Goods location identification goods warehouse management system based on laser radar

Abstract

The invention relates to the technical field of laser radar identification and cargo warehouse management, and provides a cargo location identification cargo warehouse management system based on a laser radar, and the system is characterized in that a laser radar module obtains the data information of cargoes and cargo locations in a warehouse; the data analysis module analyzes and processes the goods allocation data information acquired by the laser radar module, and identifies and classifies the goods; the dynamic adjustment module dynamically allocates the identified and classified results according to the cargo allocation data information acquired by the laser radar module; the system management module is connected with the dynamic adjustment module and is used for configuring and managing the system, including adjusting system parameters and providing a user interface; according to the invention, the data analysis module carries out weighted fusion based on the features of the three-dimensional point cloud and the image features, and carries out recognition and classification with historical data, thereby achieving the accurate recognition of the appearance of the goods, achieving the dynamic adjustment of the module for dynamic distribution, improving the efficiency and accuracy of warehouse management, and solving a problem that the space utilization rate is reduced.

Description

Technical Field

[0001] The present invention relates to the field of laser radar recognition technology and cargo storage management technology, and in particular to a cargo storage management system for identifying cargo locations based on laser radar. Background Art

[0002] With the rapid development of the logistics industry and the continuous expansion of enterprise scale, the requirements for warehouse management are getting higher and higher. In this context, modern warehouse management technology came into being. The emergence of advanced technologies such as radio frequency identification (RFID) technology, warehouse management system (WMS) and automated warehousing system provides an effective way to solve the problems in traditional warehouse management. These technologies can improve the efficiency, accuracy and intelligence of warehouse management and meet the growing business needs of enterprises.

[0003] Advances in laser radar technology, Internet of Things technology, sensor technology, and data analysis and processing technology have provided support for efficient and accurate warehouse management. Laser radar emits a laser beam and receives the laser signal reflected from the target object, measuring the flight time, intensity, angle and other information of the laser signal to determine the distance, shape, position and motion state of the target object. In warehouse management, laser radar can be installed on the ceiling, wall or shelf of the warehouse to scan and monitor the cargo space and goods.

[0004] The existing Chinese invention patent (patent publication number: CN118365250A) is that this method installs IoT devices in the storage area that needs to be identified and conducts real-time monitoring, and abstractly defines the storage area that needs to be identified, the shelves in the storage area, and each cargo position in the shelves. Based on the abstract definition of the storage area that needs to be identified, the shelves in the storage area, and each cargo position in the shelves, and the types of items in the storage area, the storage area is divided, and each warehoused cargo is allocated by an ant colony algorithm, and the warehoused cargo is allocated to the corresponding cargo position. After the warehoused cargo, each cargo position is accurately located by position fingerprint positioning, and the status of the corresponding located cargo is queried, thereby improving the efficiency and accuracy of identifying the cargo position status. This solution can complete the classification of cargo when the cargo is put into storage and the allocation of cargo positions when the cargo is stored, but it still has certain shortcomings in the following aspects.

[0005] 1. In actual warehousing operations, goods may be damaged due to improper operations during transportation and handling. When goods are damaged, their appearance may change significantly, which will affect the correct identification and inbound and outbound management of the goods, thereby affecting the inbound and outbound process of goods and reducing the efficiency and accuracy of warehousing management.

[0006] 2. The storage planning of cargo space lacks a dynamic adjustment mechanism. With the entry and exit of goods and changes in inventory, the use of cargo space is constantly changing. However, for goods that have already entered the warehouse, when the use of cargo space changes, if no adjustment is made, some cargo spaces may be idle for a long time, while other cargo spaces are overcrowded. The overall storage space cannot be reasonably allocated and fully utilized, and the cargo spaces cannot be adjusted and optimized in time, which may lead to a decrease in space utilization, and then some cargo spaces are idle for a long time, while other cargo spaces are overcrowded. Summary of the invention

[0007] In order to solve the above technical problems, the present invention provides a cargo warehouse management system for identifying cargo locations based on laser radar, which solves the problems in the prior art by fusing the three-dimensional point cloud features and image features of the cargo for identification and classification, and performing dynamic cargo location allocation and adjustment.

[0008] The technical solution of the present invention is as follows:

[0009] A cargo storage management system for identifying cargo locations based on laser radar, comprising:

[0010] A laser radar module, which is used to scan and collect cargo locations and cargo in the warehouse to obtain data information about the cargo locations;

[0011] A data analysis module, which is data-connected to the laser radar module and is used to analyze and process the cargo location data information acquired by the laser radar module and identify and classify the cargo;

[0012] A dynamic adjustment module, which is data-connected to the data analysis module and the laser radar module, is used to dynamically allocate goods according to the cargo location data information obtained by the laser radar module and the results of identification and classification by the data analysis module, monitor the in-and-out of goods and inventory changes, and dynamically adjust the cargo location;

[0013] The system management module is data-connected with the dynamic adjustment module and is used to configure and manage the system, including adjusting system parameters and providing an intuitive user interface.

[0014] Preferably, the laser radar module includes a laser radar device and an image acquisition device, which are used to obtain three-dimensional point cloud information and image information of the cargo location respectively;

[0015] By setting appropriate parameters and frequency, the lidar equipment can automatically adjust the scanning speed according to the frequency of goods entering and leaving the warehouse. When the frequency of entering and leaving the warehouse increases, the scanning rate becomes faster. At the same time, the image acquisition device can realize synchronous data acquisition.

[0016] Preferably, the data analysis module includes an analysis and identification unit, a database unit and a classification unit;

[0017] The analysis and identification unit is used to analyze the information obtained by the laser radar module, perform weighted fusion based on the features of the three-dimensional point cloud and the image features, and compare and identify the historical cargo feature data;

[0018] The database unit is used to collect information acquired by the historical laser radar module, establish a database of feature classification, automatically collect historical data when there is no historical data, and automatically enrich and improve the cargo feature data;

[0019] The classification unit is used to classify the goods scanned in real time by the laser radar according to the comparison between the analysis and recognition unit result and the database unit feature classification.

[0020] Preferably, the analysis and identification unit realizes identification and classification of goods based on the fusion of main curvature features and deep learning features of goods;

[0021] According to the information of the goods scanned by the laser radar module, the point cloud data set of the goods is obtained: P = {P1, P2, ..., P w}, for the set of these point cloud data P = {P1, P2, ..., P w}, w is the total number of these points, let a be the index of the point, P a It means the ath point, and the principal curvature is used to describe the curvature of the point cloud data surface;

[0022] First, calculate the local neighborhood of the point cloud. For each point P a , determine the specified number of neighboring points around it to form a local neighborhood N(P a ); Then, calculate the covariance matrix C of the local neighborhood:

[0023]

[0024] Where * represents the transpose of the matrix, k is the number of neighborhood points, is the average coordinate of the neighborhood points, which represents an important intermediate quantity in the covariance matrix of the calculated point cloud and is used to describe the overall position information of the neighborhood points. The calculation formula is:

[0025]

[0026] Perform eigenvalue decomposition on the covariance matrix C. The covariance matrix C is a μ×μ matrix, where μ is the dimension of the matrix, indicating that there are μ diagonal elements in the matrix, which means the number of coordinate components of point p. The eigenvalue decomposition formula of the covariance matrix is:

[0027] C=QΛQ -1

[0028] Among them, the orthogonal matrix Q refers to a square matrix whose transpose is equal to its inverse matrix, and Λ is a diagonal matrix composed of eigenvalues. In the eigenvalue decomposition, this formula holds true, where the diagonal elements (λ1, λ2, ..., λμ) of Λ are the eigenvalues ​​of the matrix C;

[0029] The eigenvalue decomposition is arranged from large to small to obtain the eigenvalues ​​λ1, λ2, ..., and the principal curvatures k1 and k2 are calculated by the following formula: k1 = λ1, k2 = λ2;

[0030] In the analysis of goods with incomplete regular shapes, the principal curvature is equal to the eigenvalue because the eigenvalue reflects the degree of change of the local surface of the point cloud in different directions and is closely related to the curvature of the surface. The two principal curvatures are sufficient to describe the curvature of the surface at that point.

[0031] Extract features from the cargo image, assuming that the image feature vector extracted by the deep neural network is γ;

[0032] The curvature features and deep learning feature vectors are fused using a weighted fusion method. The fused feature vector is (ω1k,ω2γ), where k = {k1, k2} is the principal curvature of the point cloud, and ω1 and ω2 are weights, which are set according to the importance of the curvature features and image features.

[0033] Preferably, the dynamic adjustment module includes a real-time storage database unit, an analysis unit and an adjustment unit;

[0034] The real-time warehouse database unit establishes a database based on the cargo information acquired by the laser radar module through real-time scanning, including the real-time cargo occupancy and cargo quantity of the warehouse;

[0035] The analysis unit calculates and determines the in-and-out frequency and storage requirements of different goods according to the real-time transformation in the real-time warehouse database unit, and allocates the most suitable cargo space for the goods according to this information, and automatically generates a dynamic adjustment plan by monitoring the use of cargo space in real time;

[0036] The adjustment unit is used to perform real-time dynamic adjustment according to the dynamic adjustment scheme generated by the analysis unit.

[0037] Preferably, the analysis unit sets a time interval of Δt to perform laser radar scanning, updates the cargo location status and cargo information, and t represents time; randomly generates an allocation plan based on the current cargo location status and cargo information, regards each cargo location allocation plan as a particle, and finds the optimal cargo location allocation strategy through information sharing and collaboration between particles;

[0038] The volume of cargo i obtained by laser radar scanning is H i , the capacity of cargo space j is Lj ; i represents the i-th cargo, j represents the j-th cargo location;

[0039] Record the cargo position status, using binary variable 0 j Indicates the occupancy status of storage location j, 0 j =1 means occupied, 0 means j =0 means free; calculate the frequency of in and out storage and storage demand assessment;

[0040] Calculate the frequency of inbound and outbound storage, and count the number of inbound and outbound times of item i in the time interval [t1, t2] R in and the number of shipments R out , the frequency of in and out storage is D i :

[0041]

[0042] Storage requirement calculation, considering the volume H of the cargo i and special storage requirement factor O i , set according to the classification of goods, storage demand S i equal:

[0043]

[0044] Encode the cargo location allocation plan as a particle. Suppose there are n cargo locations and m types of goods in the warehouse. Use an integer array of length m to represent a particle. Because each element corresponds to the cargo location number of a type of goods, the array length is equal to the number of types of goods. Each element represents the cargo location number assigned to the corresponding goods. A certain number of particles are randomly generated as the initial particle group. Use an array to represent the particles. Each element in the array is associated with a type of goods. If there are m types of goods, an array of length m is required to record the cargo location information of each type of goods.

[0045] Assume that the weight of the frequency of in-and-out storage is α, and the weight of the storage demand is β; the initial particle swarm corresponds to the fitness function E of the cargo allocation plan. The fitness function E is defined as the inverse of the sum of the comprehensive consideration values ​​of all goods. The higher the fitness value, the better the cargo allocation plan, that is:

[0046]

[0047] Each particle has a position vector and a velocity vector. Let x θi represents the cargo location number assigned to the i-th cargo in the θ-th particle, v θi Indicates the change speed of the cargo location number of the ith cargo in the θth particle; particle update, for each particle, v represents the speed, x represents the position, and its speed and position are updated according to the formula:

[0048] Speed ​​update formula:

[0049]

[0050] in, is the speed of change of the cargo location number of the i-th cargo allocation of particle θ in the t+1th iteration. The speed update is to adjust the search step and direction of the particle to find more fitness functions. B is the inertia weight. f1 and f2 are learning factors used to control the step size of the algorithm search. The initial value is randomly set and adjusted according to the performance of the algorithm operation. For the inertia weight B, it is taken between [0,1]. The algorithm calculation speed is fast. Appropriately reduce B to make the movement of the particle more dependent on the current optimal information. For the learning factor, when the algorithm finds that it is too fast during the search process, appropriately reduce the learning factor to reduce the search step size;

[0051] r1 and r2 are random numbers in the interval [0, 1], which are used to prevent the algorithm from stagnating and falling into the local optimal solution. It is a strategy to increase the randomness of the algorithm. θi is the best position that particle θ has ever experienced, the cargo location number assigned to the i-th cargo in the individual optimal solution. These particles constitute a particle swarm, which is the basic unit for the algorithm to search and optimize. i is the best position experienced by the entire initial particle swarm, and the cargo location number assigned to the i-th cargo in the global optimal solution;

[0052] Position update formula:

[0053]

[0054] in is the cargo location number assigned to the jth cargo by particle θ in the t+1th iteration. The fitness function is calculated for each updated location. When the fitness function no longer changes, the algorithm stops.

[0055] When an idle position 0 is detected j =0, select particles with high fitness from the current particle swarm, analyze the goods with low frequency of entry and exit or small storage demand, adjust these goods to deep idle cargo spaces, and update the cargo space allocation plan represented by the particles.

[0056] Preferably, the system management module includes a setting unit and a display unit;

[0057] The setting unit sets the authority levels of different users and provides a system setting function, allowing managers to adjust system parameters according to actual needs.

[0058] Preferably, the display unit provides a visual interface, which displays data in the database unit and the real-time storage database unit, allowing managers to view warehouse status, cargo information, and cargo space usage.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] 1. The present invention analyzes and processes the cargo location data obtained by the laser radar module through a data analysis module, performs weighted fusion based on the features of the three-dimensional point cloud and the image features, and compares and identifies the data with the historical cargo feature data, thereby achieving accurate identification of changes in the appearance of the cargo, solving the problem that changes in the appearance of the cargo due to damage to the cargo affect the correct judgment and in-and-out management, and improving the efficiency and accuracy of warehouse management.

[0061] 2. The present invention uses a dynamic adjustment module to dynamically allocate cargo location information based on the cargo location information obtained by the laser radar module and the results of identification and classification by the data analysis module, monitor the entry and exit of goods and inventory changes, and dynamically adjust the cargo location, thereby realizing dynamic adjustment of cargo location storage planning and solving the problem of decreased space utilization due to the lack of a dynamic adjustment mechanism in cargo location storage planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a work flow chart of the present invention;

[0063] Figure 2 It is a workflow diagram of the analysis and identification unit of the present invention;

[0064] Figure 3 It is a workflow diagram of the analysis unit of the present invention. DETAILED DESCRIPTION

[0065] The following embodiments of the present invention are described in further detail in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0066] The present invention provides a cargo storage management system for identifying cargo locations based on laser radar, comprising:

[0067] LiDAR module: The LiDAR module is used to scan and collect the cargo spaces and cargo in the warehouse and obtain data information about the cargo spaces and cargo spaces;

[0068] The data analysis module is connected to the laser radar module and is used to analyze and process the cargo location data information obtained by the laser radar module and identify and classify the cargo;

[0069] The dynamic adjustment module is connected with the data analysis module and the laser radar module to dynamically allocate the cargo location data information obtained by the laser radar module and the results of the data analysis module identification and classification, monitor the in-and-out and inventory changes of the cargo, and dynamically adjust the cargo location;

[0070] The system management module is connected to the dynamic adjustment module for configuring and managing the system, including adjusting system parameters and providing an intuitive user interface.

[0071] Embodiment 1: Figure 1-3 As shown, in this embodiment, taking a storage center as an example, with the continuous development of business, warehouse management faces the problem that damaged goods affect identification and classification, resulting in low storage efficiency.

[0072] A lidar module is installed in the warehouse, including a lidar device and an image acquisition device. The lidar device scans the cargo space and cargo with appropriate parameters and frequency to generate three-dimensional point cloud information of the cargo space; the image acquisition device simultaneously acquires image information of the cargo.

[0073] The laser radar equipment can automatically adjust the scanning speed according to the frequency of goods entering and leaving the warehouse. When the frequency of entering and leaving the warehouse increases, the scanning rate becomes faster. At the same time, the image acquisition device can synchronously collect image data of the scanned goods.

[0074] The data analysis module includes an analysis and recognition unit, a database unit and a classification unit;

[0075] The analysis and identification unit is used to analyze the information obtained by the laser radar module, perform weighted fusion based on the features of the three-dimensional point cloud and the image features, and compare and identify them with historical cargo feature data;

[0076] The database unit is used to collect information obtained by the historical laser radar module, establish a database of feature classification, automatically collect information when there is no historical data, and automatically enrich and improve the cargo feature data. If the new data contains feature information that has not been fully extracted before or more accurate feature information, the database unit will supplement and update the original cargo feature data;

[0077] The classification unit is used to compare features based on the analysis and recognition unit and the database unit, and then classify according to the feature category; for example, shape-related features, color or texture-related features;

[0078] The analysis and recognition unit recognizes and classifies the goods based on the fusion of the main curvature features and deep learning features of the goods. The lidar equipment obtains point cloud data of the surface of the goods and uses the point cloud processing algorithm to obtain the curvature features of each point on the surface of the goods. The deep learning features are usually extracted using a deep neural network model.

[0079] According to the information of the goods scanned by the laser radar module, the point cloud data set of the goods is obtained: P = {P1, P2, ..., P w}, for the set of these point cloud data P = {P1, P2, ..., P w}, w is the total number of these points, let a be the index of the point, P a It means the ath point, and the principal curvature is used to describe the curvature of the point cloud surface;

[0080] First, calculate the local neighborhood of the point cloud. For each point P a , determine the specified number of neighboring points around it to form a local neighborhood N(P a ); Then, calculate the covariance matrix C of the local neighborhood:

[0081]

[0082] Where * represents the transpose of the matrix, k is the number of neighborhood points, is the average coordinate of the neighborhood points, which represents an important intermediate quantity in the covariance matrix of the calculated point cloud and is used to describe the overall position information of the neighborhood points. The calculation formula is:

[0083]

[0084] Perform eigenvalue decomposition on the covariance matrix C. The covariance matrix C is a μ×μ matrix, where μ is the dimension of the matrix, indicating that there are μ diagonal elements in the matrix, which means the number of coordinate components of point p. The eigenvalue decomposition formula of the covariance matrix is:

[0085] C=QΛQ -1

[0086] Among them, the orthogonal matrix Q refers to a square matrix whose transpose is equal to its inverse matrix, and Λ is a diagonal matrix composed of eigenvalues. In the eigenvalue decomposition, this formula holds true, where the diagonal elements (λ1, λ2, ..., λμ) of Λ are the eigenvalues ​​of the matrix C;

[0087] The eigenvalue decomposition is arranged from large to small to obtain the eigenvalues ​​λ1, λ2, ..., and the principal curvatures k1 and k2 are calculated by the following formula: k1 = λ1, k2 = λ2;

[0088] In the analysis of goods with incomplete regular shapes, the principal curvature is equal to the eigenvalue because the eigenvalue reflects the degree of change of the local surface of the point cloud in different directions and is closely related to the curvature of the surface. The two principal curvatures are sufficient to describe the curvature of the surface at that point.

[0089] Extract features from the cargo image, assuming that the image feature vector extracted by the deep neural network is γ;

[0090] The curvature feature and the deep learning feature vector are fused by using a weighted fusion method. The fused feature vector is (ω1k,ω2γ), where k = {k1, k2} is the principal curvature of the point cloud, and ω1 and ω2 are weights, which are set according to the importance of the curvature feature and the image feature.

[0091] For example, for a newly-stocked product, its packaging may have been dented during transportation. Through the collaborative work of the LiDAR module and the image acquisition device, the data analysis module can accurately extract the 3D point cloud features and image features of the product, and after weighted fusion, compare them with the historical product features in the database. Although the appearance has changed, the type of the product can still be accurately identified to ensure the smooth progress of the warehousing process;

[0092] The data analysis module can identify goods more accurately through weighted fusion of three-dimensional point cloud features and image features, especially when the appearance of the goods changes due to damage during transportation and handling. The module can still accurately determine the type of goods. During the whole process, the lidar module obtains point cloud data on the surface of the goods, and obtains the principal curvature feature through a series of operations such as calculating the local neighborhood and eigenvalue decomposition of the covariance matrix. It reflects geometric information such as the degree of curvature of the surface of the goods. At the same time, a deep neural network model is used to extract feature vectors from the cargo image.

[0093] The two features are weighted and fused to form a new fused feature vector. The existing technology may rely only on a single feature or traditional recognition method. When the appearance of the goods is damaged or deformed due to transportation and handling, a single feature may not be able to accurately determine the type of goods. In this solution, the three-dimensional point cloud feature provides information from the geometric shape, and the image feature provides information from the visual appearance. After the weighted fusion of the two, even if the appearance of the goods changes, the type of goods can still be comprehensively judged from multiple angles. By comparing with the rich historical goods feature data in the database, the accuracy of goods identification in warehouse management is improved. Compared with the existing technology, this solution can better cope with the changes in the appearance of goods and improve the accuracy of warehouse management.

[0094] Embodiment 2: Figure 1-3 As shown, in this embodiment, as the business of a certain storage center changes, the storage cargo space situation also changes continuously, resulting in cargo accumulation in some cargo spaces, while other cargo spaces are idle and vacant for a long time. The cargo space planning lacks dynamic adjustment and the space utilization rate decreases.

[0095] The database unit collects information obtained by the historical lidar module, establishes a feature classification database, and automatically collects historical data when there is no historical data, continuously enriching and improving cargo feature data and improving recognition accuracy and efficiency.

[0096] The dynamic adjustment module includes a real-time storage database unit, an analysis unit and an adjustment unit;

[0097] The real-time warehouse database unit establishes a database based on the cargo information acquired by the laser radar module through real-time scanning, including the real-time cargo occupancy and cargo quantity of the warehouse;

[0098] The analysis unit calculates the frequency of entry and exit and storage requirements of different goods according to the real-time transformation in the real-time warehouse database unit. Based on this information, the most suitable storage space is allocated to the goods. By monitoring the use of storage space in real time, a dynamic adjustment plan is automatically generated.

[0099] The adjustment unit is used to perform real-time dynamic adjustment according to the dynamic adjustment scheme generated by the analysis unit.

[0100] The analysis unit sets the time interval to Δt to perform lidar scanning and update the cargo location status and cargo information, where t represents time. It randomly generates allocation plans based on the current cargo location status and cargo information, and regards each cargo location allocation plan as a particle. It finds the optimal cargo location allocation strategy through information sharing and collaboration between particles.

[0101] The volume of cargo i obtained by laser radar scanning is H i , the capacity of cargo space j is L j ; i represents the i-th cargo, j represents the j-th cargo location;

[0102] Record the cargo position status, using binary variable 0 j Indicates the occupancy status of storage location j, 0 j =1 means occupied, 0 means j =0 means free; calculate the frequency of in and out storage and storage demand assessment;

[0103] Calculate the frequency of inbound and outbound storage, and count the number of inbound and outbound times of item i in the time interval [t1, t2] R in and the number of shipments R out , the frequency of in and out storage is D i :

[0104]

[0105] Storage requirement calculation, considering the volume H of the cargo i and special storage requirement factor O i , set according to the classification of goods, storage demand S i equal:

[0106]

[0107] Encode the cargo location allocation plan as a particle. Suppose there are n cargo locations and m types of goods in the warehouse. Use an integer array of length m to represent a particle. Because each element corresponds to the cargo location number of a type of goods, the array length is equal to the number of types of goods. Each element represents the cargo location number assigned to the corresponding goods. A certain number of particles are randomly generated as the initial particle group. Use an array to represent the particles. Each element in the array is associated with a type of goods. If there are m types of goods, an array of length m is required to record the cargo location information of each type of goods. "Integer array of length m" means that there are m integer elements in this array, and each element can store an integer array, which is the cargo location number assigned to each type of goods.

[0108] Assume that the weight of the frequency of in-and-out storage is α, and the weight of the storage demand is β; the initial particle swarm corresponds to the fitness function E of the cargo allocation plan. The fitness function E is defined as the inverse of the sum of the comprehensive consideration values ​​of all goods. The higher the fitness value, the better the cargo allocation plan, that is:

[0109]

[0110] Each particle has a position vector and a velocity vector. Let x θi represents the cargo location number assigned to the i-th cargo in the θ-th particle, v θi Indicates the change speed of the cargo location number of the ith cargo in the θth particle; particle update, for each particle, v represents the speed, x represents the position, and its speed and position are updated according to the formula:

[0111] Speed ​​update formula:

[0112]

[0113] in, It is the speed of change of the cargo location number of the i-th cargo allocation of particle θ in the t+1th iteration. The speed update is to adjust the search step and direction of the particle to find more fitness functions. B is the inertia weight. f1 and f2 are learning factors used to control the step size of the algorithm search. The initial value is set randomly and adjusted according to the performance of the algorithm operation. For the inertia weight B, it is usually taken between [0,1]. If the algorithm calculation speed is fast, B is appropriately reduced to make the particle movement more dependent on the current optimal information and enhance the ability to explore new areas. If the algorithm search range is too slow, B is appropriately increased to increase the stability of particle movement and the inheritance of historical speed.

[0114] For the learning factor, if the algorithm is found to be too fast during the search process, the learning factor is appropriately lowered to reduce the search step size; if the algorithm is found to be too slow, the learning factor is appropriately increased to expand the search step size to find a better solution. These adjustments require multiple attempts and analyses during the actual algorithm operation.

[0115] r1 and r2 are random numbers in the interval [0, 1], which are used to prevent the algorithm from stagnating and falling into the local optimal solution. It is a strategy to increase the randomness of the algorithm. θi is the best position that particle θ has ever experienced, the cargo location number assigned to the i-th cargo in the individual optimal solution. These particles constitute a particle swarm, which is the basic unit for the algorithm to search and optimize. i is the best position experienced by the entire initial particle swarm, and the cargo location number assigned to the i-th cargo in the global optimal solution;

[0116] Position update formula:

[0117]

[0118] in is the cargo location number assigned to the jth cargo by particle θ in the t+1th iteration. The fitness function is calculated for each updated location. When the fitness function no longer changes, the algorithm stops.

[0119] When an idle position 0 is detected j =0, select particles with high fitness from the current particle swarm, analyze the goods with low frequency of entry and exit or small storage demand, adjust these goods to deep idle cargo spaces, and update the cargo space allocation plan represented by the particles.

[0120] For example, during a certain period of time, the frequency of goods leaving the warehouse is high, while the amount of household goods entering the warehouse increases. The lidar scan finds that the use of the cargo space has changed, with some cargo spaces becoming crowded and some being idle. The analysis unit calculates the frequency of entry and exit and storage requirements of different goods based on this information. For clothing goods, due to the high frequency of leaving the warehouse, it may be necessary to adjust them to a cargo space closer to the exit; for household goods, more suitable cargo spaces are allocated based on their volume and storage requirements.

[0121] The dynamic adjustment module dynamically adjusts the allocation of goods to the cargo space in real time according to the goods in and out of the warehouse and inventory changes, thereby improving space utilization. Compared with the existing technology where the cargo space storage planning is relatively manual and lacks a dynamic adjustment mechanism, this solution can better adapt to changes in the actual use of the warehouse and avoid the problem of long-term idleness or overcrowding of cargo spaces.

[0122] By finding the optimal cargo space allocation strategy, we can quickly and effectively allocate the most suitable cargo space for different goods and improve the efficiency of warehouse management.

[0123] The system management module includes a setting unit and a display unit; the setting unit sets the authority levels of different users and provides system setting functions, allowing managers to adjust system parameters according to actual needs.

[0124] The display unit provides a visual interface that displays data from the database unit and the real-time storage database unit, allowing managers to view warehouse status, cargo information, and cargo space usage.

[0125] Compared with the existing technology, the system management module provides an intuitive user interface, allowing managers to easily view warehouse status, cargo information and cargo space usage, making it easier to make decisions and manage. The setting unit can set the relevant parameters calculated by the system according to actual conditions, improving the standardization and applicability of the system.

[0126] The embodiments of the present invention are provided for the purpose of illustration and description. Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations of the present invention. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A cargo storage management system for identifying cargo locations based on laser radar, characterized in that: include: A laser radar module, which is used to scan and collect cargo locations and cargo in the warehouse to obtain data information about the cargo locations; A data analysis module, which is data-connected to the laser radar module and is used to analyze and process the cargo location data information acquired by the laser radar module and identify and classify the cargo; A dynamic adjustment module, which is data-connected to the data analysis module and the laser radar module, is used to dynamically allocate goods according to the cargo location data information obtained by the laser radar module and the results of identification and classification by the data analysis module, monitor the in-and-out of goods and inventory changes, and dynamically adjust the cargo location; The system management module is data-connected with the dynamic adjustment module and is used to configure and manage the system, including adjusting system parameters and providing an intuitive user interface.

2. A laser radar-based cargo storage management system for cargo location identification as claimed in claim 1, characterized in that: The laser radar module includes a laser radar device and an image acquisition device, which are used to obtain three-dimensional point cloud information and image information of the cargo location respectively; By setting appropriate parameters and frequency, the lidar equipment can automatically adjust the scanning speed according to the frequency of goods entering and leaving the warehouse. When the frequency of entering and leaving the warehouse increases, the scanning rate becomes faster. At the same time, the image acquisition device can realize synchronous data acquisition.

3. The cargo storage management system based on laser radar for identifying cargo locations as claimed in claim 1, characterized in that: The data analysis module includes an analysis and identification unit, a database unit and a classification unit; The analysis and identification unit is used to analyze the information obtained by the laser radar module, perform weighted fusion based on the features of the three-dimensional point cloud and the image features, and compare and identify the historical cargo feature data; The database unit is used to collect information acquired by the historical laser radar module, establish a database of feature classification, automatically collect historical data when there is no historical data, and automatically enrich and improve the cargo feature data; The classification unit is used to classify the goods scanned in real time by the laser radar module based on the comparison between the analysis and recognition unit results and the database unit feature classification.

4. A cargo storage management system for identifying cargo locations based on laser radar as claimed in claim 3, characterized in that: The analysis and recognition unit realizes the recognition and classification of goods based on the fusion of the main curvature features and deep learning features of the goods; According to the information of the goods scanned by the laser radar module, the point cloud data set of the goods is obtained: P = {P1, P2, ..., P w }, for the set of these point cloud data P = {P1, P2, ..., P w }, w is the total number of these points, let a be the index of the point, P a It means the ath point, and the principal curvature is used to describe the curvature of the point cloud data surface; First, calculate the local neighborhood of the point cloud. For each point P a , determine the specified number of neighboring points around it to form a local neighborhood N(P a ); Then, calculate the covariance matrix C of the local neighborhood: Where * represents the transpose of the matrix, k is the number of neighborhood points, is the average coordinate of the neighborhood points, which represents an important intermediate quantity in the covariance matrix of the calculated point cloud and is used to describe the overall position information of the neighborhood points. The calculation formula is: Perform eigenvalue decomposition on the covariance matrix C. The covariance matrix C is a μ×μ matrix, where μ is the dimension of the matrix, indicating that there are μ diagonal elements in the matrix, which means the number of coordinate components of point p. The eigenvalue decomposition formula of the covariance matrix is: C=QΛQ -1 Among them, the orthogonal matrix Q refers to a square matrix whose transpose is equal to its inverse matrix, and Λ is a diagonal matrix composed of eigenvalues. In the eigenvalue decomposition, this formula holds true, where the diagonal elements (λ1, λ2, ..., λμ) of Λ are the eigenvalues ​​of the matrix C; The eigenvalue decomposition is arranged from large to small to obtain the eigenvalues ​​λ1, λ2, ..., and the principal curvatures k1 and k2 are calculated by the following formula: k1 = λ1, k2 = λ2; In the analysis of goods with incomplete regular shapes, the principal curvature is equal to the eigenvalue because the eigenvalue reflects the degree of change of the local surface of the point cloud in different directions and is closely related to the curvature of the surface. The two principal curvatures are sufficient to describe the curvature of the surface at that point. Extract features from the cargo image, assuming that the image feature vector extracted by the deep neural network is γ; The curvature features and deep learning feature vectors are fused using a weighted fusion method. The fused feature vector is (ω1k,ω2γ), where k = {k1, k2} is the principal curvature of the point cloud, and ω1 and ω2 are weights, which are set according to the importance of the curvature features and image features.

5. The laser radar-based cargo storage management system for cargo location identification as claimed in claim 1, characterized in that: The dynamic adjustment module includes a real-time storage database unit, an analysis unit and an adjustment unit; The real-time warehouse database unit establishes a database based on the cargo data information acquired by the laser radar module through real-time scanning, including the real-time cargo occupancy and cargo quantity of the warehouse; The analysis unit calculates and determines the in-and-out frequency and storage requirements of different goods according to the real-time data transformation in the real-time warehouse database unit, and allocates the most suitable cargo space for the goods according to this information, and automatically generates a dynamic adjustment plan by monitoring the use of cargo space in real time; The adjustment unit is used to perform real-time dynamic adjustment according to the dynamic adjustment scheme generated by the analysis unit.

6. A cargo storage management system for identifying cargo locations based on laser radar as claimed in claim 5, characterized in that: The analysis unit sets the time interval to Δt to perform laser radar scanning, update the cargo location status and cargo information, and t represents time; randomly generates allocation plans based on the current cargo location status and cargo information, regards each cargo location allocation plan as a particle, and finds the optimal cargo location allocation strategy through information sharing and collaboration between particles; The volume of cargo i obtained by laser radar scanning is H i , the capacity of cargo space j is L j ; i represents the i-th cargo, j represents the j-th cargo location; Record the cargo position status, using binary variable 0 j Indicates the occupancy status of storage location j, 0 j =1 means occupied, 0 means j =0 means free; calculate the frequency of in and out storage and storage demand assessment; Calculate the frequency of inbound and outbound storage, and count the number of inbound and outbound times of item i in the time interval [t1, t2] R in and the number of shipments R out , the frequency of in and out storage is D i : Storage requirement calculation, considering the volume H of the cargo i and special storage requirement factor O i , set according to the classification of goods, storage demand S i equal: Encode the cargo location allocation plan as a particle. Suppose there are n cargo locations and m types of goods in the warehouse. Use an integer array of length m to represent a particle. Because each element corresponds to the cargo location number of a type of goods, the array length is equal to the number of types of goods. Each element represents the cargo location number assigned to the corresponding goods. A certain number of particles are randomly generated as the initial particle group. Use an array to represent the particles. Each element in the array is associated with a type of goods. If there are m types of goods, an array of length m is required to record the cargo location information of each type of goods. Assume that the weight of the frequency of in-and-out storage is α, and the weight of the storage demand is β; the initial particle swarm corresponds to the fitness function E of the cargo allocation plan. The fitness function E is defined as the inverse of the sum of the comprehensive consideration values ​​of all goods. The higher the fitness value, the better the cargo allocation plan, that is: Each particle has a position vector and a velocity vector. Let x θi represents the cargo location number assigned to the i-th cargo in the θ-th particle, v θi Indicates the change speed of the cargo location number of the ith cargo in the θth particle; Particle update, for each particle, v represents the speed, x represents the position, and its speed and position are updated according to the formula: Speed ​​update formula: in, is the speed of change of the cargo location number of the i-th cargo allocation of particle θ in the t+1th iteration. The speed update is to adjust the search step and direction of the particle to find more fitness functions. B is the inertia weight. f1 and f2 are learning factors used to control the step size of the algorithm search. The initial value is randomly set and adjusted according to the performance of the algorithm operation. For the inertia weight B, it is taken between [0,1]. The algorithm calculation speed is fast. Appropriately reduce B to make the movement of the particle more dependent on the current optimal information. For the learning factor, when the algorithm finds that it is too fast during the search process, appropriately reduce the learning factor to reduce the search step size; r1 and r2 are random numbers in the interval [0, 1], which are used to prevent the algorithm from stagnating and falling into the local optimal solution. It is a strategy to increase the randomness of the algorithm. θi is the best position that particle θ has ever experienced, the cargo location number assigned to the i-th cargo in the individual optimal solution. These particles constitute a particle swarm, which is the basic unit for the algorithm to search and optimize. i is the best position experienced by the entire initial particle swarm, and the cargo location number assigned to the i-th cargo in the global optimal solution; Position update formula: in is the cargo location number assigned to the jth cargo by particle θ in the t+1th iteration. The fitness function is calculated for each updated location. When the fitness function no longer changes, the algorithm stops. When an idle position 0 is detected j =0, select particles with high fitness from the current particle swarm, analyze the goods with low frequency of entry and exit or small storage demand, adjust these goods to deep idle cargo spaces, and update the cargo space allocation plan represented by the particles.

7. The laser radar-based cargo storage management system for cargo location identification as claimed in claim 1, characterized in that: The system management module includes a setting unit and a display unit; The setting unit sets the authority levels of different users and provides a system setting function, allowing managers to adjust system parameters according to actual needs.

8. A cargo storage management system for identifying cargo locations based on laser radar as claimed in claim 7, characterized in that: The display unit provides a visual interface, which displays data in the database unit and the real-time storage database unit, allowing managers to view warehouse status, cargo information, and cargo space usage.

Images