CN112381751A - Online intelligent detection system and method based on image processing algorithm - Google Patents
Online intelligent detection system and method based on image processing algorithm Download PDFInfo
- Publication number
- CN112381751A CN112381751A CN202010646900.9A CN202010646900A CN112381751A CN 112381751 A CN112381751 A CN 112381751A CN 202010646900 A CN202010646900 A CN 202010646900A CN 112381751 A CN112381751 A CN 112381751A
- Authority
- CN
- China
- Prior art keywords
- image
- module
- black
- algorithm
- pixel
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 62
- 238000001514 detection method Methods 0.000 title claims abstract description 62
- 238000012545 processing Methods 0.000 title claims abstract description 35
- 238000000034 method Methods 0.000 title claims abstract description 25
- 230000009467 reduction Effects 0.000 claims abstract description 17
- 238000003708 edge detection Methods 0.000 claims abstract description 14
- 238000000605 extraction Methods 0.000 claims abstract description 11
- 238000013473 artificial intelligence Methods 0.000 claims abstract description 8
- 238000013528 artificial neural network Methods 0.000 claims abstract description 8
- 238000007619 statistical method Methods 0.000 claims abstract description 8
- 238000004364 calculation method Methods 0.000 claims description 11
- 238000011105 stabilization Methods 0.000 claims description 11
- 230000008859 change Effects 0.000 claims description 9
- 238000001914 filtration Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 7
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 238000012937 correction Methods 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims description 4
- 238000012935 Averaging Methods 0.000 claims description 3
- 238000005315 distribution function Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000005259 measurement Methods 0.000 abstract description 3
- 238000004519 manufacturing process Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 238000007689 inspection Methods 0.000 description 3
- 238000009795 derivation Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000010191 image analysis Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000003908 quality control method Methods 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000010923 batch production Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000002087 whitening effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration using histogram techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30141—Printed circuit board [PCB]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Geometry (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention provides an online intelligent detection system and method based on an image processing algorithm, and relates to the technical field of intelligent detection, wherein the system comprises an image extraction module, an image processing module, a noise reduction module intelligent detection module and an artificial intelligent learning module, wherein the image extraction module is used for acquiring RGB color images of an object to be detected on line; after the image processing module carries out equalization processing on the color image histogram, carrying out edge detection; the noise reduction module is used for converting the color image into a black-and-white image and reducing noise points of the black-and-white image; the intelligent detection module divides the black and white image into a plurality of blocks to be detected, performs statistical analysis on the surface characteristics in the blocks to obtain detection results, and marks the blocks by using a marking algorithm; the artificial intelligence learning module optimizes the noise reduction module and the intelligent detection module by utilizing a neural network algorithm. The method is used for solving the technical problems that the detection method adopted by the prior art cannot unify the measurement standard, and is high in cost and low in efficiency.
Description
Technical Field
The invention relates to the technical field of intelligent detection, in particular to an online intelligent detection system and method based on an image processing algorithm.
Background
China is used as a world factory, and large-scale and large-batch production of mechanical, electronic and electrical equipment is very common. Generally, for quality control in mass production, each large factory adopts a quality inspection department to perform sampling inspection, and a large amount of labor force manual inspection strategies are invested in cooperation with a production department. The reliability of such conventional testing strategies is related to the level of quality control departments, the quality of labor, and factory management (standard work procedures).
For example, after a circuit board is produced in a large scale, the hole shape of the circuit board needs to be detected to ensure that the hole shape is qualified and not rough, but the manual detection has the following disadvantages: 1. the detection method cannot unify the measurement standards, has large error and poor stability, and has high requirements on workers; 2. today, the labor price is increasingly rising, the cost of the traditional detection strategy is higher, and the traditional detection strategy is not compliant with the current production requirement; 3. the efficiency is low.
Disclosure of Invention
In view of the above drawbacks of the prior art, an object of the present invention is to provide an online intelligent detection system and method based on an image processing algorithm, which are used to solve the technical problems that the detection method adopted in the prior art cannot unify the measurement standards, and is high in cost and low in efficiency.
The invention provides an intelligent detection system based on an image processing algorithm, which comprises an image extraction module, an image processing module, a noise reduction module, an intelligent detection module and an artificial intelligence learning module,
the image extraction module is used for acquiring RGB color images of the object to be detected on line;
the image processing module performs edge detection after the color image histogram is equalized;
the noise reduction module is used for converting the color image into a black-and-white image and reducing the noise point of the black-and-white image;
the intelligent detection module divides the black and white image into a plurality of lumps to be detected, performs statistical analysis on the surface characteristics in the lumps to obtain a detection result, and marks the lumps by using a marking algorithm;
the artificial intelligence learning module optimizes the noise reduction module and the intelligent detection module by using a neural network algorithm, and optimizes parameters so as to improve the success rate of detection.
In an embodiment of the present invention, the image extraction module includes a dynamic self-stabilization holder and an image library, and a high-definition camera is disposed on the dynamic self-stabilization holder, and the high-definition camera is configured to acquire RGB color channel images on line and store one of the frames of color images in the image library.
In an embodiment of the present invention, the image processing module includes a contrast adjusting unit and an edge detecting unit, the contrast adjusting unit utilizes a histogram equalization algorithm to enhance local brightness of the color image, and the edge detecting unit utilizes an edge detecting algorithm to find a boundary of the color image and extract a preset feature point in the color image to establish a position coordinate system.
The noise reduction module converts the color image into a black and white image of 0 or 255 by utilizing a binarization algorithm so as to reduce the data volume in the image; and the two-dimensional FIR filtering algorithm is utilized to reduce the noise of the black and white image, so that the black and white image can highlight the target contour.
And the surface characteristics comprise the area, the perimeter, the centroid and the third-order standard moment of the block mass, so that the pixel size of the block mass is calculated, the pixel size is compared with a preset range, if the pixel size is within the preset range, the block mass is qualified, and a square symbol is adopted to mark the block mass.
An online intelligent detection method based on an image processing algorithm, comprising the following steps:
s1: placing an article to be detected below a high-definition camera, and collecting RGB color images of the article to be detected by the camera;
s2: after the histogram of the color image is equalized, carrying out edge detection;
s3: converting the color image into a black-and-white image, and simultaneously reducing the noise of the black-and-white image;
s4: dividing the black-white image into a plurality of blocks, carrying out statistical analysis on the surface characteristics in the blocks to obtain a detection result, and marking the blocks by using a marking algorithm.
In an embodiment of the present invention, the step S2 specifically includes:
s2.1: the RGB color channel image comprises R, G, B three pixel channels, the number of each gray value in the three pixel channels is counted, the probability P of each gray value is calculated, a new gray value of the RGB color channel image is obtained through an accumulative distribution function, and the specific calculation formula is as follows:
wherein n represents the number of pixel points, njIndicates the number of pixels with a gray value of j, SkNew gray levels (it can be known that the gray levels of 0 and L-1 are mapped without change);
s2.2: utilizing a Sobel operator to perform weighted summation on pixel values in the color image area (namely, each pixel value is respectively multiplied by each element of a convolution template, the convolution results are summed, and the sum obtained by operation is the result of convolution operation), wherein the specific calculation formula is as follows:
g(i,j)＝∑k,l f(i-k,j-l)h(k,l)＝∑k,l f((k,l)h(i-k,j-l)；
setting a threshold value for the result, outputting the calculated pixel value to be 1 if the calculated pixel value is larger than the threshold value, outputting the pixel value to be 0 if the calculated pixel value is smaller than the threshold value, and taking the boundary point of 1 and 0 as an edge area;
s2.3: projecting the black-white image to a new viewing plane through projection change, finding a projection conversion matrix for mapping the maximum pixel by using the edge area, and extracting preset characteristic points in the color image to establish a position coordinate system.
In an embodiment of the present invention, the step S3 specifically includes:
s3.1: processing the color image into a black-and-white image with the gray value of 0 or 255 by using a binarization algorithm;
s3.2: the method comprises the following steps of removing noise points of black and white images by adopting a two-dimensional FIR filtering algorithm, wherein x (n) is a time sequence, a (n) is an input signal, the input signal changes along with the change of time, and the FIR filtering algorithm y (k) finally outputs the input at each moment multiplied by corresponding weight coefficients for superposition output, namely, the signals are subjected to multiple moving averaging treatment, and the specific calculation formula is as follows:
in an embodiment of the present invention, the step S4 specifically includes:
s4.1: extracting the block mass to be detected of the black and white image by an automatic correction algorithm,
s4.2: calculating the surface characteristics of the lumps, including area, centroid, perimeter and third-order standard moment, wherein the perimeter is obtained through the length of the chain code, and the area and the centroid can be obtained in the process of directly solving the third-order standard moment through the representative point;
s4.3: calculating the pixel point size of the block mass through surface characteristics, and comparing the pixel point size with a preset range;
s4.4: if the pixel point size is in the preset range, the block is qualified, and a square mark is adopted to circle the block.
In an embodiment of the present invention, if the size of the pixel point is outside the preset range, the blob is marked as a triangle, which is not qualified.
As described above, the present invention has the following advantageous effects:
1. the invention replaces the traditional manual detection with the related detection equipment and the adaptive intelligent detection algorithm, not only innovates the detection method in the batch processing production and greatly improves the detection efficiency, but also parameterizes the batch detection standard to a certain extent, provides a standardized paradigm for the detection process, has high system intelligence degree, is suitable for automatic assembly line operation and conforms to the new era trend of the industry.
2. According to the invention, the high-definition image of the product to be tested is rapidly acquired through the high-definition camera, the dynamic self-stabilization holder can automatically adjust the position of the camera to rapidly focus the product, meanwhile, the edge area of the high-definition image is found by using an edge detection algorithm, and the preset characteristic points are extracted from the color image to establish a position coordinate system (reference system of the product to be tested), namely, the product to be tested is not required to be strictly placed at the appointed position by an operator, only the approximate position is required to be placed, the dynamic self-stabilization holder can automatically find the product to be tested and automatically focus the product to be tested, the relative error caused by slight difference of clamping positions is not required to be worried about, and.
3. By the neural network algorithm, an operator does not need to know professional programming knowledge, only needs to prepare data of qualified products in advance for the computer to detect and learn, the detection equipment can independently learn by the algorithm and establish an information base of related data of the qualified products, meanwhile, the equipment can continuously learn in a detection task according to the algorithm, and the computer becomes more and more intelligent and is competent for the detection task after repeated learning for many times. The operator does not need to have professional computer knowledge, and the technical threshold is reduced for the detection work of the factory.
Drawings
FIG. 1 is a block diagram of the system disclosed herein.
FIG. 2 is a diagram illustrating the steps of the disclosed method.
FIG. 3 is a flow chart of the present disclosure.
Fig. 4 shows an original image collected by the high definition camera disclosed in the present invention.
Fig. 5 is a diagram of the noise reduction module according to the disclosure.
Fig. 6 shows a diagram with a reference symbol for the present disclosure.
Detailed Description
The embodiments of the present invention are described below with reference to specific embodiments, and other advantages and effects of the present invention will be easily understood by those skilled in the art from the disclosure of the present specification. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention. It is to be noted that the features in the following embodiments and examples may be combined with each other without conflict.
It should be noted that the drawings provided in the following embodiments are only for illustrating the basic idea of the present invention, and the components related to the present invention are only shown in the drawings rather than drawn according to the number, shape and size of the components in actual implementation, and the type, quantity and proportion of the components in actual implementation may be changed freely, and the layout of the components may be more complicated.
Referring to fig. 1, the present invention provides an intelligent detection system based on image processing algorithm, which comprises an image extraction module, an image processing module, a noise reduction module, an intelligent detection module and an artificial intelligence learning module,
the image extraction module is used for acquiring RGB color images of the object to be detected on line;
the image processing module performs edge detection after the color image histogram is equalized;
the noise reduction module is used for converting the color image into a black-and-white image and reducing the noise point of the black-and-white image;
the intelligent detection module divides the black and white image into a plurality of lumps to be detected, performs statistical analysis on the surface characteristics in the lumps to obtain a detection result, and marks the lumps by using a marking algorithm;
the artificial intelligence learning module optimizes the noise reduction module and the intelligent detection module by using a neural network algorithm, and optimizes parameters so as to improve the success rate of detection.
In this embodiment, in the intelligent detection module, a MATLAB-based Blob analysis algorithm is used to simulate a human visual processing process, and during image analysis, an image is divided into a plurality of regions according to characteristics such as color, texture, and contour, so as to obtain a plurality of mutually disjoint closed regions, which are called blobs (blobs), and pixels in the blobs are spatially communicated and have similar image characteristics such as color texture contour, and a corresponding algorithm is used according to characteristics such as hole or cavity of a target to be detected.
Based on the above embodiment, the image extraction module comprises a dynamic self-stabilization holder and an image library, wherein a high-definition camera is arranged on the dynamic self-stabilization holder, the dynamic self-stabilization holder is provided with a Beagleboard high-performance development board carrying an ARM-A8 processor, and the self-stabilization algorithm of the dynamic holder is adapted, so that the position of the high-definition camera can be automatically adjusted to quickly focus a product, and the high-definition camera is used for collecting RGB color channel images on line and storing one frame of color image in the image library;
in this embodiment, the operator only needs to put the product to be measured at the approximate position, and the dynamic self-stabilization holder can find the product to be measured by itself and perform automatic focusing.
Based on the above embodiment, the image processing module includes a contrast adjusting unit and an edge detecting unit, the contrast adjusting unit utilizes a histogram equalization algorithm to enhance the local brightness of the color image, and the local (brightness) of the image is enhanced by the histogram equalization algorithm without affecting the overall contrast; the edge detection unit finds the boundary of the color image by using an edge detection algorithm and extracts preset feature points in the color image to establish a position coordinate system.
The noise reduction module converts the color image into a black and white image of 0 or 255 by using a binarization algorithm to reduce the data volume in the image, and the binarization algorithm can extract only a small part of information related to calculation in a large amount of detail information of the image, thereby providing feasibility for timely detection (large amount of calculation); and the two-dimensional FIR filter algorithm is utilized to reduce the noise of the black and white image and make the black and white image show the target contour, and the two-dimensional FIR filter is used as a basic element in the digital signal processing, thereby providing a mature proposal for reducing the image noise and improving the success rate of the image processing.
And the surface characteristics comprise the area, the perimeter, the centroid and the third-order standard moment of the block mass, so that the pixel size of the block mass is calculated, the pixel size is compared with a preset range, if the pixel size is within the preset range, the block mass is qualified, and a square symbol is adopted to mark the block mass.
Referring to fig. 2 and 3, an online intelligent detection method based on an image processing algorithm includes the following steps:
s1: placing an article to be detected below a high-definition camera, and acquiring an RGB color image of the article to be detected by the camera, as shown in FIG. 4;
s2: after the histogram of the color image is equalized, carrying out edge detection;
s3: converting the color image into a black-and-white image while reducing noise of the black-and-white image, as shown in fig. 5;
s4: the black-and-white image is divided into a plurality of blobs, statistical analysis is performed on the face features in the blobs to obtain detection results, and the blobs are marked by using a marking algorithm, as shown in fig. 6.
Based on the above embodiment, the step S2 specifically includes:
s2.1: the RGB color channel image comprises R, G, B three pixel channels, the number of each gray value in the three pixel channels is counted, the probability P of each gray value is calculated, a new gray value of the RGB color channel image is obtained through an accumulative distribution function, and the specific calculation formula is as follows:
wherein n represents the number of pixel points, njIndicates the number of pixels with a gray value of j, SkIs the new gray level (it is known that the gray levels of 0 and L-1 are mapped unchanged).
S2.2: utilizing a Sobel operator to perform weighted summation on pixel values in the color image area (namely, each pixel value is respectively multiplied by each element of a convolution template, the convolution results are summed, and the sum obtained by operation is the result of convolution operation), wherein the specific calculation formula is as follows:
g(i,j)＝∑k,l f(i-k,j-l)h(k,l)＝∑k,l f((k,l)h(i-k,j-l)； (2)
setting a threshold value for the result, outputting 1 if the calculated pixel value is larger than the threshold value, outputting 0 if the calculated pixel value is smaller than the threshold value,
the boundary point between 1 and 0 is the edge region, which is the pixel set of the image gray value sharp change region.
S2.3: the black-white image is projected to a new viewing plane through projection change, the edge area is utilized to find out a projection conversion matrix for mapping the maximum pixel, and preset characteristic points are extracted from a color image to establish a position coordinate system, because a product to be measured is placed at an approximate position, the shot image is not a fixed angle, and the color image can be positioned by extracting the characteristic points, so that each image is in a preset standard position coordinate system, the next operation is convenient, wherein the formula of the projection conversion matrix is as follows:
u and v are original pictures, and the coordinates of the corresponding transformed pictures are x and y, wherein
Based on the above embodiment, the step S3 specifically includes:
s3.1: processing the color image into a black-and-white image with the gray value of 0 or 255 by using a binarization algorithm, namely blackening and whitening the whole image, so that the processable data volume in the image is obviously reduced;
s3.2: the method comprises the following steps of removing noise points of black and white images by adopting a two-dimensional FIR filtering algorithm, wherein x (n) is a time sequence, a (n) is an input signal, the input signal changes along with the change of time, the FIR filtering algorithm y (k) finally outputs the input at each moment multiplied by corresponding weight coefficients, and the input signal is superposed and output, namely, the signal is subjected to multiple moving averaging processing, so that the value at a certain moment is positioned at a more accurate position, and finally, the purpose of removing the noise waves and obtaining a clearer image is achieved, and the specific calculation formula is as follows:
wherein n and k represent the number of signals.
Based on the above embodiment, the step S4 specifically includes:
s4.1: extracting the block mass to be detected of the black and white image by an automatic correction algorithm,
s4.2: calculating the surface characteristics of the lumps, including area, centroid, perimeter and third-order standard moment, wherein the perimeter is obtained through the length of the chain code, and the area and the centroid can be obtained in the process of directly solving the third-order standard moment through the representative point; the chain code is a method for describing a curve or a boundary by using a curve starting point coordinate and a boundary point direction code, and is a coding representation method of the boundary, and in the example, the chain code is a set of boundary points; the representative points are the static distance/graphic area on the X-axis and the static distance/graphic area on the Y-axis. The face (hole) characteristics in this example are given by:
wherein, nu is the average intensity, N is the number of color samples, and P is the color intensity; a first moment defining the average intensity of each color component;
σ is variance, P is color intensity, and ν is first moment; a second moment reflecting the color variance, i.e., non-uniformity, of the region to be measured;
xi is skewness, P is color intensity, and v is first moment; the third moment, defines the skewness of the color components, i.e., the asymmetry of the color.
S4.3: calculating the pixel point size of the block mass through surface characteristics, and comparing the pixel point size with a preset range;
s4.4: if the pixel point size is in the preset range, the block is qualified, and a square mark is adopted to circle the block. Xi
Based on the above embodiment, if the size of the pixel point is outside the preset range, the blob is marked as a triangle, which is not qualified.
In this embodiment, a MATLAB-based Blob analysis algorithm simulates a human visual processing process, and during image analysis, an image is divided into a plurality of regions according to characteristics such as color, texture, and contour, so as to obtain a plurality of mutually disjoint closed regions, which are called blobs (blobs), and pixels in the blobs are spatially connected and have similar image characteristics such as color, texture, and contour.
The specific embodiment is as follows: the circuit board is used as an article to be detected, and whether hole processing in the circuit board is qualified is detected, so that a target to be detected is a hole, and a block mass analysis needs to be carried out by extracting a surface feature as a hole (circle);
s1: placing the circuit board to be tested below the high-definition camera, and using the camera to test the RGB color image of the circuit board to be tested;
s1.1: the position of the high-definition camera is automatically adjusted to quickly focus the circuit board to be tested;
s1.2: the high-definition camera is used for collecting RGB color channel images of the circuit board to be detected on line;
s1.3: storing one frame of color image in an image library;
s2: after the histogram of the color image is equalized, carrying out edge detection, and establishing a position coordinate system of an image to be processed;
s3: converting the color image into a black-and-white image, and simultaneously reducing the noise of the black-and-white image to obtain a 'black-and-white' image for highlighting the contour of the target;
s4: dividing the black-white image into a plurality of blocks, carrying out statistical analysis on the surface characteristics in the blocks to obtain a detection result, and marking the blocks by using a marking algorithm.
S4.1: extracting the block mass to be detected of the black and white image by an automatic correction algorithm,
s4.2: calculating the surface characteristics of the lumps, including area, centroid, perimeter and third-order standard moment, wherein the perimeter is obtained through the length of the chain code, and the third-order standard moment, the centroid and the perimeter are obtained through the representative points;
s4.3: calculating the pixel point size of the block mass through the surface characteristics, and comparing the pixel point size with a preset range, wherein the preset range is 20-50 pixel points, namely the size of a processing hole under the resolution of a camera is 20-50 pixel points;
s4.4: if the size of the pixel point is in a preset range, the block mass is qualified, and the block mass is circled out by adopting a square mark '□'.
Based on the above embodiment, if the size of the pixel point is outside the preset range, the blob is marked as a triangle, i.e., the blob is unqualified.
Specifically, the process of optimizing the noise reduction module and the intelligent detection module by the artificial intelligence learning module by using a neural network algorithm is as follows:
selecting a plurality of samples from data obtained by manual detection as initial learning samples;
calculating partial derivatives of the loss function to all weights and biases;
thirdly, updating each weight and bias of the formula by using an SGD random gradient descent method;
and fourthly, repeating the steps 1 to three, namely learning.
Built in this example is a simple neural network comprising an input layer (two pixel parameters x1, x2 for planar features), a hidden layer (h1, h2) sandwiched between the input layer and the output layer, and an output layer o1 comprising 1 neuron (o 1 is 1 when the test sample is good and o1 is 0 when the test sample is bad), the output formula being:
o1＝f(w×[h1,h2]+b)， (8)
wherein w is a weight; b is an offset; mean Square Error (MSE) is introduced to define the loss, the smaller the loss, the better the learning effect of the neural network, i.e.:
wherein n is the number of samples; y represents the output result, 1 is qualified, and 0 is not qualified; y-tune is the true value of the variable and y-pred is the predicted value of the variable.
According to the partial derivative and the chain derivation method, the adjustment of the weight and the bias parameter can be known, so that the size of a predicted value and a loss function MSE can be changed, an optimization algorithm (SGD) called random gradient descent is introduced to train the network, and the accuracy is improved;
wherein, the chain type derivation formula is:
wherein, η is the learning rate, which determines the speed of the training network, and w 1' is the new weight.
In conclusion, the invention replaces the traditional manual detection with the related detection equipment and the intelligent detection algorithm matched with the related detection equipment, thereby not only renovating the detection method in batch processing production, but also greatly improving the detection efficiency. Therefore, the invention effectively overcomes various defects in the prior art and has high industrial utilization value.
The foregoing embodiments are merely illustrative of the principles and utilities of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or change the above-mentioned embodiments without departing from the spirit and scope of the present invention. Accordingly, it is intended that all equivalent modifications or changes which can be made by those skilled in the art without departing from the spirit and technical spirit of the present invention be covered by the claims of the present invention.
Claims (10)
1. An online intelligent detection system based on image processing algorithm is characterized in that: the system comprises an image extraction module, an image processing module, a noise reduction module, an intelligent detection module and an artificial intelligence learning module,
the image extraction module is used for acquiring RGB color images of the object to be detected on line;
the image processing module performs edge detection after the color image histogram is equalized;
the noise reduction module is used for converting the color image into a black-and-white image and reducing the noise point of the black-and-white image;
the intelligent detection module divides the black and white image into a plurality of lumps to be detected, performs statistical analysis on the surface characteristics in the lumps to obtain a detection result, and marks the lumps by using a marking algorithm;
the artificial intelligence learning module optimizes the noise reduction module and the intelligent detection module by utilizing a neural network algorithm.
2. The system of claim 1, wherein: the image extraction module comprises a dynamic self-stabilization holder and an image library, wherein a high-definition camera is arranged on the dynamic self-stabilization holder and used for collecting RGB color channel images on line and storing one frame of color image in the image library.
3. The system of claim 1, wherein: the image processing module comprises a contrast adjusting unit and an edge detection unit, the contrast adjusting unit utilizes a histogram equalization algorithm to enhance the local brightness of the color image, the edge detection unit utilizes an edge detection algorithm to find the edge area of the color image, and preset feature points are extracted from the color image to establish a position coordinate system.
4. The system of claim 1, wherein: the noise reduction module converts the color image into a black and white image of 0 or 255 by utilizing a binarization algorithm so as to reduce the data volume in the image; and the two-dimensional FIR filtering algorithm is utilized to reduce the noise of the black and white image, so that the black and white image can highlight the target contour.
5. The system of claim 1, wherein: and the surface characteristics comprise the area, the perimeter, the centroid and the third-order standard moment of the block mass, so that the pixel size of the block mass is calculated, the pixel size is compared with a preset range, if the pixel size is within the preset range, the block mass is qualified, and a square symbol is adopted to mark the block mass.
6. An online intelligent detection method based on an image processing algorithm is characterized by comprising the following steps:
s1: placing an article to be detected below a high-definition camera, and collecting RGB color images of the article to be detected by the camera;
s2: after the histogram of the color image is equalized, carrying out edge detection;
s3: converting the color image into a black-and-white image, and simultaneously reducing the noise of the black-and-white image;
s4: dividing the black-white image into a plurality of blocks, carrying out statistical analysis on the surface characteristics in the blocks to obtain a detection result, and marking the blocks by using a marking algorithm.
7. The method according to claim 6, wherein the step S2 specifically includes:
s2.1: the RGB color channel image comprises R, G, B three pixel channels, the number of each gray value in the three pixel channels is counted, the probability P of each gray value is calculated, a new gray value of the RGB color channel image is obtained through an accumulative distribution function, and the specific calculation formula is as follows:
wherein n represents the number of pixel points, njIndicates the number of pixels with a gray value of j, SkIs a new gray scaleA stage;
s2.2: utilizing a Sobel operator to perform weighted summation on pixel values in the color image area (namely, each pixel value is respectively multiplied by each element of a convolution template, the convolution results are summed, and the sum obtained by operation is the result of convolution operation), wherein the specific calculation formula is as follows:
g(i,j)＝∑k,lf(i-k,j-l)h(k,l)＝∑k,lf((k,l)h(i-k,j-l)；
setting a threshold value for the result, outputting the calculated pixel value to be 1 if the calculated pixel value is larger than the threshold value, outputting the pixel value to be 0 if the calculated pixel value is smaller than the threshold value, and taking the boundary point of 1 and 0 as an edge area;
s2.3: projecting the black-white image to a new viewing plane through projection change, finding a projection conversion matrix for mapping the maximum pixel by using the edge area, and extracting preset characteristic points in the color image to establish a position coordinate system.
8. The method according to claim 6, wherein the step S3 specifically includes:
s3.1: processing the color image into a black-and-white image with the gray value of 0 or 255 by using a binarization algorithm;
s3.2: the method comprises the following steps of removing noise points of black and white images by adopting a two-dimensional FIR filtering algorithm, wherein x (n) is a time sequence, a (n) is an input signal, the input signal changes along with the change of time, and the FIR filtering algorithm y (k) finally outputs the input at each moment multiplied by corresponding weight coefficients for superposition output, namely, the signals are subjected to multiple moving averaging treatment, and the specific calculation formula is as follows:
9. the method according to claim 6, wherein the step S4 specifically includes:
s4.1: extracting the block mass to be detected of the black and white image by an automatic correction algorithm,
s4.2: calculating the surface characteristics of the lumps, including area, centroid, perimeter and third-order standard moment, wherein the perimeter is obtained through the length of the chain code, and the area and the centroid can be obtained in the process of directly solving the third-order standard moment through the representative point;
s4.3: calculating the pixel point size of the block mass through surface characteristics, and comparing the pixel point size with a preset range;
s4.4: if the pixel point size is in the preset range, the block is qualified, and a square mark is adopted to circle the block.
10. The method of claim 9, wherein: and if the size of the pixel point is out of the preset range, marking the block into a triangle, namely, the block is unqualified.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010646900.9A CN112381751A (en) | 2020-07-07 | 2020-07-07 | Online intelligent detection system and method based on image processing algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010646900.9A CN112381751A (en) | 2020-07-07 | 2020-07-07 | Online intelligent detection system and method based on image processing algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112381751A true CN112381751A (en) | 2021-02-19 |
Family
ID=74586384
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010646900.9A Pending CN112381751A (en) | 2020-07-07 | 2020-07-07 | Online intelligent detection system and method based on image processing algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112381751A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113063353A (en) * | 2021-03-31 | 2021-07-02 | 深圳中科飞测科技股份有限公司 | Coordinate system establishing method, detection device, detection equipment and storage medium |
| CN113643352A (en) * | 2021-08-09 | 2021-11-12 | 贵州电网有限责任公司 | Natural icing on-line monitoring running wire image icing degree evaluation method |
| CN114292021A (en) * | 2021-12-30 | 2022-04-08 | 南京春辉科技实业有限公司 | System and method for adjusting preform rod in real time in quartz optical fiber drawing process |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101936918A (en) * | 2010-09-02 | 2011-01-05 | 东信和平智能卡股份有限公司 | BGA (Ball Grid Array) chip vision detecting system and detecting method thereof |
| CN107886131A (en) * | 2017-11-24 | 2018-04-06 | 佛山科学技术学院 | One kind is based on convolutional neural networks detection circuit board element polarity method and apparatus |
| CN108982512A (en) * | 2018-06-28 | 2018-12-11 | 芜湖新尚捷智能信息科技有限公司 | A kind of circuit board detecting system and method based on machine vision |
| CN110728692A (en) * | 2019-10-29 | 2020-01-24 | 中国计量大学 | Image edge detection method based on Scharr operator improvement |
| CN110956630A (en) * | 2019-12-18 | 2020-04-03 | 浙江大学 | Method, device and system for detecting plane printing defects |
-
2020
- 2020-07-07 CN CN202010646900.9A patent/CN112381751A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101936918A (en) * | 2010-09-02 | 2011-01-05 | 东信和平智能卡股份有限公司 | BGA (Ball Grid Array) chip vision detecting system and detecting method thereof |
| CN107886131A (en) * | 2017-11-24 | 2018-04-06 | 佛山科学技术学院 | One kind is based on convolutional neural networks detection circuit board element polarity method and apparatus |
| CN108982512A (en) * | 2018-06-28 | 2018-12-11 | 芜湖新尚捷智能信息科技有限公司 | A kind of circuit board detecting system and method based on machine vision |
| CN110728692A (en) * | 2019-10-29 | 2020-01-24 | 中国计量大学 | Image edge detection method based on Scharr operator improvement |
| CN110956630A (en) * | 2019-12-18 | 2020-04-03 | 浙江大学 | Method, device and system for detecting plane printing defects |
Non-Patent Citations (2)
| Title |
|---|
| 吴强等: "基于机器视觉的零件缺陷检测算法", 科学技术创新, no. 26, 15 September 2018 (2018-09-15), pages 65 - 66 * |
| 李正明;黎宏;孙俊;: "基于数字图像处理的印刷电路板缺陷检测", 仪表技术与传感器, no. 08, 15 August 2012 (2012-08-15), pages 87 - 89 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113063353A (en) * | 2021-03-31 | 2021-07-02 | 深圳中科飞测科技股份有限公司 | Coordinate system establishing method, detection device, detection equipment and storage medium |
| CN113643352A (en) * | 2021-08-09 | 2021-11-12 | 贵州电网有限责任公司 | Natural icing on-line monitoring running wire image icing degree evaluation method |
| CN113643352B (en) * | 2021-08-09 | 2024-05-24 | 贵州电网有限责任公司 | Natural icing on-line monitoring running wire image icing degree evaluation method |
| CN114292021A (en) * | 2021-12-30 | 2022-04-08 | 南京春辉科技实业有限公司 | System and method for adjusting preform rod in real time in quartz optical fiber drawing process |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109978839B (en) | Method for detecting wafer low-texture defects | |
| CN112257676B (en) | Pointer type instrument reading method and system and inspection robot | |
| CN112381751A (en) | Online intelligent detection system and method based on image processing algorithm | |
| CN104240264B (en) | The height detection method and device of a kind of moving object | |
| CN111612737B (en) | Artificial board surface flaw detection device and detection method | |
| CN116228780B (en) | Silicon wafer defect detection method and system based on computer vision | |
| CN115082477B (en) | Semiconductor wafer processing quality detection method based on light reflection removing effect | |
| CN114998571B (en) | Image processing and color detection method based on fixed-size markers | |
| CN116071315A (en) | Product visual defect detection method and system based on machine vision | |
| CN114612399A (en) | Picture identification system and method for mobile phone appearance mark | |
| CN117315670B (en) | Water meter reading area detection method based on computer vision | |
| CN114065798A (en) | Visual identification method and device based on machine identification | |
| CN115880683B (en) | Urban waterlogging ponding intelligent water level detection method based on deep learning | |
| CN116758266A (en) | Reading method of pointer type instrument | |
| CN116993654B (en) | Camera module defect detection method, device, equipment, storage medium and product | |
| CN115184362B (en) | Rapid defect detection method based on structured light projection | |
| CN116843618A (en) | Method for detecting shallow apparent appearance defects of metal parts | |
| TW200842339A (en) | Mura detection method and system | |
| CN114882095B (en) | Object height online measurement method based on contour matching | |
| CN111260625B (en) | Automatic extraction method for offset printing large image detection area | |
| CN115546141A (en) | Small sample Mini LED defect detection method and system based on multi-dimensional measurement | |
| CN212646436U (en) | Artificial board surface flaw detection device | |
| CN115471494A (en) | Wo citrus quality inspection method, device, equipment and storage medium based on image processing | |
| CN115205155A (en) | Distorted image correction method and device and terminal equipment | |
| CN115205287A (en) | Gear digital measurement and evaluation cloud method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |