Rectangular chip detection equipment and detection method thereof

Abstract

The invention discloses rectangular chip detection equipment and a detection method of the rectangular chip detection equipment, and the rectangular chip detection equipment adopts automatic control to detect chips, so that the detection efficiency of the rectangular chips is improved. According to the Internet of Things detection method, test data of the rectangular chip are transmitted to the cloud server through the Internet of Things, and production of the rectangular chip can be optimized through powerful cloud computing of the background cloud server. Through the Internet of Things detection method, the rectangular chip can be accurately positioned, and the detection speed is improved. According to the Internet of Things detection method, the depth information of the three-dimensional image of the target is obtained, and the spatial positioning precision is improved. According to the Internet of Things detection method, the complete outer contour of the rectangular chip is extracted on the basis of keeping detail information of the rectangular chip. According to the invention, the rectangular chip can be accurately positioned, and the detection speed is improved. The depth information of the three-dimensional image of the target is acquired, and the spatial positioning precision is improved.

Description

Technical Field

[0001] The present invention relates to a detection device, in particular to a rectangular chip detection device and a detection method of the rectangular chip detection device. Background Art

[0002] In the prior art, rectangular chips need to be inspected, and the appearance quality inspection of rectangular chip packaging mainly includes rectangular chip inclination inspection, printed character inspection, rectangular chip pin size inspection, and pin electrical performance inspection. For the above inspection items, manual inspection was initially used. However, due to the small size of the chip itself, manual inspection has many disadvantages, such as: (1) high labor intensity, easy fatigue of the human eye, missed inspection, false inspection, etc.; (2) inconsistent inspection standards; (3) slow speed and low efficiency; (4) rising labor costs. Therefore, as the cost of equipment gradually decreases, manual inspection obviously no longer meets the needs of industrial production, and the occasions for manual inspection of chips in industrial production are gradually decreasing. It is necessary to further study the visual inspection link of chip packaging and testing. Since it is difficult to objectively detect defects in the semiconductor chip packaging process, it directly affects the efficiency of chip packaging and testing, thereby restricting the development of the semiconductor industry. Therefore, the development of a high-precision and high-efficiency semiconductor chip image visual positioning inspection system is of far-reaching significance for effectively promoting the semiconductor chip packaging and testing industry.

[0003] At present, visual positioning equipment also applies AI intelligent mode, which aggregates data through the Internet of Things and uses AI deep learning to obtain the best data to achieve efficient and high-quality chip production. Summary of the invention

[0004] In view of the deficiencies in the prior art, the technical problem to be solved by the present invention is to provide a rectangular chip detection device and a detection method for the rectangular chip detection device. The purpose of designing the rectangular chip detection device is to improve chip detection efficiency and detection quality. The Internet of Things detection method of the present invention is to transmit the test data of the rectangular chip to the cloud server through the Internet of Things, and then through the backend cloud server cloud computing, the rectangular chip can be optimized for production. The Internet of Things detection method of the present invention can accurately locate the rectangular chip, thereby improving the detection speed.

[0005] In order to solve the above technical problems, the present invention is implemented by the following scheme: A rectangular chip detection device of the present invention includes a frame, the upper end of the frame is a platen, and the platen is set as a feeding area, a detection area and a unloading area from one end to the other end, and the rectangular chip detection device also includes:

[0006] A first X-axis moving mechanism is arranged on one side of the upper surface of the table, the first X-axis moving mechanism has a first X-axis power source and a transport mechanism drivingly connected to the first X-axis power source, the transport mechanism is driven by the first X-axis power source to move between the feeding area, the detection area and the unloading area, and the transport mechanism has a first Z-axis power source and a transport manipulator drivingly connected to the first Z-axis power source;

[0007] A dust removal and orientation detection mechanism provided in the feeding area, the dust removal and orientation detection mechanism comprising an elevated clean air dust removal section, an orientation detection section arranged side by side with the clean air dust removal section, and a first Y-axis moving mechanism passing through the clean air dust removal section and the orientation detection section, wherein a path moved by the first Y-axis moving section of the first Y-axis moving mechanism and a path moved by the transport robot have an intersection in an orthogonal space;

[0008] A double-station detection group arranged in the detection area, any one of which includes a loading platform transfer mechanism and a detection mechanism, the loading platform transfer mechanism has a multi-dimensional driving mechanism, and the trajectory moved by the transfer activity part of the loading platform transfer mechanism and the trajectory moved by the detection activity part of the detection mechanism have an intersection in an orthogonal space;

[0009] The unloading mechanism is arranged in the unloading area, and the unloading mechanism is a conveying line arranged in the Y-axis direction, and the conveying line is used to receive the qualified rectangular chips sent by the handling robot.

[0010] Furthermore, the first X-axis power source is a first X-axis motor, a driving shaft of the first X-axis motor is connected to a first X movable part, the first movable part performs X-axis motion on a long track, and the first Z-axis power source is fixed on its side, the first Z-axis power source is a first Z-axis motor, and its driving shaft is driven and connected to the handling robot.

[0011] Furthermore, the handling robot is a suction device, the suction port of which is set according to the outer shape of the matrix metal contact points of the rectangular chip and the suction head of the suction device is a flexible silicone head.

[0012] Furthermore, the clean air and dust removal part has a first "U"-shaped frame and a clean air blowing device arranged on the first "U"-shaped frame;

[0013] The orientation detection unit comprises a second U-shaped frame, a second X-axis moving mechanism and a first CCD acquisition camera, wherein the second U-shaped frame is provided with a first X-axis track, the power source of the second X-axis moving mechanism is a second X-axis motor, the second X-axis motor is fixed to the second U-shaped frame, and is driven and connected to a second X-axis movable part, the second X-axis movable part is slidably connected to the first X-axis track, and the first CCD acquisition camera is also fixed on the outer side thereof, the lens of the first CCD acquisition camera faces downward and a lever is provided on the edge of the lower shell of the first CCD acquisition camera;

[0014] The power source of the first Y-axis moving mechanism is a first linear motor, a first horizontal rotary motor is installed on the upper end of the first mover of the first linear motor, a drive shaft of the first horizontal rotary motor is upwardly connected with an adsorption plate, and the adsorption plate is the first Y-axis moving part.

[0015] Furthermore, the adsorption plate is a rectangular structural plate having an adsorption hole connected to the negative pressure channel and a plurality of adjustment grooves arranged on the surface of the adsorption plate, and each adjustment groove is provided with a limit block protruding from the notch on the adjustment groove;

[0016] The adsorption surface of the adsorption plate is also provided with two back-facing straight grooves, which are located on both sides of the adsorption hole. After the lever is driven, its lower end can extend into the straight groove and move along the straight groove.

[0017] Furthermore, the material loading platform transfer mechanism is provided with a second linear motor fixed to the table and arranged in the Y-axis direction, an angular rotation drive mechanism installed at the mover end of the second linear motor, a U-shaped bracket provided at the mover end of the second linear motor, a support shell rotatably fixed to the U-shaped bracket, a second horizontal rotary motor installed on the support shell, and an adsorption platform fixed to the drive shaft of the second horizontal rotary motor, and the adsorption platform is the transfer movable part;

[0018] The angular rotation driving mechanism is a rotating motor, a driving shaft of the rotating motor is connected to the supporting shell, and the driving shaft of the rotating motor drives the supporting shell to perform a circular motion.

[0019] Furthermore, the two groups of detection mechanisms share a third U-shaped frame, a second X-axis track is provided on the top of the third U-shaped frame, and second X-axis power sources are installed at both ends of the second X-axis track, and the two groups of second X-axis power sources respectively drive the detection mechanisms connected to the two groups of second X-axis power sources;

[0020] Both detection mechanisms are slidably connected to the second X-axis track, and are provided with a Z-axis driving mechanism and a CCD camera module connected to the Z-axis driving mechanism, and the Z-axis driving mechanism drives the CCD camera module to rise and fall;

[0021] The CCD camera module is provided with two second CCD acquisition cameras, which are arranged at an angle and in an inverted "eight" shape. The shooting directions of the lenses of the two second CCD acquisition cameras can intersect.

[0022] The detection method of the detection device of the present invention is used for spatial positioning of a rectangular chip digital image, comprising:

[0023] A data processing module processes the received data;

[0024] An image acquisition module is connected to a CCD acquisition device, and the image data collected by the module is transmitted to the data processing module. The CCD acquisition device has a dual-camera visual system, and the camera directions of the dual-camera visual system are distributed at an angle;

[0025] An Internet of Things network module, connected to the data processing module, and the image data processed by the data processing module is uploaded to the cloud server through the Internet of Things network module;

[0026] The detection method further includes a digital image coordinate calculation method, which includes the following steps:

[0027] Step 1: Use dual cameras to capture two sets of images of the product at different angles, calibrate the world coordinates between the feature points of the calibration object through the data processing module, extract the feature points on the calibration object in the camera, and use the data processing module to process and calculate the image of the pixel feature points to solve the internal and external parameters of the camera;

[0028] Step 2: obtain the internal and external parameters of the left camera and the right camera through the calibration of the dual-camera vision system, establish the numerical relationship between the image pixel coordinates and the three-dimensional coordinates of the camera, and the relative position and posture of the left camera and the right camera;

[0029] Step three, according to step two, establish a stereo vision matching model and then reconstruct in three dimensions, use the disparity map obtained by stereo matching of the stereo vision matching model, reconstruct the depth map of the matching points in the image in the camera coordinate system according to the binocular vision principle, use a horizontally placed calibration plate as a 3D reconstruction object, calculate the height of each matching point on the calibration plate relative to the camera coordinate system, calculate the 3D reconstruction error, and calculate the average distance between the plane of the horizontally placed calibration plate and the camera coordinate system through the dual-camera vision system;

[0030] Step 4: Process the original image in sequence: ① Binarize the original image to separate the product image from the background; ② Remove unnecessary color features and texture features; ③ Use the closing operation method to remove the jagged edges after binarization; ④ Use the Canny algorithm to detect the edge of the image and extract the edge of the image;

[0031] Step 5: After processing the original image in step 4, calculate the pixel value of the product centroid in the image, extract the outer contour of the product, calculate the centroid using a rectangle-based centroid algorithm, and determine the coordinates of the centroid;

[0032] Step six, correct the position angle of the centroid in step five, use the centroid coordinates as the rotation center of the image, fit a straight line according to the least squares method, determine the angle of the rectangular chip outline image, calculate the inclination angle of the rectangular chip outline image, and correct the angle to make the rectangular chip image form a front view image.

[0033] Furthermore, in step 1, the calculation method for solving the internal and external parameters of the camera includes an internal parameter calculation method and an external parameter calculation method;

[0034] 1) Internal parameter calculation method: Set the X and Y directions to L x , L y , find the distance between the camera and the object, set the distance to L, assume that the focal length of the camera is f, and the number of pixels of the object image in the X and Y directions is N x 、N y ,get:

[0035]

[0036] Where K u , K v They represent the number of pixels represented by the image coordinates in the unit distance in the X and Y directions respectively. L can be calculated based on the pinhole imaging principle.

[0037] According to formula (2.1), we get the following equation:

[0038] N 1x ×L / L x =N 2x ×(Ll) / L x (2.2);

[0039] Here l is the distance that the calibration object steps along the optical axis. Solving L yields:

[0040] L=l×N 2x / (N 2x -N 1x )(2.3);

[0041] Substituting (2.2) into (2.3), we obtain (2.4):

[0042] f=N x N 2x l / L x k u (N 2x -N 1x)(2.4);

[0043] 2) External parameter calculation method: Assume that the model parameters of the two cameras are the same when calibrating the object to be calibrated. When installing the cameras, ensure that the positions of the two cameras are parallel, and ensure that the coordinates of the two cameras in the Y direction are the same. At this time, only the deviation in the X direction needs to be calculated. Assume that there is a point P in the field of view, then the coordinates of the left and right cameras are (X l ,Y l ,Z l )、(X r ,Y r ,Z r ), where Z is the depth coordinate value of point P, f is the focal length of the camera, f1 is the focal length of the left camera, and f r is the focal length of the right camera;

[0044] If the right camera is moved to the origin of the world coordinate system, the left camera and point P will also move while keeping the relative geometric relationship unchanged, where Z is the distance from point P to the small hole. Move the left camera to the origin of the world coordinate system, and the solution is:

[0045]

[0046] The left camera's world coordinate system coincides with the left camera's world coordinate system, and the optical axes of the two cameras are parallel. The rotation matrix of the left camera's world coordinate system is The translation matrix is There is a distance B between the right camera and the left camera in the X direction. The rotation matrix of the right camera and the world coordinates is The translation matrix is Where T xr =B, and the center distance B is obtained by the formula AX=XB.

[0047] Furthermore, in step three, the establishment of the stereoscopic vision matching model includes one of a feature matching model, a phase matching model and a region matching model.

[0048] Furthermore, the region matching model estimates the disparity value in the whole image by scanning similarity calculation in a way of constructing a support window, thereby obtaining a dense disparity map;

[0049] The regional matching model uses the features in the matching area as matching primitives, and finds the best matching point through a local optimization similarity measurement function. The regional stereo matching uses the left image as the reference image and the right image as the matching image. In the reference image, a subwindow W1 of size (2m+1)×(2n+1) is set in its surrounding neighborhood with a certain pixel point as the center. A movable window Wr of the same size is also set in the search image, so that the window searches along the outer pole line within the search range of the matching image. The similarity is calculated using a similarity measurement function, and the point with the highest similarity is set as the matching point, and the pixel grayscale value is used as the matching feature.

[0050] Furthermore, in step 5, a rectangle-based centroid algorithm is used, which is as follows:

[0051] Definition {P n (x)} is about the point set (x i )(i=0,2,...,m): (1)

[0053] P0(x)=1

[0054] P1(x)=(x-α1)P0(x)

[0055] P k+1 (x) = (x-α k+1 ) k (x)-β k P k-1 (x);

[0056] In formula (1): P k (x) is a k-order polynomial with the first coefficient being 1. According to P k (x) Orthogonality constraints, α k+1 , β k The expression is:

[0057]

[0058] A quadratic orthogonal function is fitted to the 4×4 sample points around the center pixel of the patch, and the weight function ω(x i ,y j )=1, (m, n) is the plane coordinate corresponding to the pixel point, and the fitting formula is:

[0059]

[0060] In formula (3): a klx represents the k-order and l-order fitting coefficients in the x direction; a kly represents the fitting coefficient of k-order and l-order in the y direction, and its expression is:

[0061]

[0062] In formula (4): X(x i ,y i ) represents the pixel (x i ,y i ) corresponds to the horizontal coordinate of the space point; Y(x i ,y i ) represents the pixel (x i ,y i ) is the ordinate of the space point corresponding to the chip. After the fitting coefficient is obtained, the centroid pixel of the rectangular chip is substituted into the fitting formula to obtain the three-dimensional space coordinates of the centroid of the patch component.

[0063] Furthermore, in step six, the correction algorithm for calculating the tilt angle of the rectangular chip outline image is as follows:

[0064] Taking the centroid coordinates as the rotation center of the image, the angle of the rectangular chip contour image is determined according to the straight line fitted by the least squares method. The tilt angle α of the rectangular chip and the horizontal offset value x of the rectangular chip are converted into a simple mathematical problem. Let OA be the extension line of the centroid coordinates in the positive direction of the x-axis, pq be the fitted straight line of the two corner points of the longest side by the least squares method, h=(x,y) is the coordinate of the intersection of pq and OA, l is the vertical distance from point p to h, and according to the right triangle relationship marked by the light-colored line in the figure, we have:

[0065] l=p y -h y (1);

[0066] The horizontal offset value x of the pin image can be obtained:

[0067] x=p x -h x (2);

[0068] Then the component outline image tilt angle α is:

[0069]

[0070] Compared with the prior art, the beneficial effects of the present invention are:

[0071] 1. The rectangular chip detection device of the present invention adopts automatic control to detect chips, thereby improving the detection efficiency of rectangular chips.

[0072] 2. The IoT detection method of the present invention transmits the test data of the rectangular chip to the cloud server through the IoT, and then optimizes the production of the rectangular chip through the powerful cloud computing of the backend cloud server. The IoT detection method of the present invention can accurately locate the rectangular chip and improve the detection speed.

[0073] 3. The Internet of Things detection method of the present invention obtains the depth information of the three-dimensional image of the target and improves the spatial positioning accuracy.

[0074] 4. The Internet of Things detection method of the present invention extracts the complete outer contour of the rectangular chip while maintaining the detailed information of the rectangular chip. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 It is a structural diagram of the rectangular chip detection equipment of the present invention.

[0076] Figure 2 It is a structural diagram of the dust removal and orientation detection mechanism of the present invention.

[0077] Figure 3 This is a diagram of the installation structure of the first horizontal rotating motor and the adsorption plate of the present invention.

[0078] Figure 4 It is a structural diagram of the material loading platform transfer mechanism of the present invention.

[0079] Figure 5 Schematic diagram of the dual-camera vision system of the present invention.

[0080] Figure 6 This is a schematic diagram of the calibration principle of the present invention.

[0081] Figure 7 Schematic diagram of the dual-camera parallax principle of the present invention.

[0082] Figure 8 This is a schematic diagram of the calibration plate and sample collection of the present invention.

[0083] Fig. 9 Schematic diagram of the stereo vision matching search process of the present invention.

[0084] Fig.10 It is a schematic diagram of three-dimensional reconstruction of the calibration plate of the present invention.

[0085] Fig.11 This is a schematic diagram of extracting the outer contour of a rectangular chip according to the present invention.

[0086] Fig.12 This is a schematic diagram of the correction angle of the present invention.

[0087] Fig.13 This is a schematic diagram of the rectangular chip extraction results of the present invention.

[0088] Markings in the accompanying drawings: frame 1, unloading mechanism 2, detection mechanism 3, second X-axis track 4, dust removal and orientation detection mechanism 5, first X-axis moving mechanism 6, defective product storage area 7, loading platform transfer mechanism 8, first mover 51, adsorption plate 52, first CCD acquisition camera 53, first linear motor 54, the air purification and dust removal part 55, first horizontal rotating motor 56, adsorption platform 81, angle rotation drive mechanism 82, second horizontal rotating motor 83, second linear motor 84, dual camera 101, rectangular chip 102, cloud server 103, AI device 104, left camera 105, right camera 106, rectangular structural plate 521, adsorption hole 522, limit block 523, straight groove 524, and lever 531. DETAILED DESCRIPTION

[0089] 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, so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making a clearer and more explicit definition of the protection scope of the present invention. Obviously, the embodiments described in the present invention are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0090] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0091] Embodiment 1:

[0092] Please refer to the attached Figure 1-4 The rectangular chip detection device of the present invention comprises a frame 1, the upper end of the frame is a platform, and the platform is set as a feeding area, a detection area and a discharge area from one end to the other end. The feeding area is used to receive rectangular chips from external equipment, and then identify qualified products and unqualified products after detection in the detection area, and finally transport qualified products to the qualified product area through the discharge area. The platform is covered with a box cover.

[0093] The rectangular chip detection equipment also includes a first X-axis moving mechanism 6, a dust removal and orientation detection mechanism 5, a double-station detection group and a feeding mechanism 2.

[0094] The working principle of the first X-axis moving mechanism 6 is as follows:

[0095] The first X-axis moving mechanism 6 is arranged on one side of the upper surface of the table, and the first X-axis moving mechanism 6 has a first X-axis power source and a transport mechanism connected to the first X-axis power source. The transport mechanism is driven by the first X-axis power source to move between the feeding area, the detection area and the unloading area. The transport mechanism has a first Z-axis power source and a transport manipulator connected to the first Z-axis power source. The first X-axis power source is a first X-axis motor, and the drive shaft of the first X-axis motor is connected to a first X movable part. The first movable part moves in the X-axis on a long track, and the first Z-axis power source is fixed on its side. The first Z-axis power source is a first Z-axis motor, and its drive shaft is connected to the transport manipulator. The transport manipulator is an adsorption device, and its suction port is set according to the outer shape of the matrix metal contact point of the rectangular chip 102, and the adsorption head of the adsorption device is a flexible silicone head. Since the metal surface of the rectangular chip 102 is the metal contact point of the lattice structure, the periphery of the metal contact point is also a rectangular plane, and the flexible silicone head is set to a rectangular opening and connected to the negative pressure system. After the negative pressure system works, the flexible silicone head can adsorb the front side of the rectangular chip 102. The rectangular chip 102 also needs to be inspected on the back side, and a flip device is set on the inspection device of the present invention to complete the inspection of the back side of the rectangular chip 102. The handling robot goes back and forth between the feeding area, the inspection area, and the unloading area.

[0096] Structure and working principle of dust removal and orientation detection mechanism 5:

[0097] The dust removal and orientation detection mechanism 5 is arranged in the feeding area, and the dust removal and orientation detection mechanism 5 has an elevated clean air dust removal part 55, an orientation detection part arranged side by side with the clean air dust removal part 55, and a first Y-axis moving mechanism passing through the clean air dust removal part 55 and the orientation detection part, and the path moved by the first Y-axis moving part of the first Y-axis moving mechanism and the path moved by the transport robot have an intersection in an orthogonal space; the clean air dust removal part 55 has a first "U"-shaped frame and a clean air blowing device arranged on the first "U"-shaped frame. The orientation detection unit comprises a second U-shaped frame, a second X-axis moving mechanism and a first CCD acquisition camera 53. The second U-shaped frame is provided with a first X-axis track. The power source of the second X-axis moving mechanism is a second X-axis motor. The second X-axis motor is fixed to the second U-shaped frame. The second X-axis motor is connected to the driving part of the second X-axis movable part. The second X-axis movable part is slidably connected to the first X-axis track. The first CCD acquisition camera 53 is also fixed on the outer side of the second X-axis movable part. The lens of the first CCD acquisition camera 53 faces downward and a lever 531 is provided on the edge of the lower shell of the first CCD acquisition camera 53. The power source of the first Y-axis moving mechanism is a first linear motor 54. The upper end of the first mover 51 of the first linear motor 54 is provided with a first horizontal rotation motor 56. The driving shaft of the first horizontal rotation motor 56 is connected with an adsorption plate 52 upward. The adsorption plate 52 is the first Y-axis moving part.

[0098] The adsorption plate 52 is a rectangular structural plate 521, which has an adsorption hole 522 connected to the negative pressure channel, and multiple adjustment grooves provided on the surface of the adsorption plate 52, each of which is provided with a stopper 523 protruding from the notch on the adjustment groove. The adsorption surface of the adsorption plate 52 is also provided with two back-facing straight grooves 524, which are located on both sides of the adsorption hole 522. After the lever 531 is driven, its lower end can extend into the straight groove 524 and can move along the straight groove 524.

[0099] The adsorption plate 52 is arranged horizontally, and moves synchronously in the Y-axis direction with the first horizontal rotary motor 56 and the first mover 51. The first mover 51 is driven to the loading point by the stator of the first linear motor 54. The loading point is Figure 1 The external manipulator delivers the processed rectangular chip 102 to the adsorption plate 52. The adsorption plate 52 is connected to the negative pressure system. After the negative pressure system is working, the adsorption plate 52 can adsorb the printed surface of the rectangular chip 102. After being driven, the adsorption plate 52 passes through the air outlet position of the air purification and dust removal unit 55 and is blown and dusted by the air purification and dust removal unit 55, and then reaches the bottom of the orientation detection unit.

[0100] The first CCD acquisition camera 53 on the orientation detection unit acquires the image of the rectangular chip 102 and sends the image information to the image data processing module to determine whether the position of the rectangular chip 102 on the adsorption plate 52 is correct. When the position of the adsorption plate 52 is offset, the first Y-axis moving mechanism moves the lever 531 along the straight groove 524 to move the rectangular chip 102 so that the two adjacent sides of the rectangular chip 102 are against the limit block 523, which is convenient for the handling robot to grab. Finally, the adsorption plate 52 and the rectangular chip 102 thereon are moved to a position where they can be grabbed by the handling robot.

[0101] The transport robot takes out the rectangular chip 102 after the orientation adjustment from the adsorption plate 52 and transfers it to the adsorption platform 81 on the loading platform transfer mechanism 8 .

[0102] The following is the structure and working principle of the dual-station detection group:

[0103] The double-station detection group is arranged in the detection area, and any one of the detection groups includes a loading platform transfer mechanism 8 and a detection mechanism 3. The loading platform transfer mechanism 8 has a multi-dimensional driving mechanism, and the trajectory of the transfer activity of the loading platform transfer mechanism 8 and the trajectory of the detection activity of the detection mechanism 3 have an intersection in the orthogonal space. The loading platform transfer mechanism 8 is provided with a second linear motor 84 fixed to the table and arranged in the Y-axis direction, an angle rotation drive mechanism 82 installed at the moving end of the second linear motor 84, a U-shaped bracket arranged at the moving end of the second linear motor 84, a support shell rotatably fixed to the U-shaped bracket, a second horizontal rotation motor 83 installed on the support shell, and an adsorption platform 81 fixed to the driving shaft of the second horizontal rotation motor 83, wherein the adsorption platform 81 is the transfer activity; the angle rotation drive mechanism 82 is a rotary motor, the driving shaft of which is connected to the support shell and the driving shaft of which drives the support shell to perform circular motion. The two detection mechanisms 3 share a third U-shaped frame, the top of which is provided with a second X-axis track 4, and both ends of the second X-axis track 4 are provided with a second X-axis power source. The second X-axis power source is a motor.

[0104] The two sets of second X-axis power sources respectively drive the detection mechanism 3 connected to the two sets of second X-axis power sources; the two sets of detection mechanisms 3 are slidably connected to the second X-axis track 4, which is provided with a Z-axis driving mechanism and a CCD camera module connected to the Z-axis driving mechanism, and the Z-axis driving mechanism drives the CCD camera module to rise and fall; the CCD camera module is provided with two second CCD acquisition cameras 101, the two second CCD acquisition cameras 101 are arranged at an angle and in an inverted "eight" structure, and the lens shooting directions of the two second CCD acquisition cameras 101 can intersect. The detection method of the CCD camera module is shown in Example 2.

[0105] After the adsorption platform 81 receives the rectangular chip 102, it is fixed to the adsorption platform 81 by adsorption through the negative pressure system. The adsorption platform 81 is driven by the multi-dimensional driving mechanism of the loading platform transfer mechanism 8 to reach the detection point, which is located at the intersection of the lenses of the two second CCD acquisition cameras 101. The two second CCD acquisition cameras 101 are driven by the second X-axis power source and the Z-axis driving mechanism so that the second CCD acquisition cameras 101 can reach the optimal shooting position.

[0106] The table is also provided with two areas of defective product storage areas 7, which are located on the adjacent sides of the material loading platform transfer mechanism 8, respectively, for temporarily storing defective rectangular chips 102. After the second CCD acquisition camera 101 detects the rectangular chip 102, it determines whether the tested product is qualified. The qualified rectangular chip 102 is sent to the handling robot for grabbing and sent to the unloading area, and the defective product is sent to the handling robot and sent to the defective product storage area 7.

[0107] The unloading mechanism 2 is arranged in the unloading area. The unloading mechanism 2 is a conveying line arranged in the Y-axis direction. The conveying line is used to receive the qualified rectangular chips 102 sent by the handling robot.

[0108] Embodiment 2:

[0109] The next step is to set up a printing detection device for detecting the other side of the rectangular chip 102. The printing detection device has the same structure as the rectangular chip detection device of the first embodiment.

[0110] The specific structure of the present invention is as follows:

[0111] Please refer to the attached Figure 5-13 , a detection method of a rectangular chip detection device of the present invention, the detection method is used for spatial positioning of a digital image of a rectangular chip 102, comprising:

[0112] A data processing module processes the received data;

[0113] An image acquisition module is connected to a CCD acquisition device, and the image data collected by the module is transmitted to the data processing module. The CCD acquisition device has a dual-camera visual system, and the camera directions of the dual-camera visual system are distributed at an angle;

[0114] An Internet of Things network module, connected to the data processing module, and the image data processed by the data processing module is uploaded to the cloud server 103 through the Internet of Things network module;

[0115] The detection method further includes a digital image coordinate calculation method, which includes the following steps:

[0116] Step 1: Use the dual camera 101 to capture two sets of images of the product at different angles, calibrate the world coordinates between the feature points of the calibration object through the data processing module, extract the feature points on the calibration object in the camera, and use the data processing module to process and calculate the image of the pixel feature points to solve the internal and external parameters of the camera;

[0117] Step 2: obtain the internal and external parameters of the left camera 105 and the right camera 106 through the calibration of the dual-camera vision system, establish the numerical relationship between the image pixel coordinates and the three-dimensional coordinates of the camera, and the relative position posture of the left camera 105 and the right camera 106; the installation distance of the left and right cameras, that is, the size of the baseline B. When the baseline B is specified, the measurement error of the positioning measurement system based on binocular vision will increase sharply with the increase of the object distance z; and when the value of the baseline B is within 0.8z~2.2z, the change of its measurement error is small;

[0118] Step three, according to step two, establish a stereo vision matching model and then reconstruct in three dimensions, use the disparity map obtained by stereo matching of the stereo vision matching model, reconstruct the depth map of the matching points in the image in the camera coordinate system according to the binocular vision principle, use a horizontally placed calibration plate as a 3D reconstruction object, calculate the height of each matching point on the calibration plate relative to the camera coordinate system, calculate the 3D reconstruction error, and calculate the average distance between the plane of the horizontally placed calibration plate and the camera coordinate system through the dual-camera vision system;

[0119] Step 4: Process the original image in sequence: ① Binarize the original image to separate the product image from the background; ② Remove unnecessary color features and texture features; ③ Use the closing operation method to remove the jagged edges after binarization; ④ Use the Canny algorithm to detect the edge of the image and extract the edge of the image;

[0120] Step 5: After processing the original image in step 4, calculate the pixel value of the product centroid in the image, extract the outer contour of the product, calculate the centroid using a rectangle-based centroid algorithm, and determine the coordinates of the centroid;

[0121] Step six, correct the position angle of the centroid in step five, use the centroid coordinates as the rotation center of the image, fit a straight line according to the least squares method, determine the angle of the rectangular chip 102 contour image, calculate the inclination angle of the rectangular chip 102 contour image, and correct the angle so that the rectangular chip 102 image forms a front view image.

[0122] A preferred technical solution of this embodiment: in step 1, the calculation method for solving the internal and external parameters of the camera includes an internal parameter calculation method and an external parameter calculation method;

[0123] 1) If Figure 2As shown, the internal parameter calculation method is: set one X and Y direction as L x , L y , find the distance between the camera and the object, set the distance to L, assume that the focal length of the camera is f, and the number of pixels of the object image in the X and Y directions is N x 、N y ,get:

[0124]

[0125] Where K u , K v They represent the number of pixels represented by the image coordinates in the unit distance in the X and Y directions respectively. L can be calculated based on the pinhole imaging principle.

[0126] According to formula (2.1), we get the following equation:

[0127] N 1x ×L / L x =N 2x ×(Ll) / L x (2.2);

[0128] Here l is the distance that the calibration object steps along the optical axis. Solving L yields:

[0129] L=l×N 2x / (N 2x -N 1x )(2.3);

[0130] Substituting (2.2) into (2.3), we obtain (2.4):

[0131] f=N x N 2x l / L x k u (N 2x -N 1x )(2.4);

[0132] 2) If Figure 3 As shown in FIG. 1 , the external parameter calculation method is as follows: Assume that the model parameters of the two cameras are the same when calibrating the object to be photographed. When installing the cameras, ensure that the positions of the two cameras are parallel, and ensure that the coordinates of the two cameras in the Y direction are the same. At this time, only the deviation in the X direction needs to be calculated. Assume that there is a point P in the field of view, then the coordinates of the left and right cameras 106 are (X l ,Y l ,Z l )、(X r ,Y r ,Z r), where Z is the depth coordinate value of point P, f is the focal length of the camera, f1 is the focal length of the left camera 105, and f r is the focal length of the right camera 106;

[0133] When the right camera 106 is moved to the origin of the world coordinate system, the left camera 105 and point P will also move with it while keeping the relative geometric relationship unchanged, where Z is the distance from point P to the small hole. When the left camera 105 is moved to the origin of the world coordinate system, the solution is:

[0134]

[0135] The left camera 105 has a world coordinate system that coincides with the left camera 105, and the optical axes of the two cameras are parallel. The rotation matrix of the left camera 105 in the world coordinate system is The translation matrix is There is a distance B between the right camera 106 and the left camera 105 in the X direction. The rotation matrix of the right camera and the world coordinates is The translation matrix is Where T xr =B, and the center distance B is obtained by the formula AX=XB.

[0136] The calibration of the dual-camera vision system obtains the internal and external parameters of the left and right cameras, establishes the numerical relationship between the image pixel coordinates and the three-dimensional coordinates of the cameras, and the relative position and posture between the two cameras. The camera internal parameters include focal length, distortion coefficient, pixel size and center coordinates, and the external parameters include the rotation matrix and translation vector of the camera coordinate system relative to the world coordinate system. The calibration plate used in the present invention is as follows: Figure 4 As shown in the figure, the size is 72×72mm, and the image array is 7×7. The calibration process requires multiple sets of calibration plate images in different postures, extracting their feature points, and fitting them. Binarization, edge detection, extracted contours, and finally the centroid pixel value of each solid rectangle are obtained for the acquired fitting plate image. The position information of the solid rectangle in the image and in the plane is the fitting sample point, which is used for the subsequent local fitting of the orthogonal function.

[0137] A preferred technical solution of this embodiment: in step three, the establishment of a stereoscopic vision matching model includes one of a feature matching model, a phase matching model and a region matching model.

[0138] A preferred technical solution of this embodiment: the region matching model estimates the disparity value by scanning similarity calculation in the whole image in a way of constructing a support window to obtain a dense disparity map;

[0139] The regional matching model uses the features in the matching area as matching primitives, and finds the best matching point through a local optimization similarity measurement function. The regional stereo matching uses the left image as the reference image and the right image as the matching image. In the reference image, a subwindow W1 of size (2m+1)×(2n+1) is set in its surrounding neighborhood with a certain pixel point as the center. A movable window Wr of the same size is also set in the search image, so that the window searches along the outer pole line within the search range of the matching image. The similarity is calculated using a similarity measurement function, and the point with the highest similarity is set as the matching point, and the pixel grayscale value is used as the matching feature.

[0140] like Figure 5 As shown, the present invention selects the regional stereo matching algorithm in local stereo matching. Regional stereo matching uses the features in the matching area as matching primitives and finds the best matching point through the local optimization similarity measurement function. Regional stereo matching uses the left image as the reference image and the right image as the matching image. In the reference image, a sub-window W1 of size (2m+1)*(2n+1) is set in the surrounding neighborhood with a certain pixel point as the center. A movable window Wr of the same size is also set in the search image, so that the window searches along the outer pole line within the search range of the matching image. The similarity is calculated using the similarity measurement function, and the point with the highest similarity is the matching point. Pixel grayscale values ​​are usually used as matching features.

[0141] According to the basic principle of dual-camera vision system, the disparity of a certain point is related to its depth, and the size of the disparity value is inversely proportional to the depth. Therefore, it is not necessary to calculate along the epipolar line in sequence. Usually, the depth range of the patch relative to the binocular vision system is obtained by estimated measurement, and the approximate disparity range d∈[dmin, dmax] of the matching point in the image is obtained. The length of the disparity search area can be within this range, which reduces the amount of calculation in the matching process and improves efficiency. In addition to improving efficiency by narrowing the search range, the image pyramid can also be used to search for matching points from coarse to fine to improve search efficiency.

[0142] Using the disparity map obtained by stereo matching, the depth map of the matching points in the image in the camera coordinate system can be reconstructed according to the principle of binocular vision. This paper first uses a horizontally placed calibration plate as the 3D reconstruction object, and calculates the 3D reconstruction error by calculating the height of each matching point on the calibration plate relative to the camera coordinate system. According to the calculation, the average distance between the plane of the horizontally placed calibration plate and the camera coordinate system calculated by the binocular vision system. Based on the average distance, the error of the 3D reconstruction is within the allowable range. The 3D reconstruction result is as follows: Figure 6 shown.

[0143] In the process of collecting, digitizing and transmitting rectangular chip images, due to the surrounding environment, equipment performance and other reasons, the original image may have problems such as insufficient contrast, blur, and noise, so the image needs to be preprocessed. The histogram equalization method is used to enhance the image contrast, which is convenient for subsequent edge detection, feature extraction and other operations. The Gaussian filtering method can better maintain the edge characteristics of the part image and retain more relevant details. The maximum inter-class variance method (also known as the Otsu method) is used for binarization, which can separate the image from the background and remove unnecessary color features and texture features. The closing operation is used to remove the jagged edges and unnecessary details after binarization. The Canny operator is used to detect the edge of the image and extract the edge of the image, which is convenient for patch feature extraction and positioning. After preprocessing the image, the pixel value of the patch centroid in the image is calculated. The outer contour of the patch is extracted, and the centroid is calculated using a rectangle-based centroid algorithm.

[0144] A preferred technical solution of this embodiment: In step 5, a rectangle-based centroid algorithm is used, and the algorithm is as follows:

[0145] Definition {P n (x)} is about the point set (x i )(i=0,2,...,m): (1)

[0147] P0(x)=1

[0148] P1(x)=(x-α1)P0(x)

[0149] P k+1 (x) = (x-α k+1 ) k (x)-β k P k-1 (x);

[0150] In formula (1): P k (x) is a k-order polynomial with the first coefficient being 1. According to P k (x) Orthogonality constraints, α k+1 , β k The expression is:

[0151]

[0152]

[0153] A quadratic orthogonal function is fitted to the 4×4 sample points around the center pixel of the patch, and the weight function ω(x i ,y j )=1, (m, n) is the plane coordinate corresponding to the pixel point, and the fitting formula is:

[0154]

[0155] In formula (3): a klx represents the k-order and l-order fitting coefficients in the x direction; a kly represents the fitting coefficient of k-order and l-order in the y direction, and its expression is:

[0156]

[0157] In formula (4): X(x i ,y i ) represents the pixel (x i ,y i ) corresponds to the horizontal coordinate of the space point; Y(x i ,y i ) represents the pixel (x i ,y i ) corresponds to the ordinate of the spatial point. After the fitting coefficient is obtained, the centroid pixel of the rectangular chip 102 is substituted into the fitting formula to obtain the three-dimensional spatial coordinates of the centroid of the patch element.

[0158] A preferred technical solution of this embodiment: Figure 7 As shown, in step six, the correction algorithm for calculating the tilt angle of the rectangular chip 102 contour image is as follows:

[0159] With the centroid coordinates as the rotation center of the image, the angle of the rectangular chip 102 contour image is determined according to the straight line fitted by the least squares method. The tilt angle α of the rectangular chip 102 and the horizontal offset value x of the rectangular chip 102 are converted into a simple mathematical problem. Let OA be the extension line of the centroid coordinates in the positive direction of the x-axis, pq be the fitted straight line of the two corner points of the longest side by the least squares method, h = (x, y) is the coordinate of the intersection of pq and OA, l is the vertical distance from point p to h, and according to the right triangle relationship marked by the light-colored line in the figure, we have:

[0160] l=p y -h y (1);

[0161] The horizontal offset value x of the pin image can be obtained:

[0162] x=p x -h x (2);

[0163] Then the component outline image tilt angle α is:

[0164]

[0165] In summary, the rectangular chip detection equipment of the present invention uses automated control to detect chips, which improves the detection efficiency of rectangular chips. The Internet of Things detection method of the present invention transmits the test data of the rectangular chip to the cloud server through the Internet of Things, and then optimizes the production of rectangular chips through the powerful cloud computing of the background cloud server. The Internet of Things detection method of the present invention can accurately locate the rectangular chip and improve the detection speed. The Internet of Things detection method of the present invention obtains the depth information of the three-dimensional image of the target and improves the spatial positioning accuracy. The Internet of Things detection method of the present invention extracts the complete outer contour of the rectangular chip on the basis of maintaining the detailed information of the rectangular chip.

[0166] The detection method of the rectangular chip detection device of the present invention first calibrates the dual-camera visual system, pre-processes the collected fitting plate image to obtain the solid rectangle center as the fitting sample point; the second step is to reconstruct the depth map of the matching point in the image in the camera coordinate system according to the dual-camera visual matching search method, and use the calibration plate as the three-dimensional reconstruction object to calculate the three-dimensional reconstruction error; the third step is to collect and select the chip patch as the positioning element, place it on the plane to be measured, and use the binocular camera to collect the image to obtain the collected left image and right image. Perform image preprocessing and center calculation to obtain the pixel value of the center of the patch in the left image and the right image. The fourth step is to select N×N sample points around the center pixel value of the patch to perform local least squares fitting of the orthogonal function, obtain the fitting coefficient of the area, and substitute the center pixel into the fitting formula, perform binocular reconstruction to obtain the spatial coordinates of the center of the rectangular patch, use the centroid coordinates as the rotation center of the image and the straight line fitted by the least squares method, and calculate the angle of correction of the chip contour image according to mathematical calculation.

[0167] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A rectangular chip detection device, comprising a frame (1), the upper end of the frame is a platform, and the platform is set as a feeding area, a detection area and a discharge area from one end to the other end, It is characterized in that The rectangular chip detection device also includes: A first X-axis moving mechanism (6) is arranged on one side of the upper surface of the table, the first X-axis moving mechanism (6) comprises a first X-axis power source and a transport mechanism drivingly connected to the first X-axis power source, the transport mechanism is driven by the first X-axis power source to move between the feeding area, the detection area and the unloading area, the transport mechanism comprises a first Z-axis power source and a transport robot drivingly connected to the first Z-axis power source; a dust removal and orientation detection mechanism (5) disposed in the feeding area, the dust removal and orientation detection mechanism (5) comprising an elevated clean air dust removal section (55), an orientation detection section disposed side by side with the clean air dust removal section (55), and a first Y-axis moving mechanism passing through the clean air dust removal section (55) and the orientation detection section, wherein a path moved by the first Y-axis moving section of the first Y-axis moving mechanism and a path moved by the transport robot have an intersection in an orthogonal space; A double-station detection group is arranged in the detection area, any one of which includes a loading platform transfer mechanism (8) and a detection mechanism (3), the loading platform transfer mechanism (8) having a multi-dimensional driving mechanism, and the trajectory of the transfer activity of the loading platform transfer mechanism (8) and the trajectory of the detection activity of the detection mechanism (3) intersecting in an orthogonal space; A material unloading mechanism (2) is arranged in the material unloading area, the material unloading mechanism (2) is a conveying line arranged in the Y-axis direction, and the conveying line is used to receive qualified rectangular chips delivered by the handling robot; The clean air dust removal section (55) comprises a first U-shaped frame and a clean air blowing device arranged on the first U-shaped frame; The orientation detection unit comprises a second "U"-shaped frame, a second X-axis moving mechanism and a first CCD acquisition camera (53); the second "U"-shaped frame is provided with a first X-axis track; the power source of the second X-axis moving mechanism is a second X-axis motor; the second X-axis motor is fixed to the second "U"-shaped frame and is drivingly connected to a second X-axis movable part; the second X-axis movable part is slidably connected to the first X-axis track; the first CCD acquisition camera (53) is also fixed on the outer side of the second X-axis movable part; the lens of the first CCD acquisition camera (53) faces downward and a lever (531) is provided on the edge of the lower shell of the first CCD acquisition camera (53); The power source of the first Y-axis moving mechanism is a first linear motor (54), a first horizontal rotary motor (56) is installed at the upper end of a first mover (51) of the first linear motor (54), a driving shaft of the first horizontal rotary motor (56) is connected upward to an adsorption plate (52), and the adsorption plate (52) is the first Y-axis moving part; The adsorption plate (52) is a rectangular structural plate, which has an adsorption hole (522) connected to the negative pressure channel, and a plurality of adjustment grooves arranged on the surface of the adsorption plate (52), and each adjustment groove is provided with a limit block (523) protruding from the notch on the adjustment groove; The adsorption surface of the adsorption plate (52) is also provided with two back-facing straight grooves (524), the two straight grooves (524) are located on both sides of the adsorption hole (522), and after the lever (531) is driven, its lower end can extend into the straight groove (524) and can move along the straight groove (524); The material loading platform transfer mechanism (8) is provided with a second linear motor (84) fixed to the table and arranged in the Y-axis direction, an angular rotation drive mechanism (82) installed at the moving end of the second linear motor (84), a U-shaped bracket arranged at the moving end of the second linear motor (84), a support shell rotatably fixed to the U-shaped bracket, a second horizontal rotation motor (83) installed on the support shell, and an adsorption platform (81) fixed to the driving shaft of the second horizontal rotation motor (83), wherein the adsorption platform (81) is the transfer movable part; The angular rotation driving mechanism (82) is a rotating motor, the driving shaft of which is connected to the supporting shell and drives the supporting shell to perform a circular motion; The two groups of detection mechanisms (3) share a third U-shaped frame, a second X-axis track (4) is provided on the top of the third U-shaped frame, and second X-axis power sources are installed at both ends of the second X-axis track (4), and the two groups of second X-axis power sources respectively drive the detection mechanisms (3) connected to the two groups of second X-axis power sources; Both detection mechanisms (3) are slidably connected to the second X-axis track (4), and are provided with a Z-axis driving mechanism and a CCD camera module connected to the Z-axis driving mechanism, and the Z-axis driving mechanism drives the CCD camera module to rise and fall; The CCD camera module is provided with two second CCD acquisition cameras (101), the two second CCD acquisition cameras (101) are arranged at an angle and in an inverted "eight" shape, and the shooting directions of the lenses of the two second CCD acquisition cameras (101) can intersect.

2. A rectangular chip detection device according to claim 1, It is characterized in that The first X-axis power source is a first X-axis motor, and the driving shaft of the first X-axis motor is connected to a first X movable part. The first movable part performs X-axis motion on a long track, and the first Z-axis power source is fixed on its side. The first Z-axis power source is a first Z-axis motor, and its driving shaft is drivingly connected to the handling robot.

3. A rectangular chip detection device according to claim 2, It is characterized in that The handling robot is a suction device, the suction port of which is arranged according to the peripheral shape of the matrix metal contact points of the rectangular chip, and the suction head of the suction device is a flexible silicone head.

4. A detection method for a detection device, comprising the rectangular chip detection device according to any one of claims 1 to 3.

5. The detection method according to claim 4 is used for spatial positioning of a digital image of a rectangular chip, It is characterized in that include: A data processing module processes the received data; An image acquisition module is connected to a CCD acquisition device, and the image data collected by the module is transmitted to the data processing module. The CCD acquisition device has a dual-camera visual system, and the camera directions of the dual-camera visual system are distributed at an angle; An Internet of Things network module, connected to the data processing module, and the image data processed by the data processing module is uploaded to the cloud server (103) through the Internet of Things network module; The detection method further includes a digital image coordinate calculation method, which includes the following steps: Step 1: using a dual camera (101) to capture two sets of images of a product at different angles, calibrating the world coordinates between the characteristic points of the captured calibration object through a data processing module, extracting characteristic points on the calibration object in the camera, and performing image processing and calculation of the pixel characteristic points through the data processing module to solve the internal and external parameters of the camera; Step 2: obtain the internal and external parameters of the left camera and the right camera through the calibration of the dual-camera vision system, establish the numerical relationship between the image pixel coordinates and the three-dimensional coordinates of the camera, and the relative position and posture of the left camera and the right camera; Step three, according to step two, establish a stereo vision matching model and then reconstruct in three dimensions, use the disparity map obtained by stereo matching of the stereo vision matching model, reconstruct the depth map of the matching points in the image in the camera coordinate system according to the binocular vision principle, use a horizontally placed calibration plate as a 3D reconstruction object, calculate the height of each matching point on the calibration plate relative to the camera coordinate system, calculate the 3D reconstruction error, and calculate the average distance between the plane of the horizontally placed calibration plate and the camera coordinate system through the dual-camera vision system; Step 4: Process the original image in sequence: ① Binarize the original image to separate the product image from the background; ② Remove unnecessary color features and texture features; ③ Use the closing operation method to remove the jagged edges after binarization; ④ Use the Canny algorithm to detect the edge of the image and extract the edge of the image; Step 5: After processing the original image in step 4, calculate the pixel value of the product centroid in the image, extract the outer contour of the product, calculate the centroid using a rectangle-based centroid algorithm, and determine the coordinates of the centroid; Step 6, correcting the position angle of the centroid in step 5, taking the centroid coordinates as the rotation center of the image, fitting a straight line according to the least squares method, determining the angle of the rectangular chip outline image, calculating the tilt angle of the rectangular chip outline image, and correcting the angle so that the rectangular chip image forms a front view image; In step 1, the calculation method for solving the internal and external parameters of the camera includes an internal parameter calculation method and an external parameter calculation method; 1) Internal parameter calculation method: Set the X and Y directions to L x , L y , find the distance between the camera and the object, set the distance to L, assume that the focal length of the camera is f, and the number of pixels of the object image in the X and Y directions is N x 、N y ,get: Where K u , K v They represent the number of pixels represented by the image coordinates in the unit distance in the X and Y directions respectively. L can be calculated based on the pinhole imaging principle. According to formula (2.1), we get the following equation: N 1x ×L / L x =N 2x ×(Ll) / L x (2.2); Here l is the distance that the calibration object steps along the optical axis. Solving L yields: L=l×N 2x / (N 2x -N 1x )(2.3); Substituting (2.2) into (2.3), we obtain (2.4): f=N x N 2x l / L x k u (N 2x -N 1x )(2.4); 2) External parameter calculation method: Assume that the model parameters of the two cameras are the same when calibrating the object to be calibrated. When installing the cameras, ensure that the positions of the two cameras are parallel, and ensure that the coordinates of the two cameras in the Y direction are the same. At this time, only the deviation in the X direction needs to be calculated. Assume that there is a point P in the field of view, then the coordinates of the left and right cameras are (X l ,Y l ,Z l )、(X r ,Y r ,Z r ), where Z is the depth coordinate value of point P, f is the focal length of the camera, and f 1 is the focal length of the left camera, f r is the focal length of the right camera; If the right camera is moved to the origin of the world coordinate system, the left camera and point P will also move while keeping the relative geometric relationship unchanged, where Z is the distance from point P to the small hole. Move the left camera to the origin of the world coordinate system, and the solution is: The left camera's world coordinate system coincides with the left camera's world coordinate system, and the optical axes of the two cameras are parallel. The rotation matrix of the left camera's world coordinate system is The translation matrix is There is a distance B between the right camera and the left camera in the X direction. The rotation matrix of the right camera and the world coordinates is The translation matrix is Where T xr =B, the center distance B is obtained by the formula AX=XB; In step three, the establishment of a stereoscopic vision matching model includes one of a feature matching model, a phase matching model and a region matching model.

6. The detection method according to claim 5, It is characterized in that The region matching model estimates the disparity value in the whole image by scanning similarity calculation in a way of constructing a support window to obtain a dense disparity map; The regional matching model uses the features in the matching area as matching primitives, and finds the best matching point through a local optimization similarity measurement function. The regional stereo matching uses the left image as the reference image and the right image as the matching image. In the reference image, a subwindow W1 of size (2m+1)×(2n+1) is set in its surrounding neighborhood with a certain pixel point as the center. A movable window Wr of the same size is also set in the search image, so that the window searches along the outer pole line within the search range of the matching image. The similarity is calculated using a similarity measurement function, and the point with the highest similarity is set as the matching point, and the pixel grayscale value is used as the matching feature.

7. The detection method according to claim 5, It is characterized in that In step 5, a rectangle-based centroid algorithm is used, which is as follows: Definition {P n (x)} is about the point set (x i )(i=0,2,...,m): (1) P 0 (x)=1 P 1 (x)=(x-α 1 )P 0 (x) P k+1 (x)=(x-α k+1 )P k (x)-β k P k-1 (x); In formula (1): P k (x) is a k-order polynomial with the first coefficient being 1. According to P k (x) Orthogonality constraints, α k+1 , β k The expression is: (2) A quadratic orthogonal function is fitted to the 4×4 sample points around the center pixel of the patch, and the weight function ω(x i ,y j )=1, (m, n) is the plane coordinate corresponding to the pixel point, and the fitting formula is: (3) In formula (3): a klx represents the k-order and l-order fitting coefficients in the x direction; a kly represents the fitting coefficient of k-order and l-order in the y direction, and its expression is: (4) In formula (4): X(x i ,y i ) represents the pixel (x i ,y i ) corresponds to the horizontal coordinate of the space point; Y(x i ,y i ) represents the pixel (x i ,y i ) is the ordinate of the space point corresponding to the chip. After the fitting coefficient is obtained, the centroid pixel of the rectangular chip is substituted into the fitting formula to obtain the three-dimensional space coordinates of the centroid of the patch component.

8. The detection method according to claim 5, It is characterized in that In step 6, the correction algorithm for calculating the tilt angle of the rectangular chip outline image is as follows: Taking the centroid coordinates as the rotation center of the image, the angle of the rectangular chip contour image is determined according to the straight line fitted by the least squares method. The tilt angle α of the rectangular chip and the horizontal offset value x of the rectangular chip are converted into a simple mathematical problem. Let OA be the extension line of the centroid coordinates in the positive direction of the x-axis, pq be the fitted straight line of the two corner points of the longest side by the least squares method, h=(x,y) is the coordinate of the intersection of pq and OA, l is the vertical distance from point p to h, and according to the right triangle relationship marked by the light-colored line in the figure, we have: l=p y -h y (1); The horizontal offset value x of the pin image can be obtained: x=p x -h x (2); Then the component outline image tilt angle α is:

Images