Cargo classification prediction and recognition method based on artificial intelligence

Abstract

The application provides an artificial intelligence-based cargo classification prediction and recognition method, comprising the following steps: S1, data collection and labeling: collecting images and label information of warehouse goods, and labeling the category of each piece of goods; S2, data preprocessing: preprocessing the collected images, and cleaning and arranging the label information accordingly; S3, feature extraction: using a convolutional neural network in deep learning to extract features from the images, and using NLP technology to extract relevant features from the label information; S4, classification model construction: establishing a YOLO deep learning classification model, inputting the extracted features into the model for training, and predicting the category of the goods from the features; S5, verifying the generalization ability of the YOLO deep learning classification model on unknown data by using a genetic algorithm, the application can automatically identify and classify various goods, greatly improving work efficiency and accuracy, reducing the manual classification error rate, and promoting the intelligentization of related industries.

Description

Technical Field

[0001] This invention relates to the field of cargo classification technology, and in particular to a cargo classification prediction and identification method based on artificial intelligence. Background Technology

[0002] With societal progress, the volume and speed of goods circulation are constantly increasing. As a crucial part of logistics management, warehousing management faces new challenges in ensuring smooth logistics. Warehouses can be classified according to different standards, and companies can choose to build or lease different types of warehouses to store various goods based on their own conditions. A large logistics warehouse often has multiple storage areas, categorized and stored according to the nature of the goods. To save space, some goods may need to be stacked, which places demands on the classification and packaging of goods.

[0003] Currently, most logistics warehousing adopts a standard method of pre-booking goods storage and managing goods by category. This method often requires advance inventory of storage locations and on-site guidance from warehouse management personnel to store goods. Moreover, the time of goods entering the warehouse is not fixed, making it impossible to manage the goods in real time. This results in low efficiency of goods classification management, making it impossible to quickly classify the goods entering the warehouse and reducing warehouse management efficiency.

[0004] Therefore, it is necessary to provide a new AI-based method for cargo classification prediction and identification to solve the above-mentioned technical problems. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides an artificial intelligence-based method for predicting and identifying cargo classification.

[0006] The artificial intelligence-based cargo classification prediction and identification method provided by the present invention includes the following steps: S1, data collection and labeling: collecting images and label information of warehouse goods, and labeling each item of goods with a category;

[0007] S2. Data preprocessing: Preprocess the acquired images and clean and organize the label information accordingly.

[0008] S3. Feature Extraction: Features are extracted from images using convolutional neural networks in deep learning. For label information, NLP technology is used to extract relevant features.

[0009] S4. Classification Model Construction: Establish a YOLO deep learning classification model, input the extracted features into the model for training, and predict the category of goods from the features;

[0010] S5. Validate the generalization ability of the YOLO deep learning classification model on unknown data using a genetic algorithm, specifically including the following:

[0011] S501. Encode the variables according to the model requirements;

[0012] S502. According to the chromosome coding rules, a batch of chromosomes is randomly generated as the initial population;

[0013] S503. Construct a fitness function based on the YOLO deep learning classification model to evaluate the superiority or inferiority of chromosomes;

[0014] S504. Selecting chromosomes with good fitness through selection operations allows for the inheritance of chromosomes; exchanging information through crossover operations allows for the generation of new chromosomes through mutation operations.

[0015] S505. After selection, crossover, and mutation operations, chromosomes form a new population. Fitness assessment is repeatedly performed to promote continuous evolution of the population. This cycle continues until the optimal solution is found, at which point evolution stops.

[0016] Furthermore, the NLP technology for extracting label information features includes the following steps:

[0017] 1) Label the preprocessed label information with semantic and contextual labels, and use the vector space model to extract features from the label data;

[0018] 2) Transform the textual feature data into vectors that can be recognized by the classification model, and use these vectors as the basis for classification.

[0019] 3) Input the vectors into the classification model for semantic and contextual classification.

[0020] Furthermore, the preprocessing of the image includes a mixed processing method of color correction, noise reduction, cropping, and scaling.

[0021] Furthermore, the images and tag information of the stored goods include pictures of the goods, RFID tags, size information of the goods, and usage information.

[0022] Another aspect of the present invention provides an artificial intelligence-based cargo classification and identification method, which is based on the aforementioned artificial intelligence-based cargo classification prediction method, and further includes the following steps:

[0023] S1. Based on the smart cameras installed on the warehouse shelves or on the sorting line, take pictures of the goods in real time and read the data, and calculate the category of the goods through the trained classification model.

[0024] S2. Obtain the set of transportation routes D = {(D pqLet K be the coordinate set of the goods sorting racks, K = {K0, K1, K2, ..., Kn}. n}, K0 is the coordinate point of the cargo, K n Let D be the coordinates of the nth goods sorting rack. pq For sorting shelves K p To K q The distance;

[0025] S3. Each AGV transport vehicle is equipped with an RFID reader and a lidar. The RFID reader identifies the category label pre-attached to the goods to determine the goods information, and the lidar identifies landmark signals to avoid obstacles.

[0026] S4. Based on n cargo sorting racks, transport the goods in batches, and assign the value p of each item to each cargo rack. j and cargo weight w j Defined as non-negative, based on the maximum load capacity W of the AGV transport trolley, satisfying... Select combinations of high-value goods that meet the maximum load capacity of the AGV transport vehicles to guide the AGV transport vehicles on the transportation route to classify and sort the goods in batches. Prioritize loading goods of the same type onto the same AGV transport vehicle. If all goods of the same type can be loaded, it is a reasonable and effective batch transportation and route selection.

[0027] Furthermore, the structure of the AGV transport vehicle includes a body with omnidirectional wheels, a mechanical clamp, and a shelf for placing goods. The mechanical clamp is located at the back of the body, the shelf is located at the top of the body, and the RFID reader and lidar are located directly opposite the body.

[0028] The mechanical clamp includes two opposing clamps and a drive unit for driving the two clamps to move toward or away from each other. The two clamps are multi-segment telescopic.

[0029] Compared with related technologies, the artificial intelligence-based cargo classification prediction and identification method provided by this invention has the following beneficial effects:

[0030] 1. This invention uses n cargo sorting racks to transport goods in batches. It selects combinations of high-value goods that meet the maximum load capacity of the AGV transport vehicle and guides the AGV transport vehicle on the transport route to sort the goods in batches. Priority is given to loading goods of the same type onto the same AGV transport vehicle. If all goods of the same type can be loaded, it is a reasonable and effective batching and route plan, which can be reasonably allocated according to transportation costs and vehicle utilization.

[0031] 2. This invention can automatically identify and classify various goods, greatly improving work efficiency and accuracy, reducing the error rate of manual classification, and promoting the intelligentization of related industries.

[0032] 3. In this invention, the mechanical clamps of the AGV transport vehicle are located at the back of the vehicle body to avoid the situation where the cargo obstructs the view and affects the accuracy of recognition during transportation. In addition, the two clamps are designed with multi-segment telescopic extension and retraction for protection when not in transportation. Attached Figure Description

[0033] Figure 1 A flowchart illustrating Embodiment 1 of the present invention;

[0034] Figure 2 A flowchart of Embodiment 2 provided by the present invention;

[0035] Figure 3 A schematic diagram of the structure of the AGV transport vehicle provided by the present invention. Figure 1 ;

[0036] Figure 4 A schematic diagram of the structure of the AGV transport vehicle provided by the present invention. Figure 2 .

[0037] The following are labeled in the diagram: 1. Vehicle body; 2. Storage rack; 3. Clamping plate; 4. Drive unit; 5. RFID reader; 6. LiDAR. Detailed Implementation

[0038] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0039] Please refer to the following: Figure 1 , Figure 2 , Figure 3 as well as Figure 4 ,in, Figure 1 A flowchart illustrating Embodiment 1 of the present invention; Figure 2 The flowchart is for Embodiment 2 of the present invention. Figure 3 A schematic diagram of the structure of the AGV transport vehicle provided by the present invention. Figure 1 ; Figure 4 A schematic diagram of the structure of the AGV transport vehicle provided by the present invention. Figure 2 .

[0040] Example 1

[0041] In the specific implementation process, such as Figure 1 As shown, the artificial intelligence-based cargo classification prediction and identification method provided by this invention includes the following steps:

[0042] S1. Data Collection and Labeling: Collect images and labeling information of stored goods, and label each item with a category. The images and labeling information of stored goods include pictures of the goods, RFID tags, size information and usage information of the goods.

[0043] S2. Data preprocessing: Preprocess the acquired images, including color correction, noise reduction, cropping, and scaling, and cleaning and organizing the label information accordingly.

[0044] S3. Feature Extraction: Features are extracted from images using convolutional neural networks in deep learning. For label information, NLP technology is used to extract relevant features.

[0045] The NLP technique for extracting label information features includes the following steps:

[0046] 1) Label the preprocessed label information with semantic and contextual labels, and use the vector space model to extract features from the label data;

[0047] 2) Transform the textual feature data into vectors that can be recognized by the classification model, and use these vectors as the basis for classification.

[0048] 3) Input the vectors into the classification model for semantic and contextual classification;

[0049] Semantic classification: Input is a vector transformed from feature data in text form. The output of each hidden layer neuron is calculated to obtain a preliminary prediction result. Then, the error is calculated by minimizing the root mean square error of the preliminary prediction result. The error bias is adjusted and the weights are updated. For each label information, if the error is less than the set threshold or the number of iterations has been reached, the training ends.

[0050] S4. Classification Model Construction: Build a YOLO deep learning classification model, input the extracted features into the model for training, and predict the category of goods from the features;

[0051] S5. Validate the generalization ability of the YOLO deep learning classification model on unknown data using a genetic algorithm, specifically including the following:

[0052] S501. Encode the variables according to the model requirements;

[0053] S502. According to the chromosome coding rules, a batch of chromosomes is randomly generated as the initial population;

[0054] S503. Construct a fitness function based on the YOLO deep learning classification model to evaluate the superiority or inferiority of chromosomes;

[0055] S504. Selecting chromosomes with good fitness through selection operations allows for the inheritance of chromosomes; exchanging information through crossover operations allows for the generation of new chromosomes through mutation operations.

[0056] S505. After selection, crossover, and mutation operations, chromosomes form a new population. Fitness evaluation is repeatedly performed to promote the continuous evolution of the population. This cycle continues until the optimal solution is found and then evolution stops.

[0057] By extracting label information features using NLP technology, adjusting error bias, and updating weights, the recognition accuracy of label information can be further improved, providing favorable conditions for classification models to identify the type of goods.

[0058] This invention uses a deep learning classification model to automatically identify and classify various goods, greatly improving work efficiency and accuracy, reducing the error rate of manual classification, and promoting the intelligentization of related industries. It can be used in areas such as warehouse automation management, logistics sorting systems, and e-commerce platform product classification.

[0059] Example 2

[0060] In the specific implementation process, the artificial intelligence-based cargo classification and identification method is based on the artificial intelligence-based cargo classification prediction method in Example 1, and refers to... Figure 2 As shown, it also includes the following steps:

[0061] S1. Based on the smart cameras installed on the warehouse shelves or on the sorting line, take pictures of the goods in real time and read the data, and calculate the category of the goods through the trained classification model.

[0062] S2. Obtain the set of transportation routes D = {(D pq Let K be the coordinate set of the goods sorting racks, K = {K0, K1, K2, ..., Kn}. n}, K0 is the coordinate point of the cargo, K n Let D be the coordinates of the nth goods sorting rack. pq For sorting shelves K p To K q The distance;

[0063] S3. Each AGV transport vehicle is equipped with an RFID reader 5 and a lidar 6. The RFID reader 5 identifies the category label pre-attached to the goods to determine the goods information, and the lidar 6 identifies landmark signals to avoid obstacles.

[0064] S4. Based on n cargo sorting racks, transport the goods in batches, and assign the value p of each item to each cargo rack. j and cargo weight w jDefined as non-negative, based on the maximum load capacity W of the AGV transport trolley, satisfying... Select high-value combinations of goods that meet the maximum load capacity of the AGV transport vehicles, and guide the AGV transport vehicles on the transportation route to classify and sort the goods in batches. Prioritize loading the same type of goods onto the same AGV transport vehicle. If all goods of the same type can be loaded, it is a reasonable and effective batch transportation and route selection. The allocation can be made reasonably based on transportation costs and vehicle utilization.

[0065] Specifically, there are AGV transport vehicles between two adjacent cargo sorting racks. The computer control software plans the route and issues instructions to the AGV transport vehicles to travel along the predetermined route. If there is a situation where an AGV transport vehicle is not fully loaded with the same type of goods on the transport route, it can be dispatched according to the AGV transport vehicle closest to the target coordinate point of the cargo sorting rack. It should be noted that this dispatch is preferably performed on AGV transport vehicles that are empty.

[0066] Example 3

[0067] In the specific implementation process, such as Figure 3 As shown, following the same structure as detailed in Embodiment 2, the AGV transport vehicle structure includes a vehicle body 1 with omnidirectional wheels, a mechanical clamp, and a shelf 2 for placing goods. The mechanical clamp is located at the back of the vehicle body 1 to avoid the goods obstructing the view during transportation. The shelf 2 is located at the top of the vehicle body 1, and the RFID reader 5 and the lidar 6 are located directly opposite the vehicle body 1.

[0068] The mechanical clamp includes two opposing clamping plates 3 and a driving component 4 for driving the two clamping plates 3 to move towards or away from each other. It should be noted that the driving component 4 is a motor, with a gear connected to the output end of the motor and racks installed on the back of the two clamping plates 3, so that the gear drives the two racks to move towards or away from each other. In addition, the two clamping plates 3 are multi-segment telescopic design, which is retracted for protection during non-transportation processes.

[0069] Specifically, the AGV transport vehicle has two transport modes, which can be selected according to the type of goods:

[0070] 1. For ordinary goods, use two clamps 3 to hold the goods and then flip them over to place them into the shelf 2;

[0071] 2. For fragile goods, they should be transported by clamping them with two clamps 3, and should not be flipped over to avoid tipping over.

[0072] The vehicle body 1 is also equipped with a receiving module for receiving computer control commands and a backup wireless control module, which can exchange data with the central control system via any of the following methods: Wi-Fi, Bluetooth, or ZigBee, to receive task commands and send status information.

[0073] The circuits and controls involved in this invention are all existing technologies and will not be described in detail here.

[0074] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0075] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or basic characteristics. Therefore, the embodiments should be considered exemplary and non-limiting in all respects. The scope of the invention is defined by the appended claims rather than the foregoing description. Therefore, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention, and no reference numerals in the claims should be construed as limiting the scope of the claims.

[0076] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A cargo classification and identification method based on artificial intelligence, characterized in that, Includes the following steps: S1. Data Collection and Labeling: Collect images and label information of stored goods, and label each item with its category. S2. Data preprocessing: Preprocess the acquired images and clean and organize the label information accordingly. S3. Feature Extraction: Features are extracted from images using convolutional neural networks in deep learning. For label information, NLP technology is used to extract relevant features. S4. Classification Model Construction: Build a YOLO deep learning classification model, input the extracted features into the model for training, and predict the category of goods from the features; S5. Validate the generalization ability of the YOLO deep learning classification model on unknown data using a genetic algorithm, specifically including the following: S501. Encode the variables according to the model requirements; S502. According to the chromosome coding rules, a batch of chromosomes is randomly generated as the initial population; S503. Construct a fitness function based on the YOLO deep learning classification model to evaluate the superiority or inferiority of chromosomes; S504. Selecting chromosomes with good fitness through selection operations allows for the inheritance of chromosomes; exchanging information through crossover operations allows for the generation of new chromosomes through mutation operations. S505. After selection, crossover, and mutation operations, chromosomes form a new population. Fitness evaluation is repeatedly performed to promote the continuous evolution of the population. This cycle continues until the optimal solution is found and then evolution stops. It also includes the following steps: S6. Based on the smart cameras installed on the warehouse shelves or the smart cameras on the sorting line, take pictures of the goods in real time and read the data, and calculate the category of the goods through the trained classification model. S7. Obtain the set of transportation routes Where K is the coordinate set of the cargo sorting rack. , For the coordinates of the goods, Let n be the coordinates of the nth goods sorting rack. For sorting goods racks arrive The distance; S8. Each AGV transport vehicle is equipped with an RFID reader (5) and a lidar (6). The RFID reader (5) identifies the category label pre-attached to the goods to determine the goods information, and the lidar (6) identifies the landmark signal to avoid obstacles. S9. Based on n cargo sorting racks, transport the goods in batches, and assign the value of each item to... and cargo weight Defined as non-negative, based on the maximum load capacity W of the AGV transport trolley, satisfying... The system selects high-value combinations of goods that meet the maximum load capacity of the AGV transport vehicles, guiding the AGV transport vehicles along the transport route to classify and sort the goods in batches. Priority is given to loading the same type of goods onto the same AGV transport vehicle. If all goods of the same type can be loaded, it is a reasonable and effective batch transport and route selection.

2. The cargo classification and identification method based on artificial intelligence according to claim 1, characterized in that, The NLP technology for extracting label information features includes the following steps: 1) Label the preprocessed label information with semantic and contextual labels, and use the vector space model to extract features from the label data; 2) Transform the textual feature data into vectors that can be recognized by the classification model, and use these vectors as the basis for classification. 3) Input the vectors into the classification model for semantic and contextual classification.

3. The artificial intelligence-based cargo classification and identification method according to claim 2, characterized in that, The image preprocessing includes a combination of color correction, noise reduction, cropping, and scaling.

4. The artificial intelligence-based cargo classification and identification method according to claim 3, characterized in that, The images and tag information of the stored goods include pictures of the goods, RFID tags, size information and usage information of the goods.

5. The artificial intelligence-based cargo classification and identification method according to claim 1, characterized in that, If an AGV transport vehicle is not fully loaded with the same type of goods on the transport route, it can be scheduled according to the AGV transport vehicle closest to the target coordinate point of the goods classification rack. The preferred scheduling is to use an AGV transport vehicle that is empty.

6. The artificial intelligence-based cargo classification and identification method according to claim 5, characterized in that, The structure of the AGV transport vehicle includes a vehicle body (1) with omnidirectional wheels, a mechanical clamp, and a shelf (2) for placing goods. The mechanical clamp is located at the back of the vehicle body (1), the shelf (2) is located at the top of the vehicle body (1), and the RFID reader (5) and the lidar (6) are located directly opposite the vehicle body (1). The mechanical clamp includes two opposing clamps (3) and a drive (4) for driving the two clamps (3) to move toward or away from each other. The two clamps (3) are multi-segment telescopic.

Images