Deep learning-based shipping container damage detection method and system

Abstract

The present application relates to the technical field of shipping container damage detection, and in particular to a deep learning-based shipping container damage detection method and system. The method comprises: establishing a plurality of shipping container categories, and creating an image analysis model according to all shipping container categories; setting an image acquisition strategy for a shipping container to be inspected, and acquiring, according to the image acquisition strategy, an image data packet of the shipping container to be inspected; generating a preprocessing result of the image data packet according to the image analysis model; generating a first-level inspection strategy for the shipping container to be inspected according to the preprocessing result, and generating a damage risk value of the shipping container to be inspected; and constructing a plurality of shipping container categories on the basis of different shipping container equipment parameters, and constructing a sub-image model for each shipping container category by using deep learning technology, so as to adapt to different types of shipping container damage conditions occurring during actual production operations at different ports, and performing preliminary damage detection by means of acquiring image data of a shipping container to be inspected, thereby improving shipping container damage detection efficiency and reducing detection operating and maintenance costs.

Description

A Deep Learning-Based Method and System for Container Damage Detection Technical Field

[0001] This application relates to the field of container damage detection technology, and in particular to a container damage detection method and system based on deep learning. Background Technology

[0002] With the rapid development of global trade and the continuous growth of port throughput, the requirements for port efficiency are also increasing. As a key unit in port loading and unloading operations, the degree of damage to containers is receiving increasing attention. Some containers have developed damage such as breakage, holes, and corrosion as they have been used for many years.

[0003] The degree of damage to containers is related to the safety of cargo and container ships, therefore, it is crucial to research efficient inspection methods for container damage detection. Traditional container damage detection methods, such as lidar detection, often require sophisticated gate hardware and are slow, increasing container management and operating costs. Summary of the Invention

[0004] The purpose of this application is to provide a container damage detection method and system based on deep learning to solve the above-mentioned technical problems, aiming to improve the efficiency and accuracy of container damage detection and reduce the overall operation and maintenance cost of containers.

[0005] In some embodiments of this application, multiple container categories are constructed based on different container equipment parameters, and sub-image models of each container category are constructed using deep learning technology. This adapts to different types of container damage situations in actual port production processes. Preliminary damage detection is performed by collecting image data of the container to be inspected, thereby improving the efficiency of container damage detection.

[0006] In some embodiments of this application, the image data packets of the container to be inspected are analyzed to make a preliminary judgment on the container to be inspected, and a secondary inspection is performed on the risk sub-regions therein. During the secondary inspection, multi-source data (laser data, infrared data, ultrasonic data, etc.) are collected for auxiliary judgment, thereby timely warning and diagnosis of potential residual risks, improving the diagnostic accuracy of container residuals and reducing the overall inspection and maintenance costs.

[0007] In some embodiments of this application, a deep learning-based method for detecting container damage is provided, including:

[0008] Establish multiple container categories and build an image analysis model based on all container categories;

[0009] Set the image acquisition strategy for the container to be inspected, and acquire the image data packet of the container to be inspected according to the image acquisition strategy;

[0010] Preprocessing results for generating image data packets based on image analysis models;

[0011] Based on the preprocessing results, a primary inspection strategy for the container to be inspected is generated, and a damage risk value for the container to be inspected is generated.

[0012] When setting multiple container categories, including:

[0013] Establish a sequence of container categories A, A=(a1,a2…a3) i …a n ), where a i Let be the i-th container category; n is the number of container categories.

[0014] In some embodiments of this application, when establishing an image analysis model based on all container categories, the following are included:

[0015] Based on the container category sequence A, a is set sequentially. i For the target container category;

[0016] Generate historical data packets for the target container category;

[0017] Generate a detection evaluation value c for the target container category based on historical data packets;

[0018] Based on the detection evaluation value c, the area of ​​the segmented sub-region of the target container category is set, and multiple segmented sub-regions of the target container category are set;

[0019] Establish a sequence B of segmented sub-regions for the target container category, B = (b1, b2, ..., bb2). i …b m ), where b i represents the i-th segmented sub-region in the target container category; m represents the number of segmented sub-regions within the target container category.

[0020] Generate training data packets for each segmented sub-region, and build a sub-image model of the target container category based on all training data packets;

[0021] Generate the expected risk value for each segmented sub-region based on historical data packets, and set the image acquisition amount for each segmented sub-region based on all expected risk values.

[0022] The sub-acquisition strategy for the target container category is set based on the image acquisition volume of each segmented sub-region;

[0023] Sub-image models and sub-acquisition strategies for each container category are set sequentially;

[0024] An image analysis model is established based on the full sub-image model and the full sub-acquisition strategy.

[0025] In some embodiments of this application, the historical data packets generate a detection evaluation value c for the target container category, including:

[0026] ;

[0027] Where e1 is the preset first weighting coefficient; e2 is the preset second weighting coefficient; Q1 is the preset first fixed coefficient; Q2 is the preset second fixed coefficient; θ1 is the number of damage evaluation indicators; α 1i s is the influencing factor of the i-th damage evaluation index; i θ1 represents the reference value for the i-th damage evaluation index generated based on historical data packets of the target container category; θ2 represents the number of equipment evaluation indicators. w is the influencing factor of the i-th equipment evaluation index; i This is the reference value for the i-th equipment evaluation index within the target container category.

[0028] In some embodiments of this application, the expected risk value of each segmented sub-region is generated, including:

[0029] Based on the sequence of sub-regions B, b is set sequentially. i Divide the target into sub-regions;

[0030] Generate the expected risk value d for the target segmented sub-region;

[0031]

[0032] Where θ3 represents the number of regional evaluation indicators; µ i Let j be the influencing factor of the i-th regional evaluation index; i The reference value for the evaluation index of the i-th region in the target segmented sub-region;

[0033] The expected risk value for each segmented sub-region is generated sequentially;

[0034] Establish a sequence of expected risk values ​​D, D=(d1,d2…d i …d m ), where d i is the expected risk value of the i-th segmented sub-region in the target container category; m is the number of segmented sub-regions in the target container category.

[0035] In some embodiments of this application, an image acquisition strategy for the container to be detected is set, including:

[0036] Obtain the characteristic parameters of the container to be inspected;

[0037] Generate the similarity evaluation values between the container to be detected and each container category;

[0038] Establish a sequence P of similarity evaluation values, P = (p1, p2…p i …p n ), where p i is the similarity evaluation value between the container to be detected and the i-th container category; n is the number of container categories;

[0039] ;

[0040] where u is the number of characteristic evaluation indicators; β r is the influence factor of the r-th characteristic evaluation indicator; v r is the reference value of the r-th characteristic evaluation indicator in the container to be detected; v' ir is the reference value of the r-th characteristic evaluation indicator in the i-th container category;

[0041] Set the sub - acquisition strategy of the container category corresponding to the maximum value p max in the sequence P of similarity evaluation values as the image acquisition strategy of the container to be detected.

[0042] In some embodiments of the present application, generate a primary detection strategy for the container to be detected according to the pre - processing result, including:

[0043] Establish a sequence B1 of detection sub - regions of the container to be detected according to the image acquisition strategy of the container to be detected, B1 = (b 11 , b 12 …b 1i …b 1n1 ), where b 1i is the i-th detection sub - region of the container to be detected; n1 is the number of detection sub - regions of the container to be detected;

[0044] Generate the primary damage value of each detection sub - region according to the pre - processing result;

[0045] Establish a sequence F of primary damage values, F = (f1, f2…f i …f n1 ), where fi is the primary damage value of the i-th detection sub - region; n1 is the number of detection sub - regions of the container to be detected;

[0046] Preset a first damage value threshold F1 and a second damage value threshold F2, and F1 < F2; [[ID=5 eight]]

[0047] If f i < F1, set the i-th detection sub - region as a safe sub - region;

[0048] If F1 < f i<F2, set the i-th detection sub-region as a risk sub-region;

[0049] If f i > F2, set the i-th detection sub-region as a damaged sub-region;

[0050] Set the primary detection strategy for the container to be inspected based on all the risk sub-regions.

[0051] In some embodiments of the present application, the primary detection strategy includes:

[0052] Establish a risk sub-region sequence B2, B2=(b 21 , b 22 … b 2i … b 2n2 ), where b 2i is the i-th risk sub-region of the container to be inspected; n2 is the number of risk sub-regions of the container to be inspected;

[0053] Set b 2i as the target risk sub-region in sequence according to the risk sub-region sequence B2;

[0054] Obtain the feedback data packet of the target risk sub-region and generate the secondary damage value g of the target risk sub-region;

[0055] ;

[0056] where x is the number of data categories to be collected for the target risk sub-region; µ i is the influence factor of the i-th type of data; t i is the expected damage value generated based on the collected i-th type of data;

[0057] Generate the secondary damage values of each risk sub-region in sequence.

[0058] In some embodiments of the present application, generating the damage risk value of the container to be inspected includes:

[0059] Generate the damage risk value h based on the secondary damage values of each risk sub-region and all the damaged sub-regions;

[0060] [[ID=�5]] ;

[0061] where e3 is the preset third weight coefficient; e4 is the preset fourth weight coefficient; Q3 is the preset third fixed coefficient; Q4 is the preset fourth fixed coefficient; f 2i is the secondary damage value of the i-th risk sub-region; n2 is the number of risk sub-regions of the container to be inspected; f2i is the secondary damage value of the i-th risk sub-region of the container to be inspected; f' is the threshold of the secondary damage value; Y(i) is the selection coefficient; if (f 2i-f')>0, Y(i)=1; if (f 2i -f')<0, Y(i)=0; θ4 is the number of damaged area indicators; η i z is the influencing factor of the index of the i-th damaged area; i This is the reference value for the index of the i-th damaged area in the container to be inspected;

[0062] Preset damage risk threshold H1;

[0063] If h > H1, generate a damage warning command for the container to be inspected.

[0064] In some embodiments of this application, a container damage detection system based on deep learning is provided, comprising:

[0065] The central control unit is used to establish multiple container categories and build an image analysis model based on all container categories;

[0066] The detection unit includes multiple detection sub-modules, and the detection unit is used to set the image acquisition strategy for the container to be inspected.

[0067] The detection unit is also used to acquire image data packets of the container to be detected according to the image acquisition strategy;

[0068] The central control unit includes:

[0069] The first processing module is used to generate preprocessed results of image data packets based on the image analysis model;

[0070] The second processing module is used to generate a primary inspection strategy for the container to be inspected based on the preprocessing results, and to generate a damage risk value for the container to be inspected.

[0071] The early warning module is used to determine whether to generate a damage early warning command based on the damage risk value;

[0072] The third processing module is used to establish a sequence of container categories A, A=(a1,a2…a…). i …a n ), where a i Let be the i-th container category; n is the number of container categories.

[0073] In some embodiments of this application, the central control unit further includes:

[0074] The fourth processing module is used to build image analysis models based on all container categories;

[0075] The fourth processing module is also used for:

[0076] Based on the container category sequence A, a is set sequentially. i For the target container category;

[0077] Generate historical data packets for the target container category;

[0078] Generate a detection evaluation value c for the target container category based on historical data packets;

[0079] Based on the detection evaluation value c, the area of ​​the segmented sub-region of the target container category is set, and multiple segmented sub-regions of the target container category are set;

[0080] Establish a sequence B of segmented sub-regions for the target container category, B = (b1, b2, ..., bb2). i …b m ), where b i represents the i-th segmented sub-region in the target container category; m represents the number of segmented sub-regions within the target container category.

[0081] Generate training data packets for each segmented sub-region, and build a sub-image model of the target container category based on all training data packets;

[0082] Generate the expected risk value for each segmented sub-region based on historical data packets, and set the image acquisition amount for each segmented sub-region based on all expected risk values.

[0083] The sub-acquisition strategy for the target container category is set based on the image acquisition volume of each segmented sub-region;

[0084] Sub-image models and sub-acquisition strategies for each container category are set sequentially;

[0085] An image analysis model is established based on the full sub-image model and the full sub-acquisition strategy.

[0086] Compared with existing technologies, the container damage detection method and system based on deep learning proposed in this application have the following advantages:

[0087] Multiple container categories are constructed based on different container equipment parameters. Sub-image models for each container category are built using deep learning technology to adapt to different types of container damage in actual port production processes. Preliminary damage detection is performed by collecting image data of the containers to be inspected, thereby improving the efficiency of container damage detection.

[0088] By analyzing the image data packets of the container to be inspected, a preliminary judgment is made on the container, and a secondary inspection is carried out on the risk sub-areas. During the secondary inspection, multi-source data (laser data, infrared data, ultrasonic data, etc.) are collected to assist in the judgment, thereby providing timely warnings and diagnoses of potential residual risks, improving the diagnostic accuracy of container residuals, and reducing the overall inspection and maintenance costs. Attached Figure Description

[0089] Figure 1 is a flowchart of a container damage detection method based on deep learning in a preferred embodiment of this application. Detailed Implementation

[0090] The specific embodiments of this application will be described in further detail below with reference to the accompanying drawings and examples. The following examples are used to illustrate this application, but are not intended to limit the scope of this application.

[0091] In the description of this application, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", 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 application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0092] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0093] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0094] As shown in Figure 1, a preferred embodiment of the container damage detection method based on deep learning in this application includes:

[0095] S101: Establish multiple container categories and build an image analysis model based on all container categories;

[0096] S102: Set the image acquisition strategy for the container to be inspected, and acquire the image data packet of the container to be inspected according to the image acquisition strategy;

[0097] S103: Preprocessing results of generating image data packets based on the image analysis model;

[0098] S104: Generate a primary inspection strategy for the container to be inspected based on the preprocessing results, and generate a damage risk value for the container to be inspected.

[0099] When setting multiple container categories, including:

[0100] Establish a sequence of container categories A, A=(a1,a2…a3) i …a n ), where a i Let be the i-th container category; n is the number of container categories.

[0101] Specifically, when building an image analysis model based on all container categories, the following is included:

[0102] Based on the container category sequence A, a is set sequentially. i For the target container category;

[0103] Generate historical data packets for the target container category;

[0104] Generate a detection evaluation value c for the target container category based on historical data packets;

[0105] Based on the detection evaluation value c, the area of ​​the segmented sub-region of the target container category is set, and multiple segmented sub-regions of the target container category are set;

[0106] Establish a sequence B of segmented sub-regions for the target container category, B = (b1, b2, ..., bb2). i …b m ), where b i represents the i-th segmented sub-region in the target container category; m represents the number of segmented sub-regions within the target container category.

[0107] Generate training data packets for each segmented sub-region, and build a sub-image model of the target container category based on all training data packets;

[0108] Generate the expected risk value for each segmented sub-region based on historical data packets, and set the image acquisition amount for each segmented sub-region based on all expected risk values.

[0109] The sub-acquisition strategy for the target container category is set based on the image acquisition volume of each segmented sub-region;

[0110] Sub-image models and sub-acquisition strategies for each container category are set sequentially;

[0111] An image analysis model is established based on the full sub-image model and the full sub-acquisition strategy.

[0112] Specifically, multiple characteristic evaluation indicators for containers are set, including but not limited to parameters such as container model, service life, and shipping time. Multiple container categories are established by random combination of different value ranges of each characteristic evaluation indicator.

[0113] Specifically, training data packets for each segmented sub-region are generated by capturing images of container damage under different lighting conditions, weather conditions, and angles for each segmented sub-region of the target container category.

[0114] Specifically, the training data package includes various damage scenarios such as severe damage, minor damage, severe holes, minor holes, missing lead seals, and rust. The number of images for each type of damage is kept consistent. Labelimg is used to annotate the collected images, and the annotated image data is processed into a format suitable for training with the Ultralytics framework. The dataset is then divided into training, validation, and test sets in a 7:2:1 ratio.

[0115] Specifically, based on deep learning technology, the initial model of each segmented sub-region is trained, the sub-alignment model of each segmented sub-region is established according to the iteration results, and the sub-image model of the target container category is generated according to all the sub-alignment models.

[0116] It is understandable that in the above embodiments, multiple container categories are constructed based on different container equipment parameters, and sub-image models of each container category are constructed using deep learning technology, thereby adapting to different types of container damage situations in the actual production process of different ports.

[0117] In a preferred embodiment of this application, the historical data packet generates a detection evaluation value c for the target container category, including:

[0118] ;

[0119] Where e1 is the preset first weighting coefficient; e2 is the preset second weighting coefficient; Q1 is the preset first fixed coefficient; Q2 is the preset second fixed coefficient; θ1 is the number of damage evaluation indicators; α 1i s is the influencing factor of the i-th damage evaluation index; i θ1 represents the reference value for the i-th damage evaluation index generated based on historical data packets of the target container category; θ2 represents the number of equipment evaluation indicators. w is the influencing factor of the i-th equipment evaluation index; i This is the reference value for the i-th equipment evaluation index within the target container category.

[0120] Specifically, by presetting a first fixed coefficient and a second fixed coefficient, all parameters in the model are normalized, so that all parameters in the model are within the same range of values.

[0121] Specifically, the survival evaluation indicators include, but are not limited to, multiple parameters such as the frequency of damage to the target container category and the probability of cargo loss after survival.

[0122] Specifically, equipment evaluation indicators include, but are not limited to, multiple parameters such as container area within a container category and frequency of containers of the target container category entering the port.

[0123] Specifically, the higher the inspection evaluation value, the more frequently the corresponding target container category of containers needs to be inspected, and the lower the possibility of damage.

[0124] Specifically, the larger the detection evaluation value, the larger the corresponding segmented sub-region area. By dynamically adjusting the area of ​​the segmented sub-region, the image acquisition volume can be reduced and the image acquisition efficiency can be improved while ensuring the accuracy of damage detection, thereby improving the overall damage detection efficiency and reducing operation and maintenance costs.

[0125] Specifically, the expected risk value for each segmented sub-region is generated, including:

[0126] Based on the sequence of segmented sub-regions B, bi is sequentially set as the target segmented sub-region;

[0127] Generate the expected risk value d for the target segmented sub-region;

[0128]

[0129] Where θ3 represents the number of regional evaluation indicators; µ i Let j be the influencing factor of the i-th regional evaluation index; i The reference value for the evaluation index of the i-th region in the target segmented sub-region;

[0130] The expected risk value for each segmented sub-region is generated sequentially;

[0131] Establish a sequence of expected risk values ​​D, D=(d1,d2…d i …d m ), where d i is the expected risk value of the i-th segmented sub-region in the target container category; m is the number of segmented sub-regions in the target container category.

[0132] Specifically, regional evaluation indicators include, but are not limited to, the probability of damage occurring within the region, the type of damage, the degree of interference with the overall container, and the significance of the damage. The higher the expected risk value, the greater the probability of damage occurring in the current sub-region, and the lower the significance of the damage, the greater the interference with the container.

[0133] Specifically, the higher the saliency, the higher the accuracy of identifying defects within the segmented sub-region through image analysis.

[0134] Specifically, the higher the expected risk value, the greater the corresponding amount of image acquisition, thereby ensuring accurate analysis of the damage status of each segmented sub-region.

[0135] It is understandable that in the above embodiments, by dynamically adjusting the image acquisition strategy for each container category, the image acquisition efficiency is improved while ensuring the accuracy of damage detection, thereby improving the overall damage detection efficiency and accuracy and reducing operation and maintenance costs.

[0136] In a preferred embodiment of this application, an image acquisition strategy for the container to be inspected is set, including:

[0137] Obtain the characteristic parameters of the container to be inspected;

[0138] Generate similarity evaluation values ​​between the container to be detected and each container category;

[0139] Establish a similar evaluation value sequence P, P=(p1, p2…p i …p n ), where p i is the similarity score between the container to be detected and the i-th container category; n is the number of container categories;

[0140] ;

[0141] Where u is the number of feature evaluation indicators; β r v is the influence factor of the r-th feature evaluation index; r Let v' be the reference value for the r-th feature evaluation index in the container to be inspected; ir This is the reference value for the r-th feature evaluation index in the i-th container category;

[0142] Set the maximum value p in the similarity evaluation value sequence P. max The sub-acquisition strategy for the corresponding container category is the image acquisition strategy for the container to be detected.

[0143] Specifically, the characteristic evaluation indicators include, but are not limited to, parameters such as container model, service life, and shipping time.

[0144] Specifically, the larger the similarity evaluation value is, the more suitable the image acquisition strategy and sub-image model within the corresponding container category are for the current container to be detected.

[0145] Specifically, a primary detection strategy for the container to be detected is generated based on the preprocessing result, including:

[0146] A sequence of detection sub-regions B1 for the container to be detected is established according to the image acquisition strategy of the container to be detected, B1 = (b 11 , b 12 …b 1i …b 1n1 ), where b 1i is the i-th detection sub-region of the container to be detected; n1 is the number of detection sub-regions of the container to be detected;

[0147] A primary damage value for each detection sub-region is generated based on the preprocessing result;

[0148] A sequence of primary damage values F is established, F = (f​​1, f2…f i …f n1 ), where f i is the primary damage value of the i-th detection sub-region; n1 is the number of detection sub-regions of the container to be detected;

[0149] A first damage value threshold F1 and a second damage value threshold F2 are preset, and F1 < F2;

[0150] If f i < F1, the i-th detection sub-region is set as a safe sub-region;

[0151] If F1 < f i < F2, the i-th detection sub-region is set as a risk sub-region;

[0152] If f i > F2, the i-th detection sub-region is set as a damaged sub-region;

[0153] A primary detection strategy for the container to be detected is set according to all the risk sub-regions.

[0154] Specifically, the preprocessing result refers to performing regional image analysis on the image data packet of the container to be detected collected by invoking the sub-image analysis model of the corresponding container category through the image acquisition strategy, and identifying the damage parameters within each detection sub-region.

[0155] Specifically, the larger the primary damage value is, the more serious the damage degree within the current segmentation sub-region is. By preprocessing the image data packet, the remaining categories within each damaged sub-region can be determined.

[0156] Specifically, the safe sub-zone refers to the container area that is currently undamaged, the risk sub-zone refers to the area where, based on image analysis, there may be a risk of damage, but the type and extent of damage cannot be accurately determined, and the damaged sub-zone refers to the area where the type of damage can be accurately determined through image analysis, and where the degree of damage is significant.

[0157] It is understood that in the above embodiments, by generating similarity evaluation values ​​between the container to be inspected and various container categories, an image acquisition strategy for the container to be inspected is quickly generated, and preliminary damage detection is performed by acquiring image data of the container to be inspected, thereby improving the efficiency of container damage detection.

[0158] In a preferred embodiment of this application, the primary detection strategy includes:

[0159] Establish a risk sub-region sequence B2, B2=(b 21 ,b 22 …b 2i …b 2n2 ), where b 2i Let be the i-th risk sub-region of the container to be inspected; n2 is the number of risk sub-regions of the container to be inspected.

[0160] Based on the risk sub-region sequence B2, set b sequentially. 2i For the target risk sub-region;

[0161] Obtain the feedback data packet of the target risk sub-region and generate the secondary residual value g of the target risk sub-region;

[0162] ;

[0163] Where x represents the number of data categories to be collected in the target risk sub-region; µ i t is the influence factor for the i-th type of data; i The expected residual value generated based on the collected i-th type of data;

[0164] Secondary residual values ​​for each risk sub-region are generated sequentially.

[0165] Specifically, the data collected includes, but is not limited to, laser feedback data, ultrasonic feedback data, infrared data, and other types of data. By analyzing and processing this data, corresponding expected damage values ​​are generated. Through the analysis of multi-source data, timely warnings can be issued regarding potential damage risks to containers.

[0166] Specifically, the higher the secondary damage value, the more severe the damage within the current risk sub-region.

[0167] Specifically, by conducting comprehensive analysis of multi-source data, damage categories can be generated for each risk sub-region.

[0168] Specifically, generating damage risk values ​​for the containers to be inspected includes:

[0169] A residual risk value h is generated based on the secondary residual value of each risk sub-region and all residual sub-regions;

[0170] ;

[0171] Where e3 is the preset third weighting coefficient; e4 is the preset fourth weighting coefficient; Q3 is the preset third fixed coefficient; Q4 is the preset fourth fixed coefficient; f 2i Let fi be the secondary damage value of the i-th risk sub-region; n2 be the number of risk sub-regions of the container to be inspected; f2i be the secondary damage value of the i-th risk sub-region of the container to be inspected; f' be the secondary damage value threshold; Y(i) be the selection coefficient; if (f 2i -f')>0, Y(i)=1; if (f 2i -f')<0, Y(i)=0; θ4 is the number of damaged area indicators; η i z is the influencing factor of the index of the i-th damaged area; i This is the reference value for the index of the i-th damaged area in the container to be inspected;

[0172] Preset damage risk threshold H1;

[0173] If h > H1, generate a damage warning command for the container to be inspected.

[0174] Specifically, by pre-setting a third and a fourth fixed coefficient, all parameters in the model are normalized, so that all parameters in the model are within the same range of values.

[0175] Specifically, the higher the damage risk value, the greater the possibility of cargo loss in the container.

[0176] Specifically, the damage risk threshold can be set based on historical parameters.

[0177] Specifically, based on the damage warning command, a table of damage categories and damage levels for each detection sub-area can be generated, and corresponding maintenance plans can be set.

[0178] It is understood that in the above embodiments, by analyzing the image data packets of the container to be inspected, a preliminary judgment is made on the container to be inspected, and a secondary inspection is performed on the risk sub-regions. During the secondary inspection, multi-source data (laser data, infrared data, ultrasonic data, etc.) are collected for auxiliary judgment, thereby timely warning and diagnosis of potential residual risks, improving the diagnostic accuracy of container residuals and reducing the overall inspection and maintenance costs.

[0179] In another preferred embodiment of the deep learning-based container damage detection method based on any of the above preferred embodiments, this preferred embodiment provides a deep learning-based container damage detection system, including:

[0180] The central control unit is used to establish multiple container categories and build an image analysis model based on all container categories;

[0181] The detection unit includes multiple detection sub-modules. The detection unit is used to set the image acquisition strategy for the container to be inspected.

[0182] The detection unit is also used to acquire image data packets of the container to be detected according to the image acquisition strategy;

[0183] The central control unit includes:

[0184] The first processing module is used to generate preprocessed results of image data packets based on the image analysis model;

[0185] The second processing module is used to generate a primary inspection strategy for the container to be inspected based on the preprocessing results, and to generate a damage risk value for the container to be inspected.

[0186] The early warning module is used to determine whether to generate a damage early warning command based on the damage risk value;

[0187] The third processing module is used to establish a sequence of container categories A, A=(a1,a2…a…). i …a n ), where a i Let be the i-th container category; n is the number of container categories.

[0188] Specifically, the detection unit is preferably an unmanned aerial vehicle (UAV) device, which is equipped with various data acquisition devices such as image acquisition devices, laser data acquisition devices, infrared devices, and ultrasonic acquisition devices.

[0189] Specifically, the central control unit also includes:

[0190] The fourth processing module is used to build image analysis models based on all container categories;

[0191] The fourth processing module is also used for:

[0192] Based on the container category sequence A, a is set sequentially. i For the target container category;

[0193] Generate historical data packets for the target container category;

[0194] Generate a detection evaluation value c for the target container category based on historical data packets;

[0195] Based on the detection evaluation value c, the area of ​​the segmented sub-region of the target container category is set, and multiple segmented sub-regions of the target container category are set;

[0196] Establish a sequence B of segmented sub-regions for the target container category, B = (b1, b2, ..., bb2). i …b m ), where b i represents the i-th segmented sub-region in the target container category; m represents the number of segmented sub-regions within the target container category.

[0197] Generate training data packets for each segmented sub-region, and build a sub-image model of the target container category based on all training data packets;

[0198] Generate the expected risk value for each segmented sub-region based on historical data packets, and set the image acquisition amount for each segmented sub-region based on all expected risk values.

[0199] The sub-acquisition strategy for the target container category is set based on the image acquisition volume of each segmented sub-region;

[0200] Sub-image models and sub-acquisition strategies for each container category are set sequentially;

[0201] An image analysis model is established based on the full sub-image model and the full sub-acquisition strategy.

[0202] According to the first concept of this application, multiple container categories are constructed based on different container equipment parameters, and sub-image models of each container category are constructed using deep learning technology. This adapts to different types of container damage in actual port production processes. Preliminary damage detection is performed by collecting image data of the container to be inspected, thereby improving the efficiency of container damage detection.

[0203] According to the second concept of this application, by analyzing the image data package of the container to be inspected, a preliminary judgment is made on the container to be inspected, and a secondary inspection is carried out on the risk sub-regions. During the secondary inspection, multi-source data (laser data, infrared data, ultrasonic data, etc.) are collected for auxiliary judgment, thereby timely warning and diagnosis of potential residual risks, improving the diagnostic accuracy of container residuals and reducing the overall inspection and maintenance costs.

[0204] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and substitutions can be made without departing from the technical principles of this application, and these improvements and substitutions should also be considered within the scope of protection of this application.

Claims

1. A deep learning-based method for detecting container damage, characterized in that, include: Establish multiple container categories and build an image analysis model based on all container categories; Set the image acquisition strategy for the container to be inspected, and acquire the image data packet of the container to be inspected according to the image acquisition strategy; Preprocessing results for generating image data packets based on image analysis models; Based on the preprocessing results, a primary inspection strategy for the container to be inspected is generated, and a damage risk value for the container to be inspected is generated. When setting multiple container categories, including: Establish a sequence of container categories A, A=(a1,a2…a3) i …a n ), where a i Let n be the i-th container category; n is the number of container categories. The primary detection strategy includes: The first-level residual value of each detection sub-region is generated based on the preprocessing results; The secondary residual values ​​for each risk sub-region are generated sequentially; The residual risk value h is generated based on the secondary residual value of each risk sub-region and all residual sub-regions; When building an image analysis model based on all container categories, the following are included: Based on the container category sequence A, a is set sequentially. i For the target container category; Generate historical data packets for the target container category; Generate a detection evaluation value c for the target container category based on historical data packets; Based on the detection evaluation value c, the area of ​​the segmented sub-region of the target container category is set, and multiple segmented sub-regions of the target container category are set; Establish a sequence B of segmented sub-regions for the target container category, B = (b1, b2, ..., bb2). i …b m ), where b i represents the i-th segmented sub-region in the target container category; m represents the number of segmented sub-regions within the target container category. Generate training data packets for each segmented sub-region, and build a sub-image model for the target container category based on all training data packets; Generate the expected risk value for each segmented sub-region based on historical data packets, and set the image acquisition amount for each segmented sub-region based on all expected risk values. The sub-acquisition strategy for the target container category is set based on the image acquisition volume of each segmented sub-region; Sub-image models and sub-acquisition strategies for each container category are set sequentially; An image analysis model is established based on the full sub-image model and the full sub-acquisition strategy.

2. The container damage detection method based on deep learning as described in claim 1, characterized in that, Historical data packets generate a detection evaluation value c for the target container category, including: ； Where e1 is the preset first weighting coefficient; e2 is the preset second weighting coefficient; Q1 is the preset first fixed coefficient; Q2 is the preset second fixed coefficient; θ1 is the number of damage evaluation indicators; α 1i s is the influencing factor of the i-th damage evaluation index; i θ2 is the reference value for the i-th damage evaluation index generated based on historical data packets of the target container category; α is the number of equipment evaluation indicators; 2i w is the influencing factor of the i-th equipment evaluation index; i This is the reference value for the i-th equipment evaluation index within the target container category.

3. The container damage detection method based on deep learning as described in claim 2, characterized in that, Generate the expected risk value for each segmented sub-region, including: Based on the sequence of sub-regions B, b is set sequentially. i Divide the target into sub-regions; Generate the expected risk value d for the target segmented sub-region; ； Where θ3 represents the number of regional evaluation indicators; µ i Let j be the influencing factor of the i-th regional evaluation index; i The reference value for the evaluation index of the i-th region in the target segmented sub-region; The expected risk value for each segmented sub-region is generated sequentially; Establish a sequence of expected risk values ​​D, D=(d1,d2…d i …d m ), where d i is the expected risk value of the i-th segmented sub-region in the target container category; m is the number of segmented sub-regions in the target container category.

4. The container damage detection method based on deep learning as described in claim 1, characterized in that, Define the image acquisition strategy for the container to be inspected, including: Obtain the characteristic parameters of the container to be inspected; Generate similarity evaluation values ​​between the container to be detected and each container category; Establish a similar evaluation value sequence P, P=(p1, p2…p i …p n ), where p i is the similarity score between the container to be detected and the i-th container category; n is the number of container categories; ； Where u is the number of feature evaluation indicators; β r v is the influence factor of the r-th feature evaluation index; r v' is the reference value for the r-th feature evaluation index in the container to be inspected; ir This is the reference value for the r-th feature evaluation index in the i-th container category; Set the maximum value p in the similarity evaluation value sequence P. max The sub-acquisition strategy for the corresponding container category is the image acquisition strategy for the container to be detected.

5. The container damage detection method based on deep learning as described in claim 4, characterized in that, Based on the preprocessing results, a primary inspection strategy is generated for the container to be inspected, including: Based on the image acquisition strategy of the container to be inspected, a sequence of detection sub-regions B1, B1=(b 11 ,b 12 …b 1i …b 1n1 ), where b 1i Let be the i-th inspection sub-region of the container to be inspected; n1 is the number of inspection sub-regions of the container to be inspected. The first-level residual value of each detection sub-region is generated based on the preprocessing results; Establish a first-order residual value sequence F, F=(f1,f2…f i …f n1 ), where f i Let n1 be the first-level damage value of the i-th inspection sub-region; n1 is the number of inspection sub-regions of the container to be inspected. A first residual value threshold F1 and a second residual value threshold F2 are preset, and F1 <F2; If f i <F1, set the i-th detected sub-region as a safe sub-region; If F1 < f i <F2, set the i-th detection sub-region as a risk sub-region; If f i >F2, define the i-th detection sub-region as the damaged sub-region; The primary inspection strategy for containers to be inspected is set based on all risk sub-regions.

6. The container damage detection method based on deep learning as described in claim 5, characterized in that, The primary detection strategy includes: Establish a risk sub-region sequence B2, B2=(b 21 ,b 22 …b 2i …b 2n2 ), where b 2i Let be the i-th risk sub-region of the container to be inspected; n2 is the number of risk sub-regions of the container to be inspected. Based on the risk sub-region sequence B2, set b sequentially. 2i For the target risk sub-region; Obtain the feedback data packet of the target risk sub-region and generate the secondary residual value g of the target risk sub-region; ； Where x represents the number of data categories to be collected in the target risk sub-region; µ i t is the influence factor for the i-th type of data; i The expected residual value generated based on the collected i-th type of data; Secondary residual values ​​for each risk sub-region are generated sequentially.

7. The container damage detection method based on deep learning as described in claim 6, characterized in that, Generate damage risk values ​​for the containers to be inspected, including: A residual risk value h is generated based on the secondary residual value of each risk sub-region and all residual sub-regions; ； Where e3 is the preset third weighting coefficient; e4 is the preset fourth weighting coefficient; Q3 is the preset third fixed coefficient; Q4 is the preset fourth fixed coefficient; f 2i Let fi be the secondary damage value of the i-th risk sub-region; n2 be the number of risk sub-regions of the container to be inspected; f2i be the secondary damage value of the i-th risk sub-region of the container to be inspected; f' be the secondary damage value threshold; Y(i) be the selection coefficient; if (f 2i -f')>0, Y(i)=1; if (f 2i -f')<0, Y(i)=0; θ4 is the number of damaged area indicators; η i z is the influencing factor of the index of the i-th damaged area; i This is the reference value for the index of the i-th damaged area in the container to be inspected; Preset damage risk threshold H1; If h > H1, generate a damage warning command for the container to be inspected.

8. A container damage detection system based on deep learning, employing the container damage detection method based on deep learning as described in any one of claims 1-7, characterized in that, include: The central control unit is used to establish multiple container categories and build an image analysis model based on all container categories; The detection unit includes multiple detection sub-modules, and the detection unit is used to set the image acquisition strategy for the container to be inspected. The detection unit is also used to acquire image data packets of the container to be detected according to the image acquisition strategy; The central control unit includes: The first processing module is used to generate preprocessed results of image data packets based on the image analysis model; The second processing module is used to generate a primary inspection strategy for the container to be inspected based on the preprocessing results, and to generate a damage risk value for the container to be inspected. The early warning module is used to determine whether to generate a damage early warning command based on the damage risk value; The third processing module is used to establish a sequence of container categories A, A=(a1,a2…a ... i …a n ), where a i Let be the i-th container category; n is the number of container categories.

9. The container damage detection system based on deep learning as described in claim 8, characterized in that, The central control unit also includes: The fourth processing module is used to build image analysis models based on all container categories; The fourth processing module is also used for: Based on the container category sequence A, a is set sequentially. i For the target container category; Generate historical data packets for the target container category; Generate a detection evaluation value c for the target container category based on historical data packets; Based on the detection evaluation value c, the area of ​​the segmented sub-region of the target container category is set, and multiple segmented sub-regions of the target container category are set; Establish a sequence B of segmented sub-regions for the target container category, B = (b1, b2, ..., bb2). i …b m ), where b i represents the i-th segmented sub-region in the target container category; m represents the number of segmented sub-regions within the target container category. Generate training data packets for each segmented sub-region, and build a sub-image model for the target container category based on all training data packets; Generate the expected risk value for each segmented sub-region based on historical data packets, and set the image acquisition amount for each segmented sub-region based on all expected risk values. The sub-acquisition strategy for the target container category is set based on the image acquisition volume of each segmented sub-region; Sub-image models and sub-acquisition strategies for each container category are set sequentially; An image analysis model is established based on the full sub-image model and the full sub-acquisition strategy.

Images