CN118505114A - Warehouse goods warehousing system and method based on radio frequency identification and image processing - Google Patents

Abstract

The invention discloses a warehouse-in system and a method for warehouse goods based on radio frequency identification and image processing, wherein the system comprises: an identification area; the periphery of the identification area is provided with a tag identifier for identifying the radio frequency tag arranged on the warehouse-in goods and a state identifier for identifying the real-time state of the warehouse-in goods; the server is respectively connected with the tag identifier and the state identifier in a signal way and is internally provided with a preprocessing module; the preprocessing module performs error elimination on the data transmitted by the tag identifier and the state identifier by using a method of minimizing the sum of squares of the deviation; the operation end is in signal connection with the server and internally provided with a man-machine interaction interface; the server is also internally provided with a warehouse management strategy library, and the server matches the storage strategy and the scheduling strategy of the warehouse goods in the warehouse management strategy library according to the label identification result and the state identification result of the warehouse goods. The invention effectively overcomes the limitation of a single technology in a complex environment by combining the radio frequency identification technology and the image processing technology.

Description

Warehouse goods warehousing system and method based on radio frequency identification and image processing

Technical Field

The invention relates to the technical field of logistics management, in particular to a warehouse-in system and method for warehouse goods based on radio frequency identification and image processing.

Background

With the rapid development of the logistics industry, the efficiency and accuracy of warehouse management become key to improving the logistics service quality. Conventional warehouse entry systems typically rely on manual entry and bar code scanning, which are time consuming and prone to error, particularly when handling large amounts of goods. Therefore, improving the speed and accuracy of cargo identification has become an important issue in improving warehouse management efficiency.

In recent years, developments in Radio Frequency Identification (RFID) technology and image processing technology have provided new solutions for warehouse cargo management. The radio frequency identification technology identifies and tracks the articles with radio frequency tags through radio waves, so that the warehousing efficiency and accuracy are improved. However, despite the advantages of RFID technology, its performance may be limited in complex environments, such as multi-tag interference, tag damage or loss, and the like.

Furthermore, image processing techniques, particularly with the aid of deep learning and advances in neural networks, have been widely used for object recognition and classification. The details and the characteristics of the goods can be accurately identified through the high-definition camera and the image identification algorithm. However, when image processing techniques are used alone, challenges such as illumination variation, occlusion, and viewing angle limitations are also faced.

Although each of the above techniques has advantages, they alone tend to be difficult to address in all situations, especially in large-scale and fast-paced modern warehouse environments. Therefore, how to effectively combine the RFID technology with the image processing technology, optimize the identification process and reduce the error is a problem that needs to be solved in the current technology.

Disclosure of Invention

In order to solve the technical problems, the invention aims to provide a more efficient and accurate warehouse entry system of warehouse goods by combining a radio frequency identification technology and an image processing technology and introducing an advanced image identification neural network and an optimization algorithm.

In order to achieve the above object, the present invention provides a technical solution comprising:

Warehouse entry system of goods based on radio frequency identification and image processing includes:

The identification area is arranged on the ground at the entrance of the warehouse, matches the size of the warehouse-in vehicle and is provided with markers around;

the periphery of the identification area is provided with a tag identifier for identifying the radio frequency tag arranged on the warehouse-in goods;

An image acquisition device for acquiring images of the warehoused goods is arranged above the identification area;

the server is respectively connected with the tag identifier and the image acquisition device in a signal way and internally provided with an image recognition neural network; the image recognition neural network performs fine adjustment by using a method for minimizing the sum of squares of deviation so as to eliminate errors of image recognition results;

The operation end is in signal connection with the server and internally provided with a man-machine interaction interface;

And the server matches and stores the radio frequency tag identification result and the image identification result of the goods and sends the radio frequency tag identification result and the image identification result to the operation end for display to a user.

In some preferred embodiments, the preprocessing module performs error cancellation by minimizing the sum of squares of deviations, including:

constructing a loss function J (theta) for measuring the difference between the cargo identification result and the cargo actual information, wherein theta is an image identification neural network parameter;

Minimizing the loss function J (θ) by gradient descent, the parameter θ being in each iteration An update, where a is the learning rate,A bias to parameter θ _j for loss function J (θ);

adjusting the image recognition neural network by using the updated theta;

constructing and calculating an optimized objective function of the difference between the goods identification result and the goods actual information:

Repeating the steps until the optimization objective function is not changed, and finishing the fine adjustment of the image recognition neural network.

In some preferred embodiments, the optimization objective function of the difference between the cargo identification result and the cargo actual information is:

argmin＝SST+W₁×SSE+W₂×SSA；

Wherein SST is a method for measuring all cargo identification result predicted values y _ij and cargo identification result ensemble average predicted values The sum of the squares of the deviations between them,SSE is the sum of squares measuring the difference between each cargo identification result predicted value y _ij and the cargo identification result average predicted value y _i of the group to which it belongs, SSE = Σ (y _ij―y_i)²; SSA is the sum of cargo identification result average predicted value y _i and cargo identification result ensemble average predicted value of different groups)The sum of the squares of the differences between them, W ₁ and W ₂ are weight parameters that measure the extent to which intra-and inter-group bias contribute to the total bias, respectively.

In some preferred embodiments, before the server matches the storage policy and the scheduling policy of the warehouse-in goods in the warehouse management policy base according to the tag identification result and the status identification result of the warehouse-in goods, the method further comprises the steps of:

Matching the radio frequency tag identification result and the real-time image identification result of the warehouse-in goods, storing if the radio frequency tag identification number and the real-time image identification number of the warehouse-in goods are consistent, otherwise, giving an alarm; if the radio frequency tag identification category of the warehouse-in goods is consistent with the real-time image identification category, storing; otherwise, alarming.

In some preferred embodiments, the W ₁ and W ₂ are initialized by means of random assignments.

In some preferred embodiments, the method for matching the radio frequency tag identification result and the image identification result of the goods by the server includes:

matching whether the identification number of the radio frequency tags of the goods is consistent with the identification number of the images, if so, storing; if not, alarming;

matching whether the radio frequency tag identification category of the goods is consistent with the image identification category, if so, storing; if not, alarming.

The invention also provides a warehouse-in method of the warehouse goods based on the radio frequency identification and the image processing, which comprises the following steps:

S1, driving and parking a transport tool carrying warehouse-in goods into an identification area which is arranged on the ground at a warehouse entrance, matches the size of a warehouse-in vehicle and is provided with markers around the warehouse-in vehicle;

S2, controlling a tag identifier arranged on the periphery of the identification area to identify the radio frequency tag on the warehouse-in goods; controlling an image acquisition device arranged above the identification area to acquire images of warehoused goods;

S3, the server acquires a cargo image from the image acquisition device, and the cargo information is identified by using a built-in image identification neural network to acquire an image identification result, wherein the image identification neural network is subjected to fine adjustment by using a method of minimizing the square sum of deviation so as to eliminate errors of the image identification result;

S4, the server acquires a label identification result from the label identifier;

and S5, matching and storing the radio frequency tag identification result and the image identification result of the goods by the server, and sending the radio frequency tag identification result and the image identification result to the operation end for display to a user.

In some preferred embodiments, the method for error cancellation by the preprocessing module in step S3 using a method for minimizing the sum of squares of deviations includes:

S301, constructing a loss function J (theta) for measuring the difference between a cargo identification result and cargo actual information, wherein theta is an image identification neural network parameter;

s302, minimizing a loss function J (theta) by adopting a gradient descent method, wherein in each iteration, the parameter theta is as follows An update, where a is the learning rate,A bias to parameter θ _j for loss function J (θ);

s303, adjusting the image recognition neural network by using the updated theta;

s304, constructing and calculating an optimization objective function of the difference between the goods identification result and the goods actual information:

S305, repeating the steps until the optimization objective function is not changed, and finishing fine adjustment of the image recognition neural network.

In some preferred embodiments, the optimization objective function of the difference between the cargo identification result and the cargo actual information is:

argmin＝SST+W₁×SSE+W₂×SSA；

Wherein SST is a method for measuring all cargo identification result predicted values y _ij and cargo identification result ensemble average predicted values The sum of the squares of the deviations between them,SSE is the sum of squares measuring the difference between each cargo identification result predicted value y _ij and the cargo identification result average predicted value y _i of the group to which it belongs, SSE = Σ (y _ij―y_i)²; SSA is the sum of cargo identification result average predicted value y _i and cargo identification result ensemble average predicted value of different groups)The sum of the squares of the differences between them, W ₁ and W ₂ are weight parameters that measure the extent to which intra-and inter-group bias contribute to the total bias, respectively.

In some preferred embodiments, step S4 further includes the step of, before the server matches the storage policy and the scheduling policy of the warehouse-in goods in the warehouse management policy library according to the tag identification result and the status identification result of the warehouse-in goods after the errors are eliminated:

Matching the radio frequency tag identification result and the real-time image identification result of the warehouse-in goods, storing if the radio frequency tag identification number and the real-time image identification number of the warehouse-in goods are consistent, otherwise, giving an alarm; if the radio frequency tag identification category of the warehouse-in goods is consistent with the real-time image identification category, storing; otherwise, alarming.

In some preferred embodiments, the W ₁ and W ₂ are initialized by means of random assignments.

Advantageous effects

1. The identification accuracy is improved: the invention effectively overcomes the limitation of a single technology in a complex environment by combining Radio Frequency Identification (RFID) technology and image processing technology. Radio frequency identification technology provides fast object recognition capabilities, while image processing technology provides fine visual information. The dual recognition mechanism greatly improves the accuracy of cargo recognition, especially when handling large amounts or similar looking cargo.

2. Error and human intervention are reduced: by introducing the image recognition neural network and the method for minimizing the sum of squares of deviation for fine tuning, the system can intelligently recognize and eliminate errors of the image recognition result. The method not only reduces errors caused by human judgment errors or equipment errors, but also reduces the dependence on human intervention, thereby improving the automation and intelligent level of the warehousing process.

3. Efficiency is improved: the invention optimizes the whole process of goods warehouse entry, from identification to information processing to data storage, and each step is carefully designed to ensure quick and high efficiency. Through automatic identification and intelligent algorithm, more cargoes can be processed in a shorter time, and the efficiency of warehouse operation is greatly improved.

Drawings

FIG. 1 is a schematic diagram of a warehouse entry system for warehoused cargo based on RFID and image processing in a preferred embodiment of the present invention;

FIG. 2 is a flow chart of a warehouse entry method based on RFID and image processing in a preferred embodiment of the invention;

Detailed Description

The present invention will be further described with reference to the accompanying drawings, in order to make the objects, technical solutions and advantages of the present invention more apparent. In the description of the present invention, it should be understood that the terms "upper," "lower," "front," "rear," "left," "right," "top," "bottom," "inner," "outer," and the like indicate or are based on the orientation or positional relationship shown in the drawings, merely to facilitate description of the present invention and to simplify the description, and do not indicate or imply that the devices or elements referred to must have a specific orientation, be configured and operated in a specific orientation, and thus should not be construed as limiting the present invention.

Example 1

As shown in fig. 1, this embodiment provides a warehouse cargo warehousing system based on radio frequency identification and image processing, including:

The identification area is arranged on the ground at the entrance of the warehouse, matches the size of the warehouse-in vehicle and is provided with markers around;

the periphery of the identification area is provided with a tag identifier for identifying the radio frequency tag arranged on the warehouse-in goods;

An image acquisition device for acquiring images of the warehoused goods is arranged above the identification area;

the server is respectively connected with the tag identifier and the image acquisition device in a signal way and internally provided with an image recognition neural network; the image recognition neural network performs fine adjustment by using a method for minimizing the sum of squares of deviation so as to eliminate errors of image recognition results;

The operation end is in signal connection with the server and internally provided with a man-machine interaction interface;

And the server matches and stores the radio frequency tag identification result and the image identification result of the goods and sends the radio frequency tag identification result and the image identification result to the operation end for display to a user.

In some preferred embodiments, the preprocessing module performs error cancellation by minimizing the sum of squares of deviations, including:

constructing a loss function J (theta) for measuring the difference between the cargo identification result and the cargo actual information, wherein theta is an image identification neural network parameter;

Minimizing the loss function J (θ) by gradient descent, the parameter θ being in each iteration An update, where a is the learning rate,A bias to parameter θ _j for loss function J (θ);

adjusting the image recognition neural network by using the updated theta;

constructing and calculating an optimized objective function of the difference between the goods identification result and the goods actual information:

Repeating the steps until the optimization objective function is not changed, and finishing the fine adjustment of the image recognition neural network.

In some preferred embodiments, the optimization objective function of the difference between the cargo identification result and the cargo actual information is:

argmin＝SST+W₁×SSE+W₂×SSA；

Wherein SST is a method for measuring all cargo identification result predicted values y _ij and cargo identification result ensemble average predicted values The sum of the squares of the deviations between them,SSE is the sum of squares measuring the difference between each cargo identification result predicted value y _ij and the cargo identification result average predicted value y _i of the group to which it belongs, SSE = Σ (y _ij―y_i)²; SSA is the sum of cargo identification result average predicted value y _i and cargo identification result ensemble average predicted value of different groups)The sum of the squares of the differences between them, W ₁ and W ₂ are weight parameters that measure the extent to which intra-and inter-group bias contribute to the total bias, respectively.

In some preferred embodiments, before the server matches the storage policy and the scheduling policy of the warehouse-in goods in the warehouse management policy base according to the tag identification result and the status identification result of the warehouse-in goods, the method further comprises the steps of:

Matching the radio frequency tag identification result and the real-time image identification result of the warehouse-in goods, storing if the radio frequency tag identification number and the real-time image identification number of the warehouse-in goods are consistent, otherwise, giving an alarm; if the radio frequency tag identification category of the warehouse-in goods is consistent with the real-time image identification category, storing; otherwise, alarming.

In some preferred embodiments, the W ₁ and W ₂ are initialized by means of random assignments.

In some preferred embodiments, the method for matching the radio frequency tag identification result and the image identification result of the goods by the server includes:

matching whether the identification number of the radio frequency tags of the goods is consistent with the identification number of the images, if so, storing; if not, alarming;

matching whether the radio frequency tag identification category of the goods is consistent with the image identification category, if so, storing; if not, alarming.

Example 2

As shown in fig. 2, the invention further provides a warehouse-in method of warehouse goods based on radio frequency identification and image processing, which comprises the following steps:

S1, driving and parking a transport tool carrying warehouse-in goods into an identification area which is arranged on the ground at a warehouse entrance, matches the size of a warehouse-in vehicle and is provided with markers around the warehouse-in vehicle;

S2, controlling a tag identifier arranged on the periphery of the identification area to identify the radio frequency tag on the warehouse-in goods; controlling an image acquisition device arranged above the identification area to acquire images of warehoused goods;

S3, the server acquires a cargo image from the image acquisition device, and the cargo information is identified by using a built-in image identification neural network to acquire an image identification result, wherein the image identification neural network is subjected to fine adjustment by using a method of minimizing the square sum of deviation so as to eliminate errors of the image identification result;

S4, the server acquires a label identification result from the label identifier;

and S5, matching and storing the radio frequency tag identification result and the image identification result of the goods by the server, and sending the radio frequency tag identification result and the image identification result to the operation end for display to a user.

In some preferred embodiments, the method for error cancellation by the preprocessing module in step S3 using a method for minimizing the sum of squares of deviations includes:

S301, constructing a loss function J (theta) for measuring the difference between a cargo identification result and cargo actual information, wherein theta is an image identification neural network parameter;

s302, minimizing a loss function J (theta) by adopting a gradient descent method, wherein in each iteration, the parameter theta is as follows J (θ) update, where a is the learning rate,A bias to parameter θ _j for loss function J (θ);

s303, adjusting the image recognition neural network by using the updated theta;

s304, constructing and calculating an optimization objective function of the difference between the goods identification result and the goods actual information:

S305, repeating the steps until the optimization objective function is not changed, and finishing fine adjustment of the image recognition neural network.

In some preferred embodiments, the optimization objective function of the difference between the cargo identification result and the cargo actual information is:

argmin＝SST+W₁×SSE+W₂×SSA；

Wherein SST is a method for measuring all cargo identification result predicted values y _ij and cargo identification result ensemble average predicted values The sum of the squares of the deviations between them,SSE is the sum of squares measuring the difference between each cargo identification result predicted value y _ij and the cargo identification result average predicted value y _i of the group to which it belongs, SSE = Σ (y _ij―y_i)²; SSA is the sum of cargo identification result average predicted value y _i and cargo identification result ensemble average predicted value of different groups)The sum of the squares of the differences between them, W ₁ and W ₂ are weight parameters that measure the extent to which intra-and inter-group bias contribute to the total bias, respectively.

In some preferred embodiments, step S4 further includes the step of, before the server matches the storage policy and the scheduling policy of the warehouse-in goods in the warehouse management policy library according to the tag identification result and the status identification result of the warehouse-in goods after the errors are eliminated:

Matching the radio frequency tag identification result and the real-time image identification result of the warehouse-in goods, storing if the radio frequency tag identification number and the real-time image identification number of the warehouse-in goods are consistent, otherwise, giving an alarm; if the radio frequency tag identification category of the warehouse-in goods is consistent with the real-time image identification category, storing; otherwise, alarming.

In some preferred embodiments, the W ₁ and W ₂ are initialized by means of random assignments.

The foregoing has shown and described the basic principles, principal features and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, and that the above embodiments and descriptions are merely illustrative of the principles of the present invention, and various changes and modifications may be made without departing from the spirit and scope of the invention, which is defined in the appended claims. The scope of the invention is defined by the appended claims and equivalents thereof.

Claims (10)

1. Warehouse goods warehouse entry system based on radio frequency identification and image processing, characterized by comprising:

The identification area is arranged on the ground at the entrance of the warehouse, matches the size of the warehouse-in vehicle and is provided with markers around;

The periphery of the identification area is provided with a tag identifier for identifying the radio frequency tag arranged on the warehouse-in goods and a state identifier for identifying the real-time state of the warehouse-in goods; the goods real-time state comprises a goods real-time image and a temperature;

The server is respectively connected with the tag identifier and the state identifier in a signal way and is internally provided with a preprocessing module; the preprocessing module performs error elimination on data transmitted by the tag identifier and the state identifier by using a method of minimizing the sum of squares of deviation;

The operation end is in signal connection with the server and internally provided with a man-machine interaction interface;

and the server is also internally provided with a warehouse management strategy library, matches the storage strategy and the scheduling strategy of the warehouse goods in the warehouse management strategy library according to the label identification result and the state identification result of the warehouse goods, and sends the storage strategy and the scheduling strategy to the operation end for display to a user.

2. The system for warehousing goods based on radio frequency identification and image processing of claim 1, wherein the preprocessing module performs error cancellation on the data transmitted by the tag identifier and the status identifier by means of minimizing the sum of squares of deviations, comprising:

Respectively constructing a loss function J (theta) for measuring the difference between the identification results of the tag identifier and the state identifier and the actual information of the goods, wherein theta is the identification parameters of the tag identifier and the state identifier;

minimizing the loss function J (θ) using gradient descent, in each iteration, identifying the parameter θ as An update, where a is the learning rate,A bias to parameter θ _j for loss function J (θ);

Adjusting parameters of the tag identifier and the state identifier by using the updated theta;

Constructing and calculating an optimized objective function of the difference between the identification results of the tag identifier and the state identifier and the actual information of the goods;

Repeating the steps until the optimization objective function is not changed, and completing error elimination.

3. The warehouse entry system of claim 2, wherein the optimized objective function for the discrepancy between the cargo identification result and the cargo actual information is:

argmin＝SST+W₁×SSE+W₂×SSA；

Wherein SST is a method for measuring all cargo identification result predicted values y _ij and cargo identification result ensemble average predicted values The sum of the squares of the deviations between them,SSE is the sum of squares measuring the difference between each cargo identification result predicted value y _ij and the cargo identification result average predicted value y _i of the group to which it belongs, SSE = Σ (y _ij―y_i)²; SSA is the sum of cargo identification result average predicted value y _i and cargo identification result ensemble average predicted value of different groups)The sum of the squares of the differences between them,W ₁ and W ₂ are weight parameters that measure the extent to which intra-and inter-group bias contribute to the total bias, respectively.

4. The warehouse entry system of claim 1, wherein before the server matches the storage policy and the dispatch policy of the warehouse entry in the warehouse management policy base according to the tag identification result and the status identification result of the warehouse entry, further comprising the steps of:

Matching the radio frequency tag identification result and the real-time image identification result of the warehouse-in goods, storing if the radio frequency tag identification number and the real-time image identification number of the warehouse-in goods are consistent, otherwise, giving an alarm; if the radio frequency tag identification category of the warehouse-in goods is consistent with the real-time image identification category, storing; otherwise, alarming.

5. The radio frequency identification and image processing based warehouse entry system as set forth in claim 1, wherein: the W ₁ and the W ₂ are initialized by means of random assignment.

6. The warehouse-in method of the warehouse goods based on the radio frequency identification and the image processing is characterized by comprising the following steps:

S1, driving and parking a transport tool carrying warehouse-in goods into an identification area which is arranged on the ground at a warehouse entrance, matches the size of a warehouse-in vehicle and is provided with markers around the warehouse-in vehicle;

S2, controlling a tag identifier and a state identifier which are arranged at the periphery of the identification area to identify the radio frequency tag and the real-time state of the warehouse-in goods;

S3, a preprocessing module arranged in the server utilizes a method of minimizing the sum of squares of deviation to eliminate errors of data transmitted by the tag identifier and the state identifier;

and S4, the server matches the storage strategy and the scheduling strategy of the warehouse-in goods in the warehouse management strategy library according to the label identification result and the state identification result of the warehouse-in goods after the errors are eliminated, and sends the storage strategy and the scheduling strategy to the operation end for displaying to a user.

7. The method for warehousing goods based on radio frequency identification and image processing as set forth in claim 6, wherein the method for error elimination by the preprocessing module using the method of minimizing the sum of squares of deviation in step S3 includes:

s301, respectively constructing a loss function J (theta) for measuring the difference between the identification results of the tag identifier and the state identifier and the actual information of the goods, wherein theta is an image identification neural network parameter;

s302, minimizing a loss function J (theta) by adopting a gradient descent method, wherein in each iteration, the parameter theta is as follows An update, where a is the learning rate,A bias to parameter θ _j for loss function J (θ);

S303, performing parameter adjustment on the tag identifier and the state identifier by using the updated theta;

s304, constructing and calculating an optimized objective function of the difference between the identification results of the tag identifier and the state identifier and the actual information of the goods;

S305, repeating the steps until the optimization objective function is not changed, and completing error elimination.

8. The method for warehousing goods based on radio frequency identification and image processing as set forth in claim 7, wherein the optimized objective function of the difference between the goods identification result and the goods actual information is:

argmin＝SST+W₁×SSE+W₂×SSA；

Wherein SST is a method for measuring all cargo identification result predicted values y _ij and cargo identification result ensemble average predicted values The sum of the squares of the deviations between them,SSE is the sum of squares measuring the difference between each cargo identification result predicted value y _ij and the cargo identification result average predicted value y _i of the group to which it belongs, SSE = Σ (y _ij―y_i)²; SSA is the sum of cargo identification result average predicted value y _i and cargo identification result ensemble average predicted value of different groups)The sum of the squares of the differences between them,W ₁ and W ₂ are weight parameters that measure the extent to which intra-and inter-group bias contribute to the total bias, respectively.

9. The method for warehousing warehoused cargo based on radio frequency identification and image processing as defined in claim 6, wherein,

Step S4, before the server matches the storage strategy and the dispatching strategy of the warehouse-in goods in the warehouse management strategy library according to the label identification result and the state identification result of the warehouse-in goods after the errors are eliminated, the method further comprises the steps of:

Matching the radio frequency tag identification result and the real-time image identification result of the warehouse-in goods, storing if the radio frequency tag identification number and the real-time image identification number of the warehouse-in goods are consistent, otherwise, giving an alarm; if the radio frequency tag identification category of the warehouse-in goods is consistent with the real-time image identification category, storing; otherwise, alarming.

10. The method for warehousing the warehouse goods based on the radio frequency identification and the image processing as set forth in claim 7, wherein: the W ₁ and the W ₂ are initialized by means of random assignment.