CN114299282A - Box type identification system and method for automatic wharf container - Google Patents
Box type identification system and method for automatic wharf container Download PDFInfo
- Publication number
- CN114299282A CN114299282A CN202210001257.3A CN202210001257A CN114299282A CN 114299282 A CN114299282 A CN 114299282A CN 202210001257 A CN202210001257 A CN 202210001257A CN 114299282 A CN114299282 A CN 114299282A
- Authority
- CN
- China
- Prior art keywords
- container
- module
- box type
- ruler
- picture
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a box type identification system and a method for an automatic wharf container, wherein the box type identification system comprises the following steps: the calibration module is used for acquiring a real-time video stream of a field lane and performing distortion removal processing on the real-time video stream; the acquisition module acquires the real-time video stream, decodes the real-time video stream into a picture and preprocesses the picture; the algorithm module is used for acquiring the picture and judging that the container is a single 20-ruler box type, a single 40-ruler box type, a double 20-ruler box type or a double 45-ruler box type through algorithm processing; and the data recording module records the history, state information and fault information of each module and stores data. The invention realizes the detection and identification of the container type of the automatic wharf container by combining AI with a machine vision detection, segmentation and positioning technology, has high detection precision, high speed and good real-time property, and is suitable for the requirement of the identification of the container type of the automatic wharf container.
Description
Technical Field
The invention relates to a box type detection, identification and positioning technology for an automatic wharf container, in particular to a box type identification system and a box type identification method for an automatic wharf container.
Background
Since the emergence of new crown epidemic situation, with the explosive growth of global demand for materials, the volume of transportation of trade goods is also the rapid growth, and the demand proportion of container constantly improves, and the volume of transportation of container also presents the explosive growth trend. From the economic accounting analysis of the voyage times, the berthing cost can be reduced by shortening the berthing time of the container ships, the voyage efficiency of the container transport ships is improved, the advantages of the unit transport cost of the ships are fully exerted, the economic benefit is improved, and higher requirements are provided for the efficiency of container loading and unloading operation.
At present, the box type of the automatic wharf crane operation container is divided into four types of single 20-ruler, single 40-ruler, single 45-ruler and double 20-ruler, wherein the double 20-ruler container is composed of two single 20-ruler containers. Under the general condition, a container on a container truck needs to be stopped at a specified area position according to a set lane before being loaded and unloaded, after the container truck is guided in place, crane equipment runs to the position above the container truck, a crane lifting appliance starts to load and unload the container to the container truck, and before the container truck is grabbed and released, the specific type of the container is distinguished in advance, so that the important effect is achieved, the operation efficiency can be greatly improved, and the operation safety can be improved.
In the prior art, the traditional manual participation or laser scanning detection mode is mostly adopted. The former container information generally depends on the information operating system of the automatic wharf tally clerk or the wharf crane operator to visually observe and identify the container type, and then the crane spreader is controlled to be arranged at the corresponding scale position corresponding to the size of the corresponding container. According to the extending length and position of the arm of the lifting appliance, the wharf crane central control system can acquire the container type of the operation, and then the container type information is fed back to the mechanism needing the information. In the prior art, some mechanisms cannot establish communication with a wharf crane central control system, the mechanism must rely on self detection of a box type, but at present, no mechanism for independently detecting the box type exists. The latter utilizes laser collection card guide system, also can only distinguish single 20 chi, 40 chi and 45 chi box, is difficult to distinguish to single 40 chi and two 20 chi boxes, has met the bottleneck in detection discernment.
Disclosure of Invention
Aiming at the defects in the prior art, the invention aims to provide a box type identification system and a box type identification method for an automatic wharf container, which are used for realizing the detection and identification of the box type of the automatic wharf container by combining AI with a machine vision detection, segmentation and positioning technology, have high detection precision, high speed and good real-time property and are suitable for the requirements of the identification of the box type of the automatic wharf container.
In order to achieve the purpose, the invention adopts the following technical scheme:
in one aspect, a box-type identification system for an automated quay container, comprising:
the calibration module is used for acquiring a real-time video stream of a field lane and performing distortion removal processing on the real-time video stream;
the acquisition module acquires the real-time video stream, decodes the real-time video stream into a picture and preprocesses the picture;
the algorithm module is used for acquiring the picture and judging that the container is a single 20-ruler box type, a single 40-ruler box type, a double 20-ruler box type or a double 45-ruler box type through algorithm processing;
and the data recording module records the history, state information and fault information of each module and stores data.
Preferably, the calibration module is an industrial camera and is arranged on the crane.
Preferably, the algorithm module further comprises:
the AI detection algorithm module I is used for realizing the rough positioning of the whole truck head and truck body in the picture;
the AI segmentation algorithm module is used for realizing the pixel level high-precision positioning of the container outline in the picture;
the coordinate system conversion algorithm module is used for calculating the length of the container;
the first judging module is used for preliminarily judging whether the container is a single 20-ruler, single 40-ruler, double 20-ruler or 45-ruler box type;
the AI detection algorithm module II is used for detecting the number of the gaps or the lock holes of the container in the picture;
and the second judging module is used for finally judging that the container is a single 20-ruler, single 40-ruler, double 20-ruler or 45-ruler box type.
In another aspect, a box-type identification system for an automated quay container, the box-type identification system for an automated quay container being adapted to perform the steps of:
s1, acquiring a real-time video stream of a field lane through the calibration module, and performing distortion removal processing on the real-time video stream;
s2, acquiring the real-time video stream through the acquisition module, decoding the real-time video stream into a picture, and preprocessing the picture;
and S3, inputting the preprocessed pictures into the algorithm module, and judging that the container is a single 20-ruler, single 40-ruler, double 20-ruler or 45-ruler box type.
Preferably, in step S1, the calibration module is an industrial camera and is installed on a crane;
before installation, camera internal reference calibration is carried out on the industrial camera;
and after the industrial camera is installed, carrying out camera external parameter calibration on the industrial camera.
Preferably, the camera internal reference calibration further includes:
and acquiring the internal reference matrix M and distortion parameters K1, K2, K3, P1 and P2 of the industrial camera by using a calibration plate and a Zhang calibration method, and performing distortion removal correction on the industrial camera through the internal reference matrix M and the distortion parameters K1, K2, K3, P1 and P2.
The camera external reference calibration further comprises:
and obtaining a rotation matrix R and a translation matrix T from the camera coordinate system to the crane coordinate system through a PnP algorithm or a nonlinear Gauss-Newton optimization algorithm by utilizing the coordinates of the calibration plate corner points under the crane coordinate system and the coordinates under the camera pixel coordinate system.
Preferably, in step S2, the preprocessing is filtering, de-drying, graying or histogram equalization processing algorithm.
Preferably, in step S3, the calculation of the algorithm module further includes the following steps:
s31, roughly positioning the truck head and vehicle body in the picture through the AI detection algorithm module, namely inputting the preprocessed picture into a pretrained target detection algorithm model of a YOLO deep learning convolution neural network to obtain a truck head and vehicle body whole vehicle detection result, and then shearing an ROI (region of interest) of the truck head and vehicle body whole vehicle to extract a truck head and vehicle body whole vehicle roughly positioning image;
s32, carrying out pixel-level high-precision positioning on the container outline in the picture through the AI segmentation algorithm module, namely obtaining a pixel-level high-precision positioning image of the container outline of the vehicle head and the vehicle body to be obtained through an AI segmentation algorithm, obtaining pixel coordinates of the vehicle head outline external frame and the upper left corner and the lower right corner of the frame through an external frame post-processing algorithm of the minimum area of the container outline, and restoring the ROI area image of the vehicle head and the vehicle body cut by the AI detection algorithm module to the original image to obtain high-precision pixel coordinates of the upper left corner and the lower right corner of the external frame and the frame of the container outline on the original image through the proportional relation between the ROI area image of the vehicle head and the original image;
s33, converting high-precision pixel coordinates of the upper left corner and the lower right corner of the container outline outer frame through the coordinate system conversion algorithm module through a calibrated camera internal reference matrix M, an external reference rotation matrix R and a translation matrix T to obtain coordinates of the upper left corner and the lower right corner of the container under a crane coordinate system, and calculating to obtain the length of the container;
s34, coordinates of the upper left corner and the lower right corner of the container under a crane coordinate system are obtained through the AI division algorithm module obtained through the first judgment module, the actual length of the container can be obtained through calculation, and the container can be preliminarily judged to be a single 20-ruler box type, a single 40-ruler box type, a double 20-ruler box type or a 45-ruler box type;
s35, detecting the number of container gaps or container lock holes in the picture through the AI detection algorithm module, inputting the segmented container picture into a pre-training target detection algorithm model of a YOLO deep learning convolution neural network to obtain a detection result of the number of the container gaps or the container lock holes;
s36, the AI detection algorithm module II is obtained through the judgment module II to obtain the detection result of the box gap of the container or the number of the lock holes of the container, namely the container is finally judged to be a single 20-foot box type, a single 40-foot box type, a double 20-foot box type or a 45-foot box type.
Preferably, the training process of the neural network algorithm model of the AI detection algorithm module i, the AI detection algorithm module ii and the AI segmentation algorithm module further includes the following steps:
a) establishing a container truck database;
b) making a container truck training, verifying and testing data set;
c) and building the neural network algorithm model.
Preferably, in the step a), the database of the container truck is established as follows:
acquiring the video stream after distortion removal of the calibration module, acquiring an image of the container truck through decoding, and labeling the image of the container truck by using a labeling tool;
in the step b), the data set is specifically made as follows:
extracting a training set, a verification set and a test set from the container truck database;
in the step c), the AI detection and segmentation neural network algorithm model is built as follows:
and training the container truck detection segmentation model by using the training set, judging whether the neural network algorithm models of the first AI detection algorithm module, the second AI detection algorithm module and the AI segmentation algorithm module meet the requirements or not through the verification set, and testing through the test set to obtain the trained neural network algorithm pre-training models of the first AI detection algorithm module, the second AI detection algorithm module and the AI segmentation algorithm module.
The box type identification system and the method for the container of the automatic wharf provided by the invention have the following beneficial effects:
1) before the crane is used, only one industrial camera needs to be installed on the crane, before the crane is installed, internal reference calibration is carried out on the camera, after the crane is installed, external reference calibration is carried out on the camera, and a conversion relation between a camera coordinate system and a crane coordinate system is established, so that the crane is simple in calibration process, convenient to use and low in cost;
2) the method comprises the steps of carrying out rough positioning detection and ROI shearing on the whole locomotive and body by adopting a trained AI detection and segmentation neural network algorithm model, then carrying out high-precision container contour segmentation on an ROI image, restoring the ROI image subjected to high-precision container contour to an original image through a proportional relation between a minimum area external frame and a sheared ROI area image, obtaining high-precision pixel coordinates of the upper left corner and the lower right corner of the container contour on the original image, obtaining the actual length of a container through calculation, preliminarily judging whether the container is a single 20-ruler box type or a single 40-ruler box type, and then judging whether the container to be identified is the double 20-ruler box type or the single 40-ruler box type through box gap or lock hole number detection, so that the container is pixel-level high-precision positioning, and is higher in speed of identifying non-overlapped objects, higher in accuracy, and effectively improving the detection positioning precision;
3) in practical application, along with the storage of new data samples by the data recording module and the increase of calibration data sets, the AI algorithm model can be trained periodically, so that the system identification precision is continuously improved.
Drawings
FIG. 1 is a schematic diagram of a framework of the box-type identification system of the present invention;
FIG. 2 is a schematic diagram of the field layout of the box-type identification system of the present invention;
FIG. 3 is a schematic right-side view of FIG. 2;
FIG. 4 is a flowchart illustrating the box type recognition method of the present invention.
Detailed Description
In order to better understand the technical solutions of the present invention, the following further describes the technical solutions of the present invention with reference to the accompanying drawings and examples.
Referring to fig. 1 to 3, the present invention provides a box-type identification system for an automated quay container, comprising:
the calibration module 100 is used for acquiring a real-time video stream of a field lane and performing distortion removal processing on the real-time video stream;
the acquisition module 101 acquires a real-time video stream, decodes the real-time video stream into a picture, and preprocesses the picture;
the algorithm module 102 acquires pictures, and judges whether the container 2 is a single 20-ruler box type, a single 40-ruler box type, a double 20-ruler box type or a 45-ruler box type through algorithm processing;
and a data recording module 103 for recording the history, state information and fault information of each module and storing data.
The calibration module 100 is installed on the crane 3 by using the industrial camera 1.
The algorithm module 102 further comprises:
the AI detection algorithm module I is used for realizing the rough positioning of the whole container truck head and truck body in the picture;
the AI segmentation algorithm module is used for realizing the pixel level high-precision positioning of the outline of the container 2 in the picture;
the coordinate system conversion algorithm module is used for calculating the length of the container 2;
the first judging module is used for preliminarily judging whether the container 2 is a single 20-ruler, single 40-ruler, double 20-ruler or 45-ruler box type according to the length of the container 2;
the AI detection algorithm module II is used for detecting the box gap of the container 2 or the number of lockholes of the container 2 in the picture;
and the judging module II finally judges that the container 2 is a single 20-ruler box type, a single 40-ruler box type, a double 20-ruler box type or a 45-ruler box type according to the box gap of the container 2 or the number of lock holes of the container 2.
Referring to fig. 1 to 4, the method for identifying a box type of an automated quay container according to the present invention comprises the following steps:
s1, acquiring a real-time video stream of the on-site lane through the calibration module 100, and performing distortion removal processing on the real-time video stream;
s2, acquiring a real-time video stream through the acquisition module 101, decoding the real-time video stream into a picture, and preprocessing the picture;
and S3, inputting the preprocessed pictures into the algorithm module 102, and judging that the container 2 is a single 20-foot, single 40-foot, double 20-foot or 45-foot box type.
In step S1, the calibration module 100 selects the industrial camera 1, installs the industrial camera 1 on the crane 3, and before installation, performs camera internal reference calibration and distortion removal correction on the industrial camera 1; after installation, the external reference calibration of the industrial camera 1 is carried out, and a rotation matrix R and a translation matrix T from a camera coordinate system to a crane coordinate system are obtained.
The camera internal reference calibration further comprises:
obtaining an internal parameter matrix M and distortion parameters K1, K2, K3, P1 and P2 of the industrial camera 1 by using a calibration plate and a Zhang calibration method, and carrying out distortion removal correction on the industrial camera 1 through the internal parameter matrix M and the distortion parameters K1, K2, K3, P1 and P2.
The camera external reference calibration further comprises:
and obtaining a rotation matrix R and a translation matrix T from the camera coordinate system to the crane coordinate system through a PnP algorithm or a nonlinear Gauss-Newton optimization algorithm by utilizing the coordinates of the calibration plate corner points under the crane coordinate system and the coordinates under the camera pixel coordinate system.
In step S2, the preprocessing is filter drying, graying, or histogram equalization processing algorithm.
In step S3, the calculation of the algorithm module further includes the following steps:
s31, roughly positioning the whole container truck head and body in a pair of pictures through an AI (Artificial intelligence) detection algorithm module, namely inputting the preprocessed pictures into a pretraining target detection algorithm model of a deep learning convolution neural network such as YOLO (Yolo) to obtain a detection result of the whole container truck head and body, and then shearing an ROI (region of interest) of the whole container truck head and body to extract a roughly positioned image of the whole container truck head and body;
s32, carrying out pixel-level high-precision positioning on the outline of the container 2 in the picture through an AI (automatic classification) segmentation algorithm module, namely obtaining a pixel-level high-precision positioning image of the outline of the container 2 from the obtained rough positioning image of the whole vehicle of the vehicle head and the vehicle body through an AI segmentation algorithm, such as semantic segmentation of FCN (fuzzy c-network), U-Net and the like or Mask-RCNN and other example segmentation algorithms, obtaining pixel coordinates of the outline of the vehicle head and the outline of the frame through a post-processing algorithm of the outline of the container 2, and restoring the ROI (region of interest) image of the whole vehicle head and the whole vehicle body sheared by the AI detection algorithm module to the original image to obtain the outline of the container 2 on the original image and the high-precision pixel coordinates of the upper left corner and the lower right corner of the outline of the frame;
s33, converting high-precision pixel coordinates of the upper left corner and the lower right corner of the outline outer frame of the container 2 through a coordinate system conversion algorithm module by using a calibrated camera internal reference matrix M, an external reference rotation matrix R and a translation matrix T to obtain coordinates of the upper left corner and the lower right corner of the container 2 under a crane coordinate system, and calculating to obtain the length of the container 2;
s34, obtaining coordinates of the upper left corner and the lower right corner of the container 2 under a crane coordinate system through the AI partitioning algorithm module obtained by the first judgment module, obtaining the actual length of the container 2 through calculation, and preliminarily judging whether the container 2 is a single 20-foot, single 40-foot, double 20-foot or 45-foot box type;
s35, detecting the box gaps of the containers 2 or the number of lock holes of the containers 2 in the two pairs of pictures through an AI detection algorithm module, inputting the pictures of the divided containers 2 into a pre-training target detection algorithm model of a YOLO deep learning convolution neural network to obtain the detection results of the box gaps of the containers 2 or the number of lock holes of the containers 2;
and S36, obtaining the detection result of the box gap of the container 2 or the number of the lock holes of the container 2 by the AI detection algorithm module II through the judgment module II, namely finally judging that the container 2 is a single 20-foot, single 40-foot, double 20-foot or 45-foot box type.
The training process of the neural network algorithm model of the AI detection algorithm module I, the AI detection algorithm module II and the AI segmentation algorithm module further comprises the following steps:
a) establishing a container truck database;
b) making a container truck training, verifying and testing data set;
c) and (5) building a neural network algorithm model.
In the step a), the database of the container truck is established as follows:
acquiring a video stream after distortion removal of the calibration module 100, acquiring an image of the container truck by decoding, and labeling the image of the container truck by using a labeling tool;
in step b), the data set is created as follows:
extracting a training set, a verification set and a test set from a container truck database;
in the step c), an AI detection and neural network segmentation algorithm model is built as follows:
and training the container truck detection segmentation model by using a training set, judging whether the neural network algorithm models of the AI detection algorithm module I, the AI detection algorithm module II and the AI segmentation algorithm module meet the requirements or not by using a verification set, and testing by using a test set to obtain the trained neural network algorithm pre-training models of the AI detection algorithm module I, the AI detection algorithm module II and the AI segmentation algorithm module.
Examples
Referring to fig. 1, the method and system for identifying a container type of an automated container in a dock according to the present invention includes:
the calibration module 100 selects the industrial camera 1 to perform internal reference calibration and distortion correction in advance before installation, and performs external reference calibration of the camera after installation;
the acquisition module 101 acquires a real-time video stream of the industrial camera 1, decodes the real-time video stream into a picture, and preprocesses the picture;
the algorithm module 102 inputs the preprocessed picture into the algorithm module 102, and a processing flow of the algorithm module 102 mainly includes: 1) an AI detection algorithm module I; 2) an AI partitioning algorithm module; 3) a coordinate system conversion algorithm module; 4) a first judging module; 5) an AI detection algorithm module II, 6) a judgment module II, which finally judges whether the container 2 is a single 20-ruler box type, a single 40-ruler box type, a double 20-ruler box type or a 45-ruler box type through six algorithm processing flows;
the data recording module 103 is mainly used for recording history records, state information and fault information, data storage and the like of each module, and mainly comprises part of original image data.
Referring to fig. 2 to 4, the present invention further provides an automatic dock container box type identification method, which includes the following steps:
1) after the position guidance of the container truck 2 on the operation lane is finished, the calibrated internal reference and distortion removal correction industrial camera 1 is mounted in advance to shoot lane videos on the crane 3.
2) The acquisition module 101 reads a real-time video stream of the industrial camera 1, decodes the real-time video stream into a picture, and preprocesses the picture; specifically, a streaming of the industrial camera 1 based on RTSP (real time streaming protocol) can be read by using a C + + and third-party OpenCV open source library to obtain a picture stream. The image is preprocessed through a filtering and drying algorithm, a graying algorithm, a histogram equalization algorithm and other algorithms, so that subsequent detection and identification are facilitated;
3) the processed pictures are input into the algorithm module 102 to perform six algorithm processing procedures such as detection, segmentation and coordinate system conversion, and finally, whether the container 2 is a single 20-foot box, a single 40-foot box, a double 20-foot box or a 45-foot box is judged.
It should be understood by those skilled in the art that the above embodiments are only for illustrating the present invention and are not to be used as a limitation of the present invention, and that changes and modifications to the above described embodiments are within the scope of the claims of the present invention as long as they are within the spirit and scope of the present invention.
Claims (10)
1. A box-type identification system for an automated quay container, comprising:
the calibration module is used for acquiring a real-time video stream of a field lane and performing distortion removal processing on the real-time video stream;
the acquisition module acquires the real-time video stream, decodes the real-time video stream into a picture and preprocesses the picture;
the algorithm module is used for acquiring the picture and judging that the container is a single 20-ruler box type, a single 40-ruler box type, a double 20-ruler box type or a double 45-ruler box type through algorithm processing;
and the data recording module records the history, state information and fault information of each module and stores data.
2. The system of claim 1, wherein the automated quay container box identification system comprises: the calibration module selects an industrial camera and is arranged on the crane.
3. The system of claim 2, wherein the algorithm module further comprises:
the AI detection algorithm module I is used for realizing the rough positioning of the whole truck head and truck body in the picture;
the AI segmentation algorithm module is used for realizing the pixel level high-precision positioning of the container outline in the picture;
the coordinate system conversion algorithm module is used for calculating the length of the container;
the first judging module is used for preliminarily judging whether the container is a single 20-ruler, single 40-ruler, double 20-ruler or 45-ruler box type;
the AI detection algorithm module II is used for detecting the number of the gaps or the lock holes of the container in the picture;
and the second judging module is used for finally judging that the container is a single 20-ruler, single 40-ruler, double 20-ruler or 45-ruler box type.
4. A box-type identification system for an automated quay container, characterized in that the following steps are performed using the box-type identification system for an automated quay container according to claim 3:
s1, acquiring a real-time video stream of a field lane through the calibration module, and performing distortion removal processing on the real-time video stream;
s2, acquiring the real-time video stream through the acquisition module, decoding the real-time video stream into a picture, and preprocessing the picture;
and S3, inputting the preprocessed pictures into the algorithm module, and judging that the container is a single 20-ruler, single 40-ruler, double 20-ruler or 45-ruler box type.
5. The method of box type identification of an automated quay container of claim 4, wherein: in the step S1, the calibration module selects an industrial camera and is installed on a crane;
before installation, camera internal reference calibration is carried out on the industrial camera;
and after the industrial camera is installed, carrying out camera external parameter calibration on the industrial camera.
6. The method for box type identification of an automated quay container of claim 5, wherein the camera-internal reference calibration further comprises:
and acquiring the internal reference matrix M and distortion parameters K1, K2, K3, P1 and P2 of the industrial camera by using a calibration plate and a Zhang calibration method, and performing distortion removal correction on the industrial camera through the internal reference matrix M and the distortion parameters K1, K2, K3, P1 and P2.
The camera external reference calibration further comprises:
and obtaining a rotation matrix R and a translation matrix T from the camera coordinate system to the crane coordinate system through a PnP algorithm or a nonlinear Gauss-Newton optimization algorithm by utilizing the coordinates of the calibration plate corner points under the crane coordinate system and the coordinates under the camera pixel coordinate system.
7. The method of box type identification of an automated quay container of claim 4, wherein: in step S2, the preprocessing is filtering and de-drying, graying or histogram equalization processing algorithm.
8. The method for box type identification of an automated quay container according to claim 4, wherein said calculation of said algorithm module in step S3 further comprises the steps of:
s31, roughly positioning the truck head and vehicle body in the picture through the AI detection algorithm module, namely inputting the preprocessed picture into a pretrained target detection algorithm model of a YOLO deep learning convolution neural network to obtain a truck head and vehicle body whole vehicle detection result, and then shearing an ROI (region of interest) of the truck head and vehicle body whole vehicle to extract a truck head and vehicle body whole vehicle roughly positioning image;
s32, carrying out pixel-level high-precision positioning on the container outline in the picture through the AI segmentation algorithm module, namely obtaining a pixel-level high-precision positioning image of the container outline of the vehicle head and the vehicle body to be obtained through an AI segmentation algorithm, obtaining pixel coordinates of the vehicle head outline external frame and the upper left corner and the lower right corner of the frame through an external frame post-processing algorithm of the minimum area of the container outline, and restoring the ROI area image of the vehicle head and the vehicle body cut by the AI detection algorithm module to the original image to obtain high-precision pixel coordinates of the upper left corner and the lower right corner of the external frame and the frame of the container outline on the original image through the proportional relation between the ROI area image of the vehicle head and the original image;
s33, converting high-precision pixel coordinates of the upper left corner and the lower right corner of the container outline outer frame through the coordinate system conversion algorithm module through a calibrated camera internal reference matrix M, an external reference rotation matrix R and a translation matrix T to obtain coordinates of the upper left corner and the lower right corner of the container under a crane coordinate system, and calculating to obtain the length of the container;
s34, coordinates of the upper left corner and the lower right corner of the container under a crane coordinate system are obtained through the AI division algorithm module obtained through the first judgment module, the actual length of the container can be obtained through calculation, and the container can be preliminarily judged to be a single 20-ruler box type, a single 40-ruler box type, a double 20-ruler box type or a 45-ruler box type;
s35, detecting the number of container gaps or container lock holes in the picture through the AI detection algorithm module, inputting the segmented container picture into a pre-training target detection algorithm model of a YOLO deep learning convolution neural network to obtain a detection result of the number of the container gaps or the container lock holes;
s36, the AI detection algorithm module II is obtained through the judgment module II to obtain the detection result of the box gap of the container or the number of the lock holes of the container, namely the container is finally judged to be a single 20-foot box type, a single 40-foot box type, a double 20-foot box type or a 45-foot box type.
9. The method of claim 8, wherein the training of the first AI detection algorithm module, the second AI detection algorithm module, and the neural network algorithm model of the AI segmentation algorithm module further comprises:
a) establishing a container truck database;
b) making a container truck training, verifying and testing data set;
c) and building the neural network algorithm model.
10. The method for box type identification of an automated quay container according to claim 9, wherein in step a), the database of container trucks is created as follows:
acquiring the video stream after distortion removal of the calibration module, acquiring an image of the container truck through decoding, and labeling the image of the container truck by using a labeling tool;
in the step b), the data set is specifically made as follows:
extracting a training set, a verification set and a test set from the container truck database;
in the step c), the AI detection and segmentation neural network algorithm model is built as follows:
and training the container truck detection segmentation model by using the training set, judging whether the neural network algorithm models of the first AI detection algorithm module, the second AI detection algorithm module and the AI segmentation algorithm module meet the requirements or not through the verification set, and testing through the test set to obtain the trained neural network algorithm pre-training models of the first AI detection algorithm module, the second AI detection algorithm module and the AI segmentation algorithm module.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210001257.3A CN114299282A (en) | 2022-01-04 | 2022-01-04 | Box type identification system and method for automatic wharf container |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210001257.3A CN114299282A (en) | 2022-01-04 | 2022-01-04 | Box type identification system and method for automatic wharf container |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114299282A true CN114299282A (en) | 2022-04-08 |
Family
ID=80975756
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210001257.3A Pending CN114299282A (en) | 2022-01-04 | 2022-01-04 | Box type identification system and method for automatic wharf container |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114299282A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115170650A (en) * | 2022-07-11 | 2022-10-11 | 深圳市平方科技股份有限公司 | Container vehicle-mounted position identification method and device, electronic equipment and storage medium |
-
2022
- 2022-01-04 CN CN202210001257.3A patent/CN114299282A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115170650A (en) * | 2022-07-11 | 2022-10-11 | 深圳市平方科技股份有限公司 | Container vehicle-mounted position identification method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20220084186A1 (en) | Automated inspection system and associated method for assessing the condition of shipping containers | |
| Mi et al. | A fast automated vision system for container corner casting recognition | |
| CN114359246B (en) | Conveyor belt detection method, device, system, electronic device and medium | |
| CN111415106A (en) | Truck loading rate identification method, device, equipment and storage medium | |
| CN112465706A (en) | Automatic gate container residual inspection method | |
| CN111767780B (en) | AI and vision combined intelligent integrated card positioning method and system | |
| CN111382762A (en) | Empty box identification method and system | |
| CN112881412B (en) | Method for detecting non-metal foreign matters in scrap steel products | |
| CN114022537B (en) | Method for analyzing loading rate and unbalanced loading rate of vehicle in dynamic weighing area | |
| CN114299282A (en) | Box type identification system and method for automatic wharf container | |
| CN116228651A (en) | Cloth defect detection method, system, equipment and medium | |
| CN115527170A (en) | Method and system for identifying closing fault of door stopper handle of automatic freight car derailing brake device | |
| KR20210122429A (en) | Method and System for Artificial Intelligence based Quality Inspection in Manufacturing Process using Machine Vision Deep Learning | |
| CN112308003B (en) | Method, system, equipment and medium for identifying loading state of scrap steel wagon | |
| Taheritanjani et al. | Automatic damage detection of fasteners in overhaul processes | |
| CN112308848A (en) | Method and system for identifying state of baffle plate of scrap steel truck, electronic equipment and medium | |
| Khokhlova et al. | Development of an Intelligent Robotic System for Recognizing and Transporting Postal Objects | |
| CN105740768A (en) | Unmanned forklift device based on combination of global and local features | |
| Khule et al. | Automated object counting for visual inspection applications | |
| CN115830509A (en) | Intelligent truck empty carriage identification auditing method based on video stream characteristics | |
| US12002196B2 (en) | Electronic device and method for marking defects of products | |
| CN112149676A (en) | Small target detection processing method for railway goods loading state image | |
| CN115393855A (en) | License plate product quality detection method, system and equipment | |
| CN109978879B (en) | Box corner in-groove state detection method based on railway wagon loading video monitoring | |
| Evstafev et al. | Surface Defect Detection and Recognition Based on CNN |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |