A container inspection system and method

Abstract

The application provides a container detection system and method, which comprises a scanning module supported by a sliding block arranged on the left and right sides and the top of a container detection gate, and the container is scanned on the box surface during container detection to obtain a real scene picture, a point cloud file and container information, the container box surface is identified for defects through the real scene picture, the point cloud of each box surface is segmented in a preset unit period, and the point cloud file of the segmented area is compared with a non-defective unit reference model to obtain a defect result of each box surface, and a data preprocessing unit outputs a detection result according to one or more of the defect identification result, the defect result and the container information.

Description

Technical Field

[0001] This invention relates to the field of container inspection technology, and in particular to a container inspection system and method. Background Technology

[0002] Currently, the annual demand in the container industry is between 3 and 3.5 million TEUs, and the total volume is increasing year by year. During long-term transportation and turnover, containers are more or less subjected to external forces such as squeezing and scraping, leading to deformation or even damage to the container body. The condition of the container body directly affects the safety of the cargo. Container body defect inspection is an essential inspection process for containers entering the port terminal, mainly to prevent disputes between transportation companies and terminals caused by container damage or deformation. Therefore, the inspection of container defects is an important part of improving the standards and efficiency of the maintenance industry.

[0003] Currently, manual inspection of containers suffers from low efficiency and inconsistent evaluation standards. Estimating dents and convexities solely by visual inspection can easily lead to disputes. The lack of images and defect data for dents and cracks makes archiving difficult, hindering repair and traceability. Existing methods for detecting container damage, such as scanning the container surface and comparing it with images of undamaged containers, suffer from limitations. These limitations overlook the fact that containers from over twenty manufacturers worldwide have inconsistent constructions and slightly different surface layouts. Therefore, if only an undamaged section of a container from one model is included in the image set for comparison, undamaged models of other container types will be missed, making it impossible to guarantee the inspection of all container types.

[0004] Therefore, overcoming the shortcomings of the existing technology is an urgent problem to be solved in this technical field. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to scan and inspect the container surface, and to inspect multiple different types of containers when different container models are incomplete.

[0006] The present invention adopts the following technical solution:

[0007] In a first aspect, a container inspection system includes: a scanning module, a camera 10, a container surface defect identification unit, a container surface defect detection unit, an information identification unit, and a data preprocessing unit, wherein:

[0008] Three slide rails 1 are respectively set on the left and right sides and the top of the gate fixing frame 8, and sliders 2 are set on each of the three slide rails 1. Each slider 2 is equipped with a scanning module. When the container 9 to be inspected enters the preset position in the gate, the sliders 2 on the left and right sides and the top of the gate fixing frame 8 slide along the slide rails 1, and the scanning modules on the sliders 2 scan the left side, right side and top side of the container 9 respectively.

[0009] The scanning module includes a 2D line scan camera 3 and a laser scanner 4;

[0010] The 2D line scan camera 3 is used to scan the surface of the container 9 to obtain real-scene images and upload them to the surface defect identification unit. The surface defect identification unit makes defect judgments based on the real-scene images and reports the defect judgment results to the data preprocessing unit.

[0011] The laser scanner 4 is used to scan the surface of container 9, obtain the point cloud file of container 9, and upload it to the container surface defect detection unit. The container surface defect detection unit divides the point cloud of each container surface into a preset unit cycle, compares the point cloud of each segmented region with the defect-free unit reference model, and obtains the defect results of each container surface based on the comparison results and reports them to the data preprocessing unit.

[0012] A camera 10 is installed on the gate fixing frame 8. The camera 10 is used to take pictures of the container 9 at the preset position and upload the pictures to the information recognition unit. The information recognition unit extracts container information based on the pictures and uploads it to the data preprocessing unit.

[0013] The data preprocessing unit is used to output detection results based on one or more of the defect identification results, the defect results, and the container information.

[0014] Preferably, the laser scanner 4 is used to scan the surface of the container 9, obtain point cloud files of the container 9 surface, and upload them to the surface defect detection unit, and further includes:

[0015] The container surface defect detection unit performs point cloud coordinate system transformation for each container surface according to the point cloud file of each container surface of the container 9, and removes the edge part of the container surface in the point cloud according to the transformed point cloud coordinate system, retaining the concave edge 7 and convex edge 6 of the container surface, to obtain the corresponding left container surface point cloud, right container surface point cloud and upper container surface point cloud under the point cloud coordinate system of each container surface.

[0016] Preferably, the preset unit period specifically includes:

[0017] Each adjacent concave edge 7 and convex edge 6 constitutes a preset unit cycle.

[0018] Preferably, the defect-free element reference model includes: a defect-free element reference model for the upper box surface and defect-free element reference models for the left and right box surfaces, wherein:

[0019] The defect-free unit reference model for the upper box surface consists of equations corresponding to the two inclined planes and two planes containing a defect-free concave edge 7 and a defect-free convex edge 6 on the upper box surface. The point cloud of the upper box surface is segmented according to the preset unit period, and the segmented point cloud is correlated with the intervals of the equations in the defect-free unit reference model. The x and y coordinates of each segmented point cloud are substituted into the equations of the corresponding interval to obtain the standard depth value Z at the corresponding position of the segmented upper box surface point cloud. a ;

[0020] The defect-free unit reference model for the left and right box surfaces consists of equations corresponding to the two inclined planes and two planes containing a defect-free concave edge 7 and a defect-free convex edge 6 on the left and right box surfaces. The point clouds of both the left and right box surfaces are segmented according to the preset unit period. The segmented point clouds are then correlated with the intervals in the equations of the defect-free unit reference model. The x and y coordinates of each segmented point cloud are substituted into the equations of the corresponding intervals to obtain the standard depth value Z at the corresponding positions of the segmented left and right box surface point clouds. c ;

[0021] Preferably, the step of comparing each segmented point cloud with a defect-free unit baseline model specifically includes:

[0022] Subtract the standard depth value Z of the corresponding part from the z-coordinate of the segmented upper box surface point cloud. a The difference z is obtained and compared with the depth threshold Z1 of the convex edge 6 and the depth threshold Z2 of the concave edge 7 of the reference model of the defect-free element on the upper box surface.

[0023] When Z2≤z≤Z1, the corresponding upper box surface segmentation area is judged to be a qualified area;

[0024] When z > Z1 or z < Z2, the corresponding upper box surface segmentation area is determined to be an unqualified area;

[0025] Subtract the standard depth value Z of the corresponding portion from the z-coordinate of the left and right box-shaped point clouds after segmentation. c The difference z is obtained, and the depth value z is compared with the depth threshold Z3 of the convex edge 6 and the depth threshold Z4 of the concave edge 7 of the reference model of the defect-free element on the left and right box surfaces.

[0026] When Z4≤z≤Z3, the corresponding left and right box surface division areas are judged to be qualified areas;

[0027] When z > Z3 or z < Z4, the corresponding left and right box surface segmentation areas are determined to be unqualified areas;

[0028] Where Z1 > 0, Z2 < 0, Z3 > 0, and Z4 < 0.

[0029] Secondly, a container inspection method, applied to the aforementioned container inspection system, the method comprising:

[0030] When container 9 enters the preset position in the gate, the sliders 2 on the left and right sides and the top of the gate fixing frame 8 are controlled to slide along the slide rail 1 so that the left, right and top surfaces of container 9 can be scanned by the scanning module on the slider 2; wherein, the scanning module includes a 2D line scan camera 3 and a laser scanner 4.

[0031] The 2D line scan camera 3 scans the container 9 to obtain a real-scene image of the container 9 surface and uploads it to the container surface defect identification unit. The container surface defect identification unit performs defect identification based on the real-scene image and reports the defect identification result to the data preprocessing unit.

[0032] The laser scanner 4 scans the surface of container 9 to obtain point cloud files of container 9 and uploads them to the container surface defect detection unit. The container surface defect detection unit divides the point cloud of each container surface into a preset unit cycle, compares the point cloud of each segmented region with the defect-free unit benchmark model, and obtains the defect results of each container surface based on the comparison results and reports them to the data preprocessing unit.

[0033] The camera 10 installed on the gate fixing frame 8 takes a picture of the container 9 at the preset position and uploads the picture to the information recognition unit. The information recognition unit extracts container information based on the picture and uploads it to the data preprocessing unit.

[0034] The data preprocessing unit outputs detection results based on one or more of the defect identification results, defect results, and container information.

[0035] Preferably, the laser scanner 4 is used to scan the surface of the container 9 to obtain point cloud files of the container 9 surface and upload them to the surface defect detection unit, and further includes:

[0036] The container surface defect detection unit performs point cloud coordinate system transformation for each container surface according to the point cloud file of each container surface of the container 9, and removes the edge part of the container surface in the point cloud according to the transformed point cloud coordinate system, retaining only the concave edge 7 and convex edge 6 of the container surface, to obtain the corresponding left container surface point cloud, right container surface point cloud and upper container surface point cloud under the point cloud coordinate system of each container surface.

[0037] Preferably, the preset unit period specifically includes:

[0038] Each adjacent concave edge 7 and convex edge 6 constitutes a preset unit cycle.

[0039] The defect-free element reference model includes: a defect-free element reference model for the upper box surface and defect-free element reference models for the left and right box surfaces, wherein:

[0040] The defect-free unit reference model for the upper box surface consists of equations corresponding to the two inclined planes and two planes containing a defect-free concave edge 7 and a defect-free convex edge 6 on the upper box surface. The point cloud of the upper box surface is segmented according to the preset unit period, and the segmented point cloud is correlated with the intervals of the equations in the defect-free unit reference model. The x and y coordinates of each segmented point cloud are substituted into the equations of the corresponding interval to obtain the standard depth value Z at the corresponding position of the segmented upper box surface point cloud. a ;

[0041] The defect-free unit reference model for the left and right box surfaces consists of equations corresponding to the two inclined planes and two planes containing a defect-free concave edge 7 and a defect-free convex edge 6 on the left and right box surfaces. The point clouds of both the left and right box surfaces are segmented according to the preset unit period. The segmented point clouds are then correlated with the intervals in the equations of the defect-free unit reference model. The x and y coordinates of each segmented point cloud are substituted into the equations of the corresponding intervals to obtain the standard depth value Z at the corresponding positions of the segmented left and right box surface point clouds. c ;

[0042] Preferably, the step of comparing each segmented point cloud with a defect-free unit baseline model specifically includes:

[0043] Subtract the standard depth value Z of the corresponding part from the z-coordinate of the segmented upper box surface point cloud. a The difference z is obtained and compared with the depth threshold Z1 of the convex edge 6 and the depth threshold Z2 of the concave edge 7 of the reference model of the defect-free element on the upper box surface.

[0044] When Z2≤z≤Z1, the corresponding upper box surface segmentation area is judged to be a qualified area;

[0045] When z > Z1 or z < Z2, the corresponding upper box surface segmentation area is determined to be an unqualified area;

[0046] Subtract the standard depth value Z of the corresponding portion from the z-coordinate of the left and right box-shaped point clouds after segmentation. c The difference z is obtained, and the depth value z is compared with the depth threshold Z3 of the convex edge 6 and the depth threshold Z4 of the concave edge 7 of the reference model of the defect-free element on the left and right box surfaces.

[0047] When Z4≤z≤Z3, the corresponding left and right box surface division areas are judged to be qualified areas;

[0048] When z > Z3 or z < Z4, the corresponding left and right box surface segmentation areas are determined to be unqualified areas;

[0049] Where Z1 > 0, Z2 < 0, Z3 > 0, and Z4 < 0.

[0050] This invention provides a container inspection system and method. Scanning modules, supported by sliders, are installed on the left, right, and top sides of the container inspection gate. During container inspection, the system scans the container surface to obtain real-world images, point cloud files, and container information. The real-world images are used to identify defects in the container surface. The point cloud of each surface is segmented according to a preset unit period. The point cloud files of the segmented areas are compared with a defect-free unit baseline model to determine the defect results for each surface. A data preprocessing unit outputs inspection results based on one or more of the defect identification results, defect results, and container information. Although different manufacturers have different production processes, and we cannot exhaustively list all container surface arrangements, the frame dimensions of containers have unified standards. The container surface can be segmented according to a preset unit period, and the resulting unit periods are regular. This invention inspects the surface of each unit period to determine whether defects exist, and can inspect multiple different container models. Attached Figure Description

[0051] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments of the present invention will be briefly described below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0052] Figure 1 This is a simplified schematic diagram of a container inspection system provided in an embodiment of the present invention;

[0053] Figure 2 This is a schematic diagram of the structure of a scanning module of a container inspection system provided in an embodiment of the present invention;

[0054] Figure 3 This is a schematic diagram of the operation of a container inspection system provided in an embodiment of the present invention;

[0055] Figure 4 This is a top view of a container in a container inspection system provided in an embodiment of the present invention;

[0056] Figure 5 This is a flowchart of a container inspection method provided in an embodiment of the present invention;

[0057] The labels in the attached diagram are as follows:

[0058] 1. Slide rail; 2. Slider; 3. 2D line scan camera; 4. Laser scanner; 5. Motor; 6. Protruding edge; 7. Concave edge; 8. Gate fixing frame; 9. Container; 10. Camera. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0060] In the description of this invention, the terms "inner", "outer", "longitudinal", "lateral", "upper", "lower", "top", "bottom", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and do not require that this invention must be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0061] Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0062] Example 1:

[0063] Embodiment 1 of the present invention provides a container inspection system.

[0064] like Figures 1-3 As shown, it includes: a scanning module, a camera 10, a box surface defect recognition unit, a box surface defect detection unit, an information recognition unit, and a data preprocessing unit, wherein:

[0065] Three slide rails 1 are respectively set on the left and right sides and the top of the gate fixing frame 8, and sliders 2 are set on each of the three slide rails 1. Each slider 2 is equipped with a scanning module. When the container 9 to be inspected enters the preset position in the gate, the sliders 2 on the left and right sides and the top of the gate fixing frame 8 slide along the slide rails 1, and the scanning modules on the sliders 2 scan the left side, right side and top side of the container 9 respectively.

[0066] The container 9 targeted in this embodiment is a standard container 9 available on the market. Alternating concave ridges 7 and convex ridges 6 are provided on the left, right and top surfaces of the container 9; and the left and right sides of the container 9 are completely symmetrical.

[0067] The preset position is the inspection area of ​​container 9. The preset position is located in the middle of the left and right sides of the gate's fixing frame 8. The gate fixing frame 8 is usually equipped with a sensor module. The preset position is located laterally to the location of the sensor module and is marked at this position. When a vehicle carrying container 9 is parked at the preset position, the sensor is triggered. The sensor informs the control terminal, which then controls the slide rail 1 and slider 2 to move to complete the scanning task.

[0068] All three slide rails 1 are parallel to the gate in the gate, ensuring that when the vehicle carrying container 9 enters the gate along the gate, the slider 2 on the slide rail 1 can scan container 9 parallel to container 9.

[0069] Motors 5 are also installed at the ends of the three slide rails 1 located on the left, right and top sides of the gate fixing frame 8. The motors 5 are controlled to open and close via a control terminal. When open, the motors 5 rotate at a constant speed, driving the slider 2 to move at a constant speed on the slide rail 1, ensuring that the scanning module scans the container 9 stably and evenly.

[0070] The scanning module includes a 2D line scan camera 3 and a laser scanner 4;

[0071] The laser scanner 4 used in this embodiment is a laser scanner 4 with a ranging accuracy of 0.5mm within a ranging range of 5 meters.

[0072] The 2D line scan camera 3 is used to scan the surface of the container 9 to obtain real-scene images and upload them to the surface defect identification unit. The surface defect identification unit makes defect judgments based on the real-scene images and reports the defect judgment results to the data preprocessing unit.

[0073] The container surface damage identification unit uses visual AI algorithms to collect a large amount of relevant data and complete AI learning and database accumulation for the images of the three sides of the container 9 scanned by the 2D line scan camera 3. The data is then reported to the data preprocessing unit. In this embodiment, the damage to the container 9 includes issues such as paint touch-ups, leaves, gaps, and dirt on the container surface.

[0074] The 2D line scan camera 3 is used to identify obvious defects on the surface of container 9, and the visual AI algorithm is used to directly identify and locate them.

[0075] The laser scanner 4 is used to scan the surface of the container 9, obtain the point cloud file of the container surface, and upload it to the container surface defect detection unit. The container surface defect detection unit divides the point cloud of each container surface in a preset unit period, compares the point cloud of each divided area with the defect-free unit reference model, and the container surface defect detection unit obtains the defect results of each container surface according to the comparison results and reports them to the data preprocessing unit;

[0076] As Figure 4 shown, where Figure 4 is a top view of the container 9. The laser scanner 4 scans the left, right, and upper surfaces of the container 9 at a constant speed to obtain the three-dimensional point cloud data on the surfaces of the left, right, and upper surfaces of the container 9, and performs coordinate transformation on the obtained three-dimensional point cloud data. The surface where the container 9 is located is used as the two-dimensional plane of the x-axis and y-axis, and the direction perpendicular to the surface of the container 9 is the z-axis direction. The convex edge 6 of the defect-free unit reference model is used as the protrusion depth threshold, and the concave edge 7 of the defect-free unit reference model is used as the depression depth threshold. When the protrusion depth or depression depth on the surface of the tested divided area is within the range of the protrusion depth threshold and the depression depth threshold, it means that the divided area is qualified, otherwise it is unqualified. After comparing all the parts divided in the preset unit period with the defect-free unit reference model, the defect positions on the entire surface of the container 9 can be accurately located and the defect degree can be determined.

[0077] The preset unit period in this embodiment is: each adjacent convex edge 6 and concave edge 7 on the surface of the container 9 is a preset unit period.

[0078] A camera 10 is provided on the gate fixing frame 8. The camera 10 is used to photograph the container 9 at the preset position and upload the photographed image to the information recognition unit. The information recognition unit extracts the container information from the photographed image and uploads it to the data preprocessing unit;

[0079] The photographing object of the camera 10 includes the container 9 and the vehicle carrying the container 9. The container information includes one or more of the container number of the container 9, the type of the container 9, and the license plate number of the vehicle carrying the container.

[0080] The data preprocessing unit is used to output a detection result according to one or more of the defect discrimination result, the defect result, and the container information.

[0081] After the data preprocessing unit obtains the detection results, the output process is as follows: Defect location information from the point cloud is extracted and integrated, and combined with real-world images of each surface, a 1:1 proportional correspondence is performed. Various defect categories and information are marked in the images, displayed as images on the software interface, and output as messages. Simultaneously, this embodiment also includes reporting the final classification according to container number 9 to the container body repair plan evaluation unit. The container body repair plan evaluation unit derives and outputs a repair plan based on the detection results for user reference. The repair plan is also uploaded to the repair quotation unit. The repair quotation unit estimates the quotation result based on the unit price of the repair category for container body 9 and related labor hour requirements, combined with the repair plan, and outputs and displays it on the software interface for user reference.

[0082] Currently, manual inspection of container 9 is inefficient and suffers from inconsistent evaluation standards. Relying solely on visual estimation of dents and convexities can easily lead to disputes. Furthermore, the lack of images and defect data for dents and cracks hinders archiving, making repair and traceability difficult. Existing methods for detecting container 9 damage, such as scanning the surface and comparing it to undamaged images, suffer from limitations. These limitations overlook the fact that the structures of containers 9 produced by over twenty global manufacturers are not identical. Slight variations in the surface layout and welding processes result in different starting positions for edge recesses. Therefore, comparing only undamaged surfaces of one model in the image set misses undamaged models of other container 9 models. While it's impossible to guarantee that all container 9 surface models are included in the image set... While the existing container 9 has a standardized frame size, the dimensions of individual protruding ridges 6 and concave ridges 7 on containers 9 of different models or from different manufacturers are uniform. Therefore, this embodiment divides the surface of the container 9 by setting each adjacent protruding ridge 6 and concave ridge 7 as a preset unit cycle. Each divided unit is compared with a standard defect-free unit reference model. If the divided part is within the standard range, it means that the divided part is qualified. If the divided part exceeds the standard range, it means that there is an abnormal protrusion or depression deformation in the container, which means that part is unqualified. After all the divided units have been compared and judged, the defect status of the entire container 9 is judged and the defect location can be accurately located. This solves the problem that different models of container 9 cannot be referenced for inspection due to different surface layouts.

[0083] Laser scanner 4 scans container 9 and obtains the original point cloud file. Simultaneously, it is necessary to extract the container surface contour point cloud from the original point cloud file to lay the foundation for subsequent removal of the container surface edges. The process is as follows:

[0084] Using the mesh structure method, based on the characteristics and dimensions of the positioning holes on the left side of container 9, the direction, length, and width of the feature point positioning holes are set, and the outline point cloud of the left side is extracted from the original point cloud of the left side; wherein, the positioning holes are through holes set at the four vertices of each surface of the container, used for container positioning.

[0085] Based on the features and dimensions of the positioning holes on the right side of container 9, the direction, length, and width of the feature point positioning holes are set, and the outline point clouds of the left and right sides are extracted from the original point clouds of the left and right sides; based on the features and dimensions of the positioning holes on the top side of container 9, the direction, length, and width of the feature point positioning holes are set, and the outline point cloud of the top side is extracted from the original point cloud of the top side.

[0086] In this embodiment, the left and right sides of container 9 are completely symmetrical, and the scanning modules on the left and right sides of the gate are also completely symmetrical and consistent.

[0087] The container surface defect detection unit performs point cloud coordinate system transformation for each container surface according to the point cloud file of each container surface of the container 9, and removes the edge part of the container surface in the point cloud according to the transformed point cloud coordinate system, retaining the concave edge 7 and convex edge 6 of the container surface, to obtain the corresponding left container surface point cloud, right container surface point cloud and upper container surface point cloud under the point cloud coordinate system of each container surface.

[0088] In this embodiment, the point cloud coordinate system transformation is as follows:

[0089] After feature point extraction, the left box surface point cloud X0O0Y0 containing the contour edge of the surface is obtained. The left box surface point cloud X0O0Y0 represents coordinate data in the initial coordinate system of the scanning module. To facilitate subsequent positioning and calculation of various coordinate points on the left box surface, the left box surface point cloud X0O0Y0 needs to undergo coordinate system transformation. In this embodiment, the surface containing the left box surface point cloud is moved to the surface containing the XOY plane in the initial coordinate system. The coordinates of the lower left corner vertex in the left box surface are moved to the zero point coordinates in the initial coordinate system, and the coordinates of the upper left corner vertex in the left box surface are moved to the Y-axis in the initial coordinate system. The transformation method is as follows:

[0090] Select the 100 points closest to each of the four positioning holes on the left box surface. Take the centroid of the XYZ coordinates of these four sets of 100 points to generate points A, B, C, and D. Select any three of these four points to fit and generate the left reference plane X1O1Y1. Then, calculate the normal to the left reference plane X1O1Y1. Calculate the initial coordinate system Z-axis and normal. Given the included angle α, take the lower left vertex of the left box surface as the reference point O1(a,b,c), where O1(a,b,c) is the coordinate position of the lower left vertex of the left box surface in the initial coordinate system. Translate the entire left box surface point cloud X0O0Y0 by a distance -a along the X-axis, a distance -b along the y-axis, and a distance -c along the z-axis, so that the coordinates of the original reference point O1 are O(0,0,0), generating the point cloud X2O2Y2. Using O1 as the fulcrum, rotate the point cloud X2O2Y2 along the normal direction by an included angle α. α, obtain the left box surface point cloud X3O3Y3 which coincides with the XOY plane of the coordinate system and is perpendicular to the Z axis; take the upper left vertex O2(d,e,0) of the left box surface, and rotate the left box surface point cloud X3O3Y3 counterclockwise around the Z axis by the angle arctan(d / e) to obtain the left box surface point cloud X4O4Y4. At this time, the coordinates of the upper left vertex are O3(0,e,0), that is, the lower left vertex of the left box surface is the origin of the coordinate system O(0,0,0), the bottom is the X axis, and the left is the Y axis, thus completing the coordinate system transformation of the left box surface point cloud.

[0091] Similarly, coordinate system transformation is performed on the right box surface and the upper box surface to obtain the point cloud X5O5Y5 on the right box surface and the point cloud X6O6Y6 on the upper box surface.

[0092] Furthermore, the box-shaped borders in the point cloud are removed, leaving only the concave edges 7 and convex edges 6, resulting in the left box-shaped point cloud X7O7Y7, the right box-shaped point cloud X8O8Y8, and the upper box-shaped point cloud X9O9Y9.

[0093] Wherein, X0O0Y0, X1O1Y1, X2O2Y2, X3O3Y3, X4O4Y4, X5O5Y5, X6O6Y6, X7O7Y7, X8O8Y8 and X9O9Y9 are all point cloud coordinate data, that is, the coordinate data of the entire left box surface in the coordinate system.

[0094] The defect-free unit reference model for the upper box surface consists of equations corresponding to the two inclined planes and two planes containing a defect-free concave edge 7 and a defect-free convex edge 6 on the upper box surface. The point cloud of the upper box surface is segmented according to the preset unit period, and the segmented point cloud is correlated with the intervals of the equations in the defect-free unit reference model. The x and y coordinates of each segmented point cloud are substituted into the equations of the corresponding interval to obtain the standard depth value Z at the corresponding position of the segmented upper box surface point cloud. a ;

[0095] One of the upper box surfaces has a concave edge 7 without defects, and the other has a convex edge 6 without defects. These are composed of two inclined planes and two flat planes. The smallest unit model of the upper box surface, M1, has the following surface equation:

[0096] a4x + b4y + c4z + d4 = 0

[0097] a5x + b5y + c5z + d5 = 0

[0098] a6x + b6y + c6z + d6 = 0

[0099] a7x+b7y+c7z+d7=0;

[0100] The four quaternary linear equations of the surface equation of the minimum unit model M1 of the upper box surface represent the two inclined surfaces and two planes of the undamaged convex edge 6 of the upper box surface, respectively. The surface equation is a piecewise function, and the four quaternary linear equations are only applicable to different fixed intervals, representing the different positions of different inclined surfaces and different planes in the coordinate system. By substituting the X and Y values ​​of the point cloud coordinates at any position on the upper box surface after coordinate system transformation into the equation of the corresponding interval of the surface equation of the minimum unit model M1, the standard depth value Z at the corresponding position can be obtained. a .

[0101] The defect-free unit reference model for the left and right box surfaces consists of equations corresponding to the two inclined planes and two planes containing a defect-free concave edge 7 and a defect-free convex edge 6 on the left and right box surfaces. The point clouds of both the left and right box surfaces are segmented according to the preset unit period. The segmented point clouds are then correlated with the intervals in the equations of the defect-free unit reference model. The x and y coordinates of each segmented point cloud are substituted into the equations of the corresponding intervals to obtain the standard depth value Z at the corresponding positions of the segmented left and right box surface point clouds. c .

[0102] One of the left or right box surfaces has a concave edge 7 without defects and a convex edge 6 without defects, which consists of two inclined planes and two planes. The smallest unit model M0 of the left or right box surface has the following surface equation:

[0103] a0x+b0y+c0z+d0=0

[0104] a1x + b1y + c1z + d1 = 0

[0105] a²x + b²y + c²z + d² = 0

[0106] a3x+b3y+c3z+d3=0.

[0107] The four quaternary linear equations of the surface equation of the minimum unit model M0 of the upper box surface represent the two inclined surfaces and two planes of the undamaged convex edge 6 on the left and right box surfaces, respectively. The surface equations are piecewise functions, and the four quaternary linear equations are only applicable to different fixed intervals, representing the different positions of different inclined surfaces and different planes in the coordinate system. By substituting the X and Y values ​​of the point cloud coordinates of any position on the left and right box surfaces after coordinate system transformation into the equation of the corresponding interval of the surface equation of the minimum unit model M0, the standard depth value Z at the corresponding position can be obtained. c .

[0108] The comparison of each segmented point cloud with a defect-free unit baseline model specifically includes:

[0109] Subtract the standard depth value Z of the corresponding part from the z-coordinate of the segmented upper box surface point cloud. a The difference z is obtained and compared with the depth threshold Z1 of the convex edge 6 and the depth threshold Z2 of the concave edge 7 of the reference model of the defect-free element on the upper box surface.

[0110] When Z2≤z≤Z1, the corresponding upper box surface segmentation area is judged to be a qualified area;

[0111] When z > Z1 or z < Z2, the corresponding upper box surface segmentation area is determined to be an unqualified area;

[0112] Subtract the standard depth value Z of the corresponding portion from the z-coordinate of the left and right box-shaped point clouds after segmentation. c The difference z is obtained, and the depth value z is compared with the depth threshold Z3 of the convex edge 6 and the depth threshold Z4 of the concave edge 7 of the reference model of the defect-free element on the left and right box surfaces.

[0113] When Z4≤z≤Z3, the corresponding left and right box surface division areas are judged to be qualified areas;

[0114] When z > Z3 or z < Z4, the corresponding left and right box surface segmentation areas are determined to be unqualified areas;

[0115] Where Z1 > 0, Z2 < 0, Z3 > 0, and Z4 < 0.

[0116] Z1, Z2, Z3, and Z4 are all the same fixed preset values, which are set by those skilled in the art taking into account the actual model and size of the container itself.

[0117] It is worth mentioning that the first concave edge 7 on the edge of container 9 does not contain a slope. When detecting this position, you can directly compare the Z-axis difference z corresponding to this segment of point cloud with the Z-axis direction difference threshold of the coordinate system XOY plane.

[0118] The defect data of the point cloud of each segment of the left box face are summarized on the point cloud X4O4Y4 of the left box face, and the defect results are output.

[0119] The defect data of the point cloud of each segment of the right box face are summarized on the point cloud X5O5Y5 of the right box face, and the defect results are output.

[0120] The defect data of the point cloud of each segment of the upper box surface are summarized on the point cloud X6O6Y6 of the left box surface, and the defect results are output.

[0121] The above results are then reported to the data preprocessing unit.

[0122] Example 2:

[0123] Based on Embodiment 1, this invention provides a container inspection method, which is applied to the container inspection system of Embodiment 1.

[0124] like Figure 5 As shown, the method flow is as follows:

[0125] In step 101, the sensor detects that container 9 has entered the preset position of the gate.

[0126] In step 102, the controller controls the sliders 2 on the left, right and top sides of the gate fixing frame 8 to slide along the slide rail 1. The left, right and top sides of the container 9 are scanned by the scanning module on the slider 2. At the same time, the process jumps to steps 103, 105 and 107.

[0127] In step 103, the 2D line scan camera 3 scans the container 9 to obtain a real-scene image of the container 9's surface and uploads it to the surface damage identification unit.

[0128] In step 104, the box surface defect identification unit performs defect identification based on the real scene image and reports the defect identification result to the data preprocessing unit, then proceeds to step 109.

[0129] The container surface damage identification unit uses visual AI algorithms to collect a large amount of relevant data and complete AI learning and database accumulation for the images of the three sides of the container 9 scanned by the 2D line scan camera 3. The data is then reported to the data preprocessing unit. In this embodiment, the damage to the container 9 includes issues such as paint touch-ups, leaves, gaps, and dirt on the container surface.

[0130] The 2D line scan camera 3 is used to identify obvious defects on the surface of container 9, and the visual AI algorithm is used to directly identify and locate them.

[0131] In step 105, the laser scanner 4 scans the container 9 to obtain the point cloud file of the container 9 surface and uploads it to the surface defect detection unit.

[0132] In step 106, the box surface defect detection unit divides the point cloud of each box surface into a preset unit period, compares the point cloud of each segmented region with the defect-free unit reference model, and obtains the defect results of each box surface based on the comparison results and reports them to the data preprocessing unit, then jumps to step 109.

[0133] The laser scanner 4 scans the left, right, and top surfaces of container 9 at a constant speed to obtain three-dimensional point cloud data of the three surfaces. The obtained three-dimensional point cloud data is then transformed into a coordinate system, with the surface of container 9 as the two-dimensional plane of the x and y axes, and the direction perpendicular to the surface of container 9 as the Z-axis. The protruding edge 6 of the defect-free unit reference model is used as the protrusion depth threshold, and the concave edge 7 of the defect-free unit reference model is used as the concavity depth threshold. When the protrusion depth or concavity depth of the tested segmented area is within the range of the protrusion depth threshold and the concavity depth threshold, the segmented area is considered qualified; otherwise, it is unqualified. By comparing all the parts segmented with the preset unit cycle with the defect-free unit reference model, the defect location of the entire surface of container 9 can be accurately located, and the degree of defect can be determined.

[0134] In this embodiment, the preset unit cycle is defined as follows: each adjacent convex edge 6 and concave edge 7 on the surface of container 9 constitutes one preset unit cycle.

[0135] In step 107, the camera 10 is installed on the gate fixing frame 8, and the camera 10 takes pictures of the container 9 at the preset position and uploads the pictures to the information recognition unit.

[0136] In step 108, the information identification unit extracts container information based on the captured photos and uploads it to the data preprocessing unit, then proceeds to step 109.

[0137] The camera 10 captures images of the container 9 and the vehicle carrying the container 9. The container information includes one or more of the container 9 number, container 9 type, and vehicle license plate number.

[0138] In step 109, the data preprocessing unit outputs one or more of the detection results of container 9 and container information based on the defect judgment result and the defect result.

[0139] After the data preprocessing unit obtains the detection results, the output process is as follows: Defect location information from the point cloud is extracted and integrated, and combined with real-world images of each surface, a 1:1 proportional correspondence is performed. Various defect categories and information are marked in the images, displayed as images on the software interface, and output as messages. Simultaneously, this embodiment also includes reporting the final classification according to container number 9 to the container body repair plan evaluation unit. The container body repair plan evaluation unit derives and outputs a repair plan based on the detection results for user reference. The repair plan is also uploaded to the repair quotation unit. The repair quotation unit estimates the quotation result based on the unit price of the repair category for container body 9 and related labor hour requirements, combined with the repair plan, and outputs and displays it on the software interface for user reference.

[0140] Laser scanner 4 scans container 9 and obtains the original point cloud file. Simultaneously, it is necessary to extract the container surface contour point cloud from the original point cloud file to lay the foundation for subsequent removal of the container surface edges. The process is as follows:

[0141] Using the mesh structure method, based on the characteristics and dimensions of the positioning holes on the left side of container 9, the direction, length, and width of the feature point positioning holes are set, and the outline point cloud of the left side is extracted from the original point cloud of the left side. Based on the characteristics and dimensions of the positioning holes on the right side of container 9, the direction, length, and width of the feature point positioning holes are set, and the outline point cloud of the right side is extracted from the original point cloud of the right side. Based on the characteristics and dimensions of the positioning holes on the top side of container 9, the direction, length, and width of the feature point positioning holes are set, and the outline point cloud of the top side is extracted from the original point cloud of the top side.

[0142] The container surface defect detection unit performs point cloud coordinate system transformation on each container surface according to the point cloud file of each container surface of the container 9, and removes the edge part of the container surface in the point cloud according to the transformed point cloud coordinate system, retaining only the concave edge 7 and convex edge 6 of the container surface, to obtain the corresponding left container surface point cloud, right container surface point cloud and upper container surface point cloud under the point cloud coordinate system of each container surface.

[0143] In this embodiment, the point cloud coordinate system transformation is as follows:

[0144] After feature point extraction, the left box surface point cloud X0O0Y0 containing the contour edge of the surface is obtained. The left box surface point cloud X0O0Y0 represents coordinate data in the initial coordinate system of the scanning module. To facilitate subsequent positioning and calculation of various coordinate points on the left box surface, the left box surface point cloud X0O0Y0 needs to undergo coordinate system transformation. In this embodiment, the surface containing the left box surface point cloud is moved to the surface containing the XOY plane in the initial coordinate system. The coordinates of the lower left corner vertex in the left box surface are moved to the zero point coordinates in the initial coordinate system, and the coordinates of the upper left corner vertex in the left box surface are moved to the Y-axis in the initial coordinate system. The transformation method is as follows:

[0145] Select the 100 points closest to each of the four positioning holes on the left box surface. Take the centroid of the XYZ coordinates of these four sets of 100 points to generate points A, B, C, and D. Select any three of these four points to fit and generate the left reference plane X1O1Y1. Then, calculate the normal to the left reference plane X1O1Y1. Calculate the initial coordinate system Z-axis and normal. Given the included angle α, take the lower left vertex of the left box surface as the reference point O1(a,b,c), where O1(a,b,c) is the coordinate position of the lower left vertex of the left box surface in the initial coordinate system. Translate the entire left box surface point cloud X0O0Y0 by a distance -a along the X-axis, a distance -b along the y-axis, and a distance -c along the z-axis, so that the coordinates of the original reference point O1 are O(0,0,0), generating the point cloud X2O2Y2. Using O1 as the fulcrum, rotate the point cloud X2O2Y2 along the normal direction by an included angle α. α, obtain the left box surface point cloud X3O3Y3 which coincides with the XOY plane of the coordinate system and is perpendicular to the Z axis; take the upper left vertex O2(d,e,0) of the left box surface, and rotate the left box surface point cloud X3O3Y3 counterclockwise around the Z axis by the angle arctan(d / e) to obtain the left box surface point cloud X4O4Y4. At this time, the coordinates of the upper left vertex are O3(0,e,0), that is, the lower left vertex of the left box surface is the origin of the coordinate system O(0,0,0), the bottom is the X axis, and the left is the Y axis, thus completing the coordinate system transformation of the left box surface point cloud.

[0146] Similarly, coordinate system transformation is performed on the right box surface and the upper box surface to obtain the point cloud X5O5Y5 on the right box surface and the point cloud X6O6Y6 on the upper box surface.

[0147] Furthermore, the box-shaped borders in the point cloud are removed, leaving only the concave edges 7 and convex edges 6, resulting in the left box-shaped point cloud X7O7Y7, the right box-shaped point cloud X8O8Y8, and the upper box-shaped point cloud X9O9Y9.

[0148] Wherein, X0O0Y0, X1O1Y1, X2O2Y2, X3O3Y3, X4O4Y4, X5O5Y5, X6O6Y6, X7O7Y7, X8O8Y8 and X9O9Y9 are all point cloud coordinate data, that is, the coordinate data of the entire left box surface in the coordinate system.

[0149] The defect-free unit reference model for the upper box surface consists of equations corresponding to the two inclined planes and two planes containing a defect-free concave edge 7 and a defect-free convex edge 6 on the upper box surface. The point cloud of the upper box surface is segmented according to the preset unit period, and the segmented point cloud is correlated with the intervals of the equations in the defect-free unit reference model. The x and y coordinates of each segmented point cloud are substituted into the equations of the corresponding interval to obtain the standard depth value Z at the corresponding position of the segmented upper box surface point cloud. a ;

[0150] One of the upper box surfaces has a concave edge 7 without defects, and the other has a convex edge 6 without defects. These are composed of two inclined planes and two flat planes. The smallest unit model of the upper box surface, M1, has the following surface equation:

[0151] a4x + b4y + c4z + d4 = 0

[0152] a5x + b5y + c5z + d5 = 0

[0153] a6x + b6y + c6z + d6 = 0

[0154] a7x+b7y+c7z+d7=0;

[0155] The four quaternary linear equations of the surface equation of the minimum unit model M1 of the upper box surface represent the two inclined surfaces and two planes of the undamaged convex edge 6 of the upper box surface, respectively. The surface equation is a piecewise function, and the four quaternary linear equations are only applicable to different fixed intervals, representing the different positions of different inclined surfaces and different planes in the coordinate system. By substituting the X and Y values ​​of the point cloud coordinates at any position on the upper box surface after coordinate system transformation into the equation of the corresponding interval of the surface equation of the minimum unit model M1, the standard depth value Z at the corresponding position can be obtained. a .

[0156] The defect-free unit reference model for the left and right box surfaces consists of equations corresponding to the two inclined planes and two planes containing a defect-free concave edge 7 and a defect-free convex edge 6 on the left and right box surfaces. The point clouds of both the left and right box surfaces are segmented according to the preset unit period. The segmented point clouds are then correlated with the intervals in the equations of the defect-free unit reference model. The x and y coordinates of each segmented point cloud are substituted into the equations of the corresponding intervals to obtain the standard depth value Z at the corresponding positions of the segmented left and right box surface point clouds. c .

[0157] One of the left or right box surfaces has a concave edge 7 without defects and a convex edge 6 without defects, which consists of two inclined planes and two planes. The smallest unit model M0 of the left or right box surface has the following surface equation:

[0158] a0x+b0y+c0z+d0=0

[0159] a1x + b1y + c1z + d1 = 0

[0160] a²x + b²y + c²z + d² = 0

[0161] a3x+b3y+c3z+d3=0.

[0162] The four quaternary linear equations of the surface equation of the minimum unit model M0 of the upper box surface represent the two inclined surfaces and two planes of the undamaged convex edge 6 on the left and right box surfaces, respectively. The surface equations are piecewise functions, and the four quaternary linear equations are only applicable to different fixed intervals, representing the different positions of different inclined surfaces and different planes in the coordinate system. By substituting the X and Y values ​​of the point cloud coordinates of any position on the left and right box surfaces after coordinate system transformation into the equation of the corresponding interval of the surface equation of the minimum unit model M0, the standard depth value Z at the corresponding position can be obtained. c .

[0163] The comparison of each segmented point cloud with a defect-free unit baseline model specifically includes:

[0164] Subtract the standard depth value Z of the corresponding part from the z-coordinate of the segmented upper box surface point cloud. a The difference z is obtained and compared with the depth threshold Z1 of the convex edge 6 and the depth threshold Z2 of the concave edge 7 of the reference model of the defect-free element on the upper box surface.

[0165] When Z2≤z≤Z1, the corresponding upper box surface segmentation area is judged to be a qualified area;

[0166] When z > Z1 or z < Z2, the corresponding upper box surface segmentation area is determined to be an unqualified area;

[0167] Subtract the standard depth value Z of the corresponding portion from the z-coordinate of the left and right box-shaped point clouds after segmentation. c The difference z is obtained, and the depth value z is compared with the depth threshold Z3 of the convex edge 6 and the depth threshold Z4 of the concave edge 7 of the reference model of the defect-free element on the left and right box surfaces.

[0168] When Z4≤z≤Z3, the corresponding left and right box surface division areas are judged to be qualified areas;

[0169] When z > Z3 or z < Z4, the corresponding left and right box surface segmentation areas are determined to be unqualified areas;

[0170] Where Z1 > 0, Z2 < 0, Z3 > 0, and Z4 < 0.

[0171] Z1, Z2, Z3, and Z4 are all the same fixed preset values, which are set by those skilled in the art taking into account the actual model and size of the container itself.

[0172] It is worth mentioning that the first concave edge 7 on the edge of container 9 does not contain a slope. When detecting this position, you can directly compare the Z-axis difference z corresponding to this segment of point cloud with the Z-axis direction difference threshold of the coordinate system XOY plane.

[0173] The defect data of the point cloud of each segment of the left box face are summarized on the point cloud X4O4Y4 of the left box face, and the defect results are output.

[0174] The defect data of the point cloud of each segment of the right box face are summarized on the point cloud X5O5Y5 of the right box face, and the defect results are output.

[0175] The defect data of the point cloud of each segment of the upper box surface are summarized on the point cloud X6O6Y6 of the left box surface, and the defect results are output.

[0176] The above results are then reported to the data preprocessing unit.

[0177] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A container inspection system, characterized in that, include: The system comprises a scanning module, a camera (10), a box surface defect identification unit, a box surface defect detection unit, an information identification unit, and a data preprocessing unit, wherein: Three slide rails (1) are respectively set on the left and right sides and the top of the gate fixing frame (8), and sliders (2) are set on the three slide rails (1). Each slider (2) is equipped with a scanning module. When the container (9) to be inspected enters the preset position in the gate, the sliders (2) on the left and right sides and the top of the gate fixing frame (8) slide along the slide rails (1), and the scanning modules on the sliders (2) scan the left side, right side and top side of the container (9) respectively. The scanning module includes a 2D line scan camera (3) and a laser scanner (4); The 2D line scan camera (3) is used to scan the surface of the container (9) to obtain real-scene images and upload them to the surface defect identification unit. The surface defect identification unit makes defect judgments based on the real-scene images and reports the defect judgment results to the data preprocessing unit. The laser scanner (4) is used to scan the surface of the container (9), obtain the point cloud file of the container (9) surface and upload it to the container surface defect detection unit. The container surface defect detection unit divides the point cloud of each container surface into a preset unit cycle, compares the point cloud of each divided area with the defect-free unit reference model, and the container surface defect detection unit obtains the defect results of each container surface based on the comparison results and reports them to the data preprocessing unit. A camera (10) is installed on the gate fixing frame (8). The camera (10) is used to take pictures of the container (9) at the preset position and upload the pictures to the information recognition unit. The information recognition unit extracts container information based on the pictures and uploads it to the data preprocessing unit. The data preprocessing unit is used to output detection results based on one or more of the defect identification results, the defect results, and the container information; The defect-free element reference model includes: a defect-free element reference model for the upper box surface and defect-free element reference models for the left and right box surfaces, wherein: The reference model for the defect-free unit on the upper box surface consists of equations corresponding to the two inclined planes and two planes containing a defect-free concave edge (7) and a defect-free convex edge (6) on the upper box surface. The point cloud on the upper box surface is segmented according to the preset unit period, and the segmented point cloud is correlated with the interval of the equation in the reference model for the defect-free unit. The x and y coordinates of each segmented point cloud are substituted into the equation of the corresponding interval to obtain the standard depth value of the corresponding position of the segmented point cloud on the upper box surface. ; The reference model for the defect-free unit on the left and right box surfaces consists of equations corresponding to the two inclined planes and two planes containing a defect-free concave edge (7) and a defect-free convex edge (6) on the left and right box surfaces. The point clouds on both the left and right box surfaces are segmented according to the preset unit period, and the segmented point clouds are correlated with the intervals of the equations in the reference model for the defect-free unit. The x and y coordinates of each segmented point cloud are substituted into the equations of the corresponding intervals to obtain the standard depth values ​​of the corresponding positions of the segmented point clouds on the left and right box surfaces. ; The equations corresponding to the two inclined planes and two planes are expressed as follows: x+ y+ z+ =0 x+ y+ z+ =0 x+ y+ z+ =0 x+ y+ z+ =0。 2. The container inspection system according to claim 1, characterized in that, The laser scanner (4) is used to scan the surface of the container (9), obtain the point cloud file of the container (9) surface and upload it to the surface defect detection unit, and also includes: The container surface defect detection unit performs point cloud coordinate system transformation for each container surface according to the point cloud file of each container surface (9), and removes the edge part of the container surface in the point cloud according to the transformed point cloud coordinate system, retaining the concave edge (7) and convex edge (6) of the container surface, and obtains the corresponding left container surface point cloud, right container surface point cloud and upper container surface point cloud under the point cloud coordinate system of each container surface.

3. The container inspection system according to claim 2, characterized in that, The preset unit period specifically includes: Each adjacent concave edge (7) and convex edge (6) constitutes a preset unit cycle.

4. The container inspection system according to claim 1, characterized in that, The comparison of each segmented point cloud with a defect-free unit baseline model specifically includes: Subtract the standard depth value of the corresponding part from the z-coordinate of the segmented upper box surface point cloud. The difference z is obtained, and the difference z is the same as the depth threshold of the convex edge (6) of the reference model of the defect-free element on the upper box surface. and the depth threshold of the concave edge (7) contrast; when ≤z≤ If so, the corresponding upper box surface segmentation area is determined to be a qualified area; When z> Or z < If so, the corresponding upper box surface segmentation area is determined to be an unqualified area; Subtract the standard depth value of the corresponding part from the z-coordinate of the segmented left and right box point clouds. The difference z is obtained, and the depth value z is compared with the depth threshold of the protrusion (6) of the reference model of the defect-free element on the left and right box surfaces. and the depth threshold of the concave edge (7) contrast; when ≤z≤ If so, the corresponding left and right box surface segmentation areas are judged to be qualified areas; When z> Or z < If so, the corresponding left and right box surface segmentation areas are determined to be unqualified areas; in, >0, <0, >0, <0.

5. A method for inspecting containers, characterized in that, The method is applied to the container inspection system as described in any one of claims 1-4, and the method includes: When the container (9) enters the preset position in the gate, the sliders (2) on the left and right sides and the top of the gate fixing frame (8) are controlled to slide along the slide rail (1) so as to scan the left, right and top surfaces of the container (9) by the scanning module on the slider (2); wherein, the scanning module includes a 2D line scan camera (3) and a laser scanner (4). The 2D line scan camera (3) scans the container (9) to obtain a real-scene image of the container (9) surface and uploads it to the container surface defect identification unit. The container surface defect identification unit makes a defect judgment based on the real-scene image and reports the defect judgment result to the data preprocessing unit. The laser scanner (4) scans the surface of the container (9) to obtain the point cloud file of the container (9) surface and uploads it to the container surface defect detection unit. The container surface defect detection unit divides the point cloud of each container surface into a preset unit cycle and compares the point cloud of each divided area with the defect-free unit reference model. The container surface defect detection unit obtains the defect results of each container surface based on the comparison results and reports them to the data preprocessing unit. The camera (10) installed on the gate fixing frame (8) takes a picture of the container (9) at the preset position and uploads the picture to the information recognition unit. The information recognition unit extracts the container information based on the picture and uploads it to the data preprocessing unit. The data preprocessing unit outputs detection results based on one or more of the defect identification results, defect results, and container information.

6. The container inspection method according to claim 5, characterized in that, The laser scanner (4) is used to scan the surface of the container (9) to obtain point cloud files of the container (9) surface and upload them to the surface defect detection unit. It also includes: The container surface defect detection unit performs point cloud coordinate system transformation for each container surface according to the point cloud file of each container surface (9), and removes the edge part of the container surface in the point cloud according to the transformed point cloud coordinate system, only retaining the concave edge (7) and convex edge (6) of the container surface, to obtain the corresponding left container surface point cloud, right container surface point cloud and upper container surface point cloud under the point cloud coordinate system of each container surface.

7. The container inspection method according to claim 6, characterized in that, The preset unit period specifically includes: Each adjacent concave edge (7) and convex edge (6) constitutes a preset unit cycle.

8. The container inspection method according to claim 7, characterized in that, The defect-free element reference model includes: a defect-free element reference model for the upper box surface and defect-free element reference models for the left and right box surfaces, wherein: The reference model for the defect-free unit on the upper box surface consists of equations corresponding to the two inclined planes and two planes containing a defect-free concave edge (7) and a defect-free convex edge (6) on the upper box surface. The point cloud on the upper box surface is segmented according to the preset unit period, and the segmented point cloud is correlated with the interval of the equation in the reference model for the defect-free unit. The x and y coordinates of each segmented point cloud are substituted into the equation of the corresponding interval to obtain the standard depth value of the corresponding position of the segmented point cloud on the upper box surface. ; The reference model for the defect-free unit on the left and right box surfaces consists of equations corresponding to the two inclined planes and two planes containing a defect-free concave edge (7) and a defect-free convex edge (6) on the left and right box surfaces. The point clouds on both the left and right box surfaces are segmented according to the preset unit period, and the segmented point clouds are correlated with the intervals of the equations in the reference model for the defect-free unit. The x and y coordinates of each segmented point cloud are substituted into the equations of the corresponding intervals to obtain the standard depth values ​​of the corresponding positions of the segmented point clouds on the left and right box surfaces. .

9. The container inspection method according to claim 8, characterized in that, The comparison of each segmented point cloud with a defect-free unit baseline model specifically includes: Subtract the standard depth value of the corresponding part from the z-coordinate of the segmented upper box surface point cloud. The difference z is obtained, and the difference z is the same as the depth threshold of the convex edge (6) of the reference model of the defect-free element on the upper box surface. and the depth threshold of the concave edge (7) contrast; when ≤z≤ If so, the corresponding upper box surface segmentation area is determined to be a qualified area; When z> Or z < If so, the corresponding upper box surface segmentation area is determined to be an unqualified area; Subtract the standard depth value of the corresponding part from the z-coordinate of the segmented left and right box point clouds. The difference z is obtained, and the depth value z is compared with the depth threshold of the protrusion (6) of the reference model of the defect-free element on the left and right box surfaces. and the depth threshold of the concave edge (7) contrast; when ≤z≤ If so, the corresponding left and right box surface segmentation areas are judged to be qualified areas; When z> Or z < If so, the corresponding left and right box surface segmentation areas are determined to be unqualified areas; in, >0, <0, >0, <0.

Images