diffuser is placed between the light strips and the slit. When the LED light strip is lit, the light emitted by it passes through the diffuser and the slit to form two uniform rectangular light strips. At this time, the rectangular light strip illuminates the surface of the object, and the two-dimensional brightness image formed by the light source on the surface of the object is collected by image acquisition equipment such as CMOS/CDT. Control the distance between the acquisition equipment and the measured object to make the captured image clearest.

[0036] The second step: use the detection algorithm to analyze the image to obtain the texture structure value of the orange peel on the surface. The specific method is:

[0037] 1. Image preprocessing: including

[0038] a) Position the image of the light source:

[0039] Binarize the two-dimensional brightness image with a given threshold of 100 (which can be adjusted according to the actual situation) to obtain a black and white image, and use the Seed Filling algorithm to find all connected areas in the black and white image, of which the two largest connected areas are the light source The main part of the image;

[0040] b) Calculate the positioning area:

[0041] Properly transform the connected area required in a) to obtain the location area; increase the height of the connected area so that the location area can contain complete image boundary information; reduce the width of the connected area to eliminate the end effect on the resulting image ( figure 1 ).

[0042] 2. Brightness sampling:

[0043] The brightness of the light source image is sampled, and a scan line is selected laterally at a certain distance in the upper and lower bright bands of the light source image area to obtain m one-dimensional brightness signals B j (j=1, 2...m), image 3 The black straight line in the middle of the bright band is a scan line, Figure 4 Is the one-dimensional brightness signal B of the scan line j.

[0044] 3. Extract orange peel signal:

[0045] For the one-dimensional brightness signal B selected in 2 j Decompose, extract the orange peel signal after denoising O j :

[0046] a) Use the EMD (Empirical Mode Decomposition) algorithm to decompose the one-dimensional luminance signal into n sub-signals of different frequencies ( Figure 5 ), and number these sub-signals as S from high to low once in frequency 1...S n;

[0047] b) Divide these sub-signals into three categories:

[0048] S 1 (The first sub-signal) is high-frequency white noise;

[0049] S 2 To S i ( ) Is orange peel signal;

[0050] S i To S n Is system noise.

[0051] Extract the orange peel signal S 2 To S i And synthesize the sub-signals to obtain a one-dimensional orange peel signal O j ( Figure 6 ):

[0052]

[0053] 4. Statistics:

[0054] a) Statistical one-dimensional orange peel signal O j The number of long and short waves in the medium: the wave signal between two adjacent maximum points or two adjacent minimum points on the wave signal is regarded as a complete wave, and the distance between them is defined as one wavelength, and the wavelength is less than The 0.6mm wave is summarized as short wave, the wave with wavelength greater than 0.6mm and less than 10mm is summarized as long wave, and the one-dimensional orange peel signal after denoising is calculated. j The number of long waves in LW j And shortwave number SW j;

[0055] b) Calculate the one-dimensional orange peel signal O j LW in j And SW j Ratio R j =LW j /SW j;

[0056] c) Calculate the ratio R of the number of long and short waves in each orange peel signal j (j=1, 2...m) average value Avg;

[0057] The third step: Convert orange peel data to orange peel evaluation through parameter calibration.

[0058] The average value Avg is linearly transformed, and its value is converted into a value OP between 0 and 99.9 through the calibration parameters (a, b) of the detection object type, OP=a*Avg+b. The converted value OP can represent the texture structure value of the orange peel on the surface of the measured object. The higher the score, the smaller the surface orange peel and the better the quality.

[0059] The content not described in detail in this specification belongs to the prior art known to those skilled in the art.