An intelligent sorting method and system for goods based on dual recognition

Abstract

The present invention discloses an intelligent goods sorting method and system based on dual recognition, which relates to the field of goods sorting. Among them, the method includes: during the transmission of goods to be sorted, using a code scanning device to collect and obtain the identification code information on the goods to be sorted; inputting it into a pre-constructed identification code information space to determine whether a first recognition result is obtained; if so, sorting the goods to be sorted according to the first recognition result, and if not, continuing to transmit the goods to be sorted; during the transmission, using the image acquisition device to collect and obtain the image information of the goods to be sorted; inputting the image information into a pre-constructed image recognition model to obtain a second recognition result; sorting the goods to be sorted according to the second recognition result. Technical effects such as improving the accuracy and automation degree of goods recognition, and further improving the quality and efficiency of goods sorting are achieved.

Description

Technical Field

[0001] The present invention relates to the field of goods sorting, and in particular, to an intelligent goods sorting method and system based on dual recognition. Background Art

[0002] A large number of goods need to be sorted every day. The existing goods sorting methods rely highly on human labor, which not only has low efficiency but also is prone to errors. Goods sorting directly affects the work efficiency and production benefits of enterprises. Many enterprises urgently need a set of goods sorting methods with low cost and good performance. To meet this demand, designing an optimized goods sorting method has important practical significance.

[0003] In the prior art, there are technical problems such as insufficient accuracy in goods recognition and low automation degree, which in turn result in low quality and efficiency of goods sorting. Summary of the Invention

[0004] This application provides an intelligent goods sorting method and system based on dual recognition, which solves the technical problems in the prior art of insufficient accuracy in goods recognition and low automation degree, resulting in low quality and efficiency of goods sorting.

[0005] In view of the above problems, this application provides an intelligent goods sorting method and system based on dual recognition.

[0006] In a first aspect, this application provides an intelligent goods sorting method based on dual recognition. The method is applied to an intelligent goods sorting system based on dual recognition, and the method includes: transporting the goods to be sorted; during the transportation process, collecting and obtaining the identification code information on the goods to be sorted by using the code scanning device; inputting the identification code information into a pre-constructed identification code information space to determine whether a first recognition result is obtained; if so, sorting the goods to be sorted according to the first recognition result, and if not, continuing to transport the goods to be sorted; during the transportation process, collecting and obtaining the image information of the goods to be sorted by using the image acquisition device; inputting the image information into a pre-constructed image recognition model to obtain a second recognition result; and sorting the goods to be sorted according to the second recognition result.

[0007] Second aspect, the present application also provides an intelligent goods sorting system based on dual recognition. Among them, the system includes: a transmission module for transmitting the goods to be sorted; a scanning code module for collecting the identification code information on the goods to be sorted by using the scanning code device during the transmission process; a judgment module for inputting the identification code information into a pre-constructed identification code information space to judge whether a first recognition result is obtained; a sorting and transmission selection module for sorting the goods to be sorted according to the first recognition result if so, and continuing to transmit the goods to be sorted if not; an image acquisition module for collecting the image information of the goods to be sorted by using the image acquisition device during the transmission process; an input module for inputting the image information into a pre-constructed image recognition model to obtain a second recognition result; and a sorting module for sorting the goods to be sorted according to the second recognition result.

[0008] One or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0009] Collect the identification code information on the goods to be sorted through the scanning code device; input the identification code information into a pre-constructed identification code information space to judge whether a first recognition result is obtained; if so, sort the goods to be sorted according to the first recognition result, if not, continue to transmit the goods to be sorted; and during the transmission process, collect the image information of the goods to be sorted by using the image acquisition device; input the image information into a pre-constructed image recognition model to obtain a second recognition result; and sort the goods to be sorted according to it. The technical effects of improving the accuracy and automation degree of goods recognition, thereby improving the quality and efficiency of goods sorting are achieved; at the same time, the intelligence and scientific nature of goods sorting are improved, the cost of goods sorting is reduced, the manpower waste caused by goods sorting is reduced, and the technical effect of efficiently and quickly completing goods sorting is achieved. Description of the Drawings

[0010] Figure 1 It is a schematic flowchart of an intelligent goods sorting method based on dual recognition of the present application;

[0011] Figure 2 It is a schematic flowchart of obtaining the image information of the goods to be sorted in an intelligent goods sorting method based on dual recognition of the present application;

[0012] Figure 3 It is a schematic flowchart of constructing an image recognition model in an intelligent goods sorting method based on dual recognition of the present application;

[0013] Figure 4This is a schematic structural diagram of an intelligent goods sorting system based on dual recognition in the present application.

[0014] Explanation of reference numerals: Transmission module 11, barcode scanning module 12, judgment module 13, sorting transmission selection module 14, image acquisition module 15, input module 16, sorting module 17. Detailed implementation manners

[0015] The present application provides an intelligent goods sorting method and system based on dual recognition, which solves the technical problems of insufficient accuracy and low automation degree in the prior art for goods recognition, thus resulting in low quality and efficiency of goods sorting. It achieves the technical effects of improving the accuracy and automation degree of goods recognition, thereby improving the quality and efficiency of goods sorting; at the same time, enhancing the intelligence and scientific nature of goods sorting, reducing the cost of goods sorting, minimizing the waste of manpower caused by goods sorting, and efficiently and quickly completing goods sorting.

[0016] Embodiment 1

[0017] Please refer to the Figure 1 , the present application provides an intelligent goods sorting method based on dual recognition. Among them, the method is applied to an intelligent goods sorting system based on dual recognition. The system includes a barcode scanning device and an image acquisition device. The method specifically includes the following steps:

[0018] Step S100: Transmit the goods to be sorted.

[0019] Step S200: During the transmission process, use the barcode scanning device to collect and obtain the identification code information on the goods to be sorted.

[0020] Specifically, the goods to be sorted are transmitted through logistics transmission equipment such as a conveyor belt. At the same time, during the transmission of the goods to be sorted, the barcode scanning device is used to identify the goods to be sorted to obtain the identification code information. Among them, the goods to be sorted are any goods automatically sorted using the intelligent goods sorting system based on dual recognition. And, the surface of the goods to be sorted has an identification code, which corresponds to the goods to be sorted one by one. The identification code information includes data information such as the type information and sorting batch information of the identification code of the goods to be sorted recognized by the barcode scanning device. It achieves the technical effects of obtaining the identification code information of the goods to be sorted through the barcode scanning device, improving the automation degree of the collection of the identification code information, avoiding the waste of manpower caused by manual collection of the identification code information, as well as the problems of high cost and low efficiency in the collection of the identification code information; at the same time, improving the efficiency and accuracy of the collection of the identification code information.

[0021] The described code scanning device is included in the intelligent goods sorting system based on dual recognition. Exemplarily, the code scanning device consists of structures such as five code scanners and sensors. Through this code scanning device, five-sided scanning of the goods to be sorted can be performed. When the goods to be sorted pass through the sensor in front of the code scanner, the code scanner is triggered to perform code scanning. The five code scanners start to scan the goods to be sorted simultaneously. One of the code scanners is the main code scanner, and the main code scanner can summarize the identification code information obtained by the other four code scanners and upload all the identification code information to the intelligent goods sorting system based on dual recognition.

[0022] After the code scanning device is installed, in order to ensure the accuracy of the obtained identification code information and prevent the impact on goods sorting due to low-quality identification code information, it is usually necessary to perform static debugging and dynamic testing on the code scanners in the code scanning device. Static debugging refers to performing a series of parameter settings on the code scanners through the software Dataman provided by the code scanner supplier Cognex. The parameter settings include the focal length, aperture size, exposure value size, gain size of the code scanner, and the selection setting of the main code scanner. At the same time, the parameter settings also include the selection of the type and quantity of the identification codes to be recognized by the code scanner, the screening of the number of digits of the identification codes to be recognized and the content to be displayed. In addition, static debugging also includes adjusting the installation position of the code scanner. Dynamic testing refers to using the code scanning device that has completed static debugging to collect the identification codes of the goods being transmitted, obtain the identification code information of the goods, and judge the quality parameters such as the clarity and integrity of the identification code information. If the clarity of the identification code information is not high and the integrity is insufficient, the code scanning device is statically debugged again until the code scanning device achieves a better code scanning effect.

[0023] Step S300: Input the identification code information into the pre-constructed identification code information space and determine whether a first recognition result is obtained;

[0024] Further, step S300 of the present application further includes:

[0025] Step S310: Construct the identification code information space;

[0026] Further, step S310 of the present application further includes:

[0027] Step S311: According to the identification codes of multiple types of goods, collect and obtain the types of multiple types of goods to obtain multiple type information;

[0028] Step S312: According to the identification codes of multiple types of goods, collect and obtain the batches of multiple types of goods to obtain multiple batch information;

[0029] Step S313: Obtain multiple entity information according to the multiple types of goods;

[0030] Step S314: Obtain a first attribute and multiple pieces of first attribute value information according to the multiple types of information;

[0031] Step S315: Obtain a second attribute and multiple pieces of second attribute value information according to the multiple batch information;

[0032] Step S316: Based on the multiple entity information, the first attribute, the multiple pieces of first attribute value information, the second attribute, and the multiple pieces of second attribute value information, construct the identification code information space based on the knowledge graph.

[0033] Specifically, based on the identification codes of multiple types of goods, the intelligent goods sorting system based on dual recognition collects information on the types and batches of multiple types of goods to obtain multiple types of information and multiple batch information. Further, multiple entity information is determined according to multiple types of goods. According to the multiple types of information, a first attribute and multiple pieces of first attribute value information are obtained. According to the multiple batch information, a second attribute and multiple pieces of second attribute value information are obtained. Based on this, combined with the idea of the knowledge graph, the identification code information space is constructed. Among them, the identification codes of the multiple types of goods include multiple identification codes corresponding to multiple types of goods sorted automatically using the intelligent goods sorting system based on dual recognition. The multiple types of information include multiple goods type information corresponding to the identification codes of multiple types of goods. The multiple batch information includes the sorting batch information corresponding to the identification codes of multiple types of goods. The sorting batch information includes the sorting line information corresponding to the identification codes of multiple types of goods. The multiple entity information includes multiple types of goods. The first attribute is the type of multiple types of goods. The multiple pieces of first attribute value information include multiple types of information. The second attribute is the batch of multiple types of goods. The multiple pieces of second attribute value information include multiple batch information. The knowledge graph is a way of expressing data information. The knowledge graph includes a schema layer and a data layer. The data layer consists of a series of facts; the schema layer is built on top of the data layer and is mainly used to express a series of facts in the data layer in a standardized manner. The identification code information space includes multiple entity information, a first attribute, multiple pieces of first attribute value information, a second attribute, multiple pieces of second attribute value information, and the corresponding relationships between them. It achieves the technical effect of constructing a reliable identification code information space through multiple types of information and multiple batch information, thereby improving the accuracy of the subsequent obtained first recognition result.

[0034] Step S320: Input the identification code information into the identification code information space for mapping correspondence, and determine whether there is a mapping result;

[0035] Step S330: If so, use the mapping result as the first recognition result;

[0036] Step S340: If not, the first recognition result is not obtained.

[0037] Specifically, the obtained identification code information is input into the constructed identification code information space for mapping correspondence, and it is determined whether there is a mapping result for the identification code information, that is, it is determined whether the identification code information has a corresponding relationship with multiple entity information, a first attribute, multiple first attribute value information, a second attribute, and multiple second attribute value information in the identification code information space. Further, if there is a mapping result for the identification code information, the mapping result is set as the first recognition result. If there is no mapping result for the identification code information, the first recognition result cannot be obtained. Wherein, the mapping result includes the entity information, the first attribute, the first attribute value information, the second attribute, and the second attribute value information corresponding to the identification code information. The first recognition result is the mapping result. It achieves the technical effect of accurately mapping and judging the identification code information through the identification code information space, adaptively obtaining the first recognition result, and improving the accuracy of goods sorting.

[0038] Step S400: If so, sort the goods to be sorted according to the first recognition result; if not, continue to transport the goods to be sorted.

[0039] Specifically, when judging whether the first recognition result is obtained, if the first recognition result is obtained, the goods to be sorted are sorted according to the first recognition result. If the first recognition result is not obtained, the goods to be sorted are continuously transported. It achieves the technical effect of adaptively sorting or transporting the goods to be sorted according to whether the first recognition result is obtained, and improving the accuracy of goods sorting.

[0040] Step S500: During the transportation process, use the image acquisition device to acquire the image information of the goods to be sorted.

[0041] Further, as shown in the appendix Figure 2 This application's step S500 further includes:

[0042] Step S510: Obtain multiple preset angles;

[0043] Step S520: Based on the multiple preset angles, use the image acquisition device to acquire multiple angle image information of the goods to be sorted;

[0044] Step S530: Use the multiple angle image information as the image information.

[0045] Specifically, for the goods to be sorted that have not obtained the first recognition result, during the continuous transmission of the goods to be sorted, multiple-angle image information of the goods to be sorted is collected by using an image acquisition device at multiple preset angles, and the image information of the goods to be sorted is obtained. Among them, the multiple preset angles include multiple image acquisition angle information of the image acquisition device. The multiple preset angles are pre-set and determined by the intelligent goods sorting system based on dual recognition. The image information of the goods to be sorted includes multiple-angle image information. The multiple-angle image information includes image data information of the goods to be sorted corresponding to multiple preset angles of the image acquisition device. It achieves the technical effect of collecting images of the goods to be sorted that have not obtained the first recognition result at multiple preset angles, obtaining the image information of the goods to be sorted, and laying a foundation for obtaining the second recognition result later.

[0046] The image acquisition device is included in the intelligent goods sorting system based on dual recognition. Exemplarily, the image acquisition device includes a camera and a sensor. To improve the clarity of the obtained image information of the goods to be sorted, a reflector and an LED lamp tube can be installed near the goods to be sorted to increase the brightness of the goods to be sorted, making the obtained image information of the goods to be sorted clearer and easier for image recognition. When the goods to be sorted pass through the sensor behind the camera, the camera is triggered to take pictures, and the camera can take pictures of the top and side of the goods to be sorted. The installation position, focal length, aperture, exposure value, and RGB value of the camera can also be adjusted so that the image information of the goods to be sorted is in the middle of the field of view and clearly visible.

[0047] Step S600: Input the image information into a pre-constructed image recognition model to obtain a second recognition result;

[0048] Further, step S600 of the present application further includes:

[0049] Step S610: Construct the image recognition model;

[0050] Further, as shown in the appendix Figure 3 Step S610 of the present application further includes:

[0051] Step S611: Collect and obtain image information of multiple types of goods to obtain a sample image information set;

[0052] Step S612: Collect and obtain the sorting category information of the multiple types of goods to obtain a sample sorting category information set;

[0053] Step S613: Divide and combine the sample image information set and the sample sorting category information set to obtain a first data set and a second data set;

[0054] Step S614: Build an image recognition model based on a deep convolutional neural network;

[0055] Step S615: Divide and label the first data set to obtain a training set, a validation set, and a test set;

[0056] Step S616: Use the training set to perform supervised training on the image recognition model until convergence or the accuracy rate reaches a preset threshold;

[0057] Step S617: Use the validation set and the test set to verify and test the image recognition model. If the accuracy rate meets the preset threshold, obtain the image recognition model.

[0058] Specifically, the intelligent goods sorting system based on dual recognition collects image information and sorting category information of multiple types of goods, obtains a sample image information set and a sample sorting category information set, divides and combines them to obtain a first data set and a second data set. Further, divide and label the first data set to obtain a training set, a validation set, and a test set. Then, use the training set to perform supervised training on the image recognition model until convergence or the accuracy rate reaches a preset threshold. After that, input the validation set and the test set into the image recognition model for verification and testing respectively. If the accuracy rate meets the preset threshold, obtain the image recognition model. Among them, the sample image information set includes the image information of multiple types of goods. The sample sorting category information set includes the sorting category information of multiple types of goods. The first data set includes part of the data information of the sample image information set and part of the data information of the sample sorting category information set. The second data set includes part of the data information of the sample image information set and part of the data information of the sample sorting category information set. And the first data set is different from the second data set. The image recognition model is a deep convolutional neural network model. The accuracy rate is a parameter information used to characterize the similarity between the sorting category information output by the image recognition model and the sorting category information in the sample sorting category information set of the input training set, validation set, and test set. The higher the similarity between the two, the higher the corresponding accuracy rate. The preset threshold is adaptively set and determined by the intelligent goods sorting system based on dual recognition according to the accuracy requirements of the image recognition model. It achieves the technical effect of building an image recognition model and providing data support for obtaining the second recognition result later.

[0059] Step S620: Verify the stability of the image recognition model. If the stability meets the preset requirements, put the image recognition model into use;

[0060] Furthermore, step S620 of the present application further includes:

[0061] Step S621: Divide and label the second data set to obtain a first stability verification set and a second stability verification set;

[0062] Step S622: Input the first stability verification set and the second stability verification set into the image recognition model respectively to obtain a plurality of first recognition results and a plurality of second recognition results, and obtain a first sorting category set and a second sorting category set;

[0063] Step S623: Calculate the stability of the image recognition model according to the first sorting category set and the second sorting category set by the following formula:

[0064]

[0065] where P i is the proportion of the i-th sorting category information in the first sorting category set, Q i is the proportion of the i-th sorting category information in the second sorting category set, and n is the number of types of sorting category information.

[0066] Specifically, after dividing and labeling the obtained second data set, a first stability verification set and a second stability verification set are obtained. Further, the first stability verification set and the second stability verification set are used as input information and input into the image recognition model respectively for predicting and recognizing sorting categories, obtaining a plurality of first recognition results and a plurality of second recognition results. Based on this, a first sorting category set and a second sorting category set are determined. After calculating in combination with the above stability calculation formula, the stability of the image recognition model is calculated, and it is judged whether the stability meets the preset requirements. If the stability meets the preset requirements, the image recognition model is put into use. Among them, the first stability verification set and the second stability verification set are included in the second data set. The plurality of first recognition results include a plurality of sorting category information output by the image recognition model corresponding to the first stability verification set. The plurality of second recognition results include a plurality of sorting category information output by the image recognition model corresponding to the second stability verification set. The first sorting category set includes a plurality of first recognition results. The second sorting category set includes a plurality of second recognition results. The preset requirements include a preset stability threshold, which is determined in advance by the intelligent goods sorting system based on dual recognition.

[0067] Specifically, since the first stability verification set and the second stability verification set are obtained by random partitioning, within the first stability verification set and the second stability verification set, the proportion of the image information of the goods of the same sorting category should be similar. Therefore, within the first sorting category set and the second sorting category set obtained by predictive recognition, the proportion of the same sorting category is also similar. Thus, the stability of the image recognition model is calculated based on the above formula. The preset requirement for this stability is preferably 0.3. If the calculated stability R value of the image recognition model is less than 0.3, then the preset requirement is met and it can be put into use.

[0068] It achieves the technical effect of using the second data set to verify the stability of the constructed image recognition model, obtaining an image recognition model whose stability meets the preset requirements, and thereby improving the accuracy of the subsequent obtained second recognition result.

[0069] Step S630: Input the image information into the image recognition model that has been put into use to obtain the second recognition result.

[0070] Step S700: Sort the goods to be sorted according to the second recognition result.

[0071] Specifically, the image information of the goods to be sorted is used as the input information and input into the image recognition model to obtain the second recognition result, and the goods to be sorted are sorted according to the second recognition result. Among them, the second recognition result includes the sorting category information corresponding to the image information of the goods to be sorted. It achieves the technical effect of obtaining an accurate second recognition result through the image recognition model, and thereby improving the quality and efficiency of goods sorting.

[0072] In summary, a goods intelligent sorting method based on dual recognition provided by the present application has the following technical effects:

[0073] 1. Collect and obtain the identification code information on the goods to be sorted through a barcode scanning device; input the identification code information into the pre-constructed identification code information space to determine whether the first recognition result is obtained; if so, sort the goods to be sorted according to the first recognition result, if not, continue to transport the goods to be sorted; and during the transportation process, use the image acquisition device to collect and obtain the image information of the goods to be sorted; input the image information into the pre-constructed image recognition model to obtain the second recognition result; and sort the goods to be sorted according to it. It achieves the technical effects of improving the accuracy and automation degree of goods recognition, and thereby improving the quality and efficiency of goods sorting; at the same time, improving the intelligence and scientific nature of goods sorting, reducing the cost of goods sorting, reducing the manpower waste caused by goods sorting, and efficiently and quickly completing goods sorting.

[0074] 2. Obtain the identification code information of the goods to be sorted through the code scanning device, improve the automation degree of the collection of the identification code information, avoid the waste of manpower caused by manual collection of the identification code information, as well as the problems of high cost and low efficiency in the collection of the identification code information; at the same time, improve the efficiency and accuracy of the collection of the identification code information.

[0075] Embodiment 2

[0076] Based on the intelligent goods sorting method based on dual recognition in the foregoing embodiment and the same inventive concept, the present invention also provides an intelligent goods sorting system based on dual recognition. Please refer to the appendix Figure 4 , the system includes:

[0077] A transmission module 11, which is used to transmit the goods to be sorted;

[0078] A code scanning module 12, which is used to collect and obtain the identification code information on the goods to be sorted by using the code scanning device during the transmission process;

[0079] A judgment module 13, which is used to input the identification code information into a pre-constructed identification code information space and judge whether a first recognition result is obtained;

[0080] A sorting and transmission selection module 14, which is used to, if so, sort the goods to be sorted according to the first recognition result, and if not, continue to transmit the goods to be sorted;

[0081] An image acquisition module 15, which is used to collect and obtain the image information of the goods to be sorted by using the image acquisition device during the transmission process;

[0082] An input module 16, which is used to input the image information into a pre-constructed image recognition model to obtain a second recognition result;

[0083] A sorting module 17, which is used to sort the goods to be sorted according to the second recognition result.

[0084] Further, the system further includes:

[0085] An identification code information space construction module, which is used to construct the identification code information space;

[0086] A judgment module, which is used to input the identification code information into the identification code information space for mapping correspondence and judge whether there is a mapping result;

[0087] The first recognition result obtaining module, which is used to, if so, use the mapping result as the first recognition result;

[0088] The first execution module, which is used to, if not, fail to obtain the first recognition result.

[0089] Furthermore, the system further includes:

[0090] The type information determination module, which is used to collect and obtain the types of multiple categories of goods according to the identification codes of the multiple categories of goods, and obtain multiple type information;

[0091] The batch information determination module, which is used to collect and obtain the batches of multiple categories of goods according to the identification codes of the multiple categories of goods, and obtain multiple batch information;

[0092] The entity information determination module, which is used to obtain multiple entity information according to the multiple categories of goods;

[0093] The first attribute information determination module, which is used to obtain a first attribute and multiple first attribute value information according to the multiple type information;

[0094] The second attribute determination module, which is used to obtain a second attribute and multiple second attribute value information according to the multiple batch information;

[0095] The identification code information space determination module, which is used to construct the identification code information space based on the multiple entity information, the first attribute, the multiple first attribute value information, the second attribute, and the multiple second attribute value information, based on the knowledge graph.

[0096] Furthermore, the system further includes:

[0097] The preset angle determination module, which is used to obtain multiple preset angles;

[0098] The angle image information determination module, which is used to collect and obtain multiple angle image information of the goods to be sorted through the image acquisition device based on the multiple preset angles;

[0099] The image information determination module, which is used to use the multiple angle image information as the image information.

[0100] Furthermore, the system further includes:

[0101] The second execution module, which is used to construct the image recognition model;

[0102] A stability verification module, which is used to verify the stability of the image recognition model. If the stability meets the preset requirements, the image recognition model will be put into use;

[0103] A third execution module, which is used to input the image information into the image recognition model that has been put into use to obtain the second recognition result.

[0104] Furthermore, the system further includes:

[0105] A sample image information set determination module, which is used to collect and obtain the image information of multiple types of goods to obtain a sample image information set;

[0106] A sample sorting category information set determination module, which is used to collect and obtain the sorting category information of the multiple types of goods to obtain a sample sorting category information set;

[0107] A data set determination module, which is used to divide and combine the sample image information set and the sample sorting category information set to obtain a first data set and a second data set;

[0108] A fourth execution module, which is used to build an image recognition model based on a deep convolutional neural network;

[0109] A first partitioning and identification module, which is used to partition and data-identify the first data set to obtain a training set, a validation set, and a test set;

[0110] A supervised training module, which is used to perform supervised training on the image recognition model using the training set until convergence or the accuracy rate reaches a preset threshold;

[0111] A verification and testing module, which is used to verify and test the image recognition model using the validation set and the test set. If the accuracy rate meets the preset threshold, the image recognition model will be obtained.

[0112] Furthermore, the system further includes:

[0113] A second partitioning and identification module, which is used to partition and data-identify the second data set to obtain a first stability verification set and a second stability verification set;

[0114] The sorting category set determination module is configured to input the first stability verification set and the second stability verification set into the image recognition model respectively, obtain a plurality of first recognition results and a plurality of second recognition results respectively, and obtain a first sorting category set and a second sorting category set;

[0115] The fifth execution module is configured to calculate the stability of the image recognition model according to the first sorting category set and the second sorting category set by the following formula:

[0116]

[0117] where P i is the proportion of the i-th sorting category information in the first sorting category set, and Q i is the proportion of the i-th sorting category information in the second sorting category set, and n is the number of types of sorting category information.

[0118] The present application provides a method for intelligent sorting of goods based on dual recognition. The method is applied to a system for intelligent sorting of goods based on dual recognition. The method includes: collecting the identification code information on the goods to be sorted through a scanning device; inputting the identification code information into a pre-constructed identification code information space to determine whether a first recognition result is obtained; if so, sorting the goods to be sorted according to the first recognition result, and if not, continuing to transport the goods to be sorted; and during the transportation process, collecting the image information of the goods to be sorted by using the image acquisition device; inputting the image information into a pre-constructed image recognition model to obtain a second recognition result; and sorting the goods to be sorted according to the second recognition result. This solves the technical problems in the prior art that the accuracy of goods recognition is insufficient and the automation degree is not high, resulting in low quality and efficiency of goods sorting. It achieves the technical effects of improving the accuracy and automation degree of goods recognition, thereby improving the quality and efficiency of goods sorting; at the same time, improving the intelligence and scientificity of goods sorting, reducing the cost of goods sorting, reducing the waste of manpower caused by goods sorting, and efficiently and quickly completing the goods sorting.

[0119] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered that the scope described in this specification is covered.

[0120] This specification and the drawings are only exemplary descriptions of the present application. If the modifications and variations of the present invention fall within the scope of the present invention and its equivalent technologies, the present invention also intends to include these modifications and variations.

Claims

1. An intelligent sorting method for goods based on dual recognition, characterized in that, the method is applied to an intelligent sorting system for goods based on dual recognition, the system includes a barcode scanning device and an image acquisition device, and the method includes: Transmitting the goods to be sorted; During the transmission process, using the barcode scanning device to collect and obtain the barcode information on the goods to be sorted; Inputting the barcode information into a pre-constructed barcode information space, and judging whether a first recognition result is obtained, including: Constructing the barcode information space; Inputting the barcode information into the barcode information space for mapping correspondence, and judging whether there is a mapping result; if so, sorting the goods to be sorted according to the first recognition result, if not, continuing to transmit the goods to be sorted; During the transmission process, using the image acquisition device to collect and obtain the image information of the goods to be sorted; Inputting the image information into a pre-constructed image recognition model to obtain a second recognition result; Sorting the goods to be sorted according to the second recognition result; Constructing the barcode information space includes: According to the barcodes of multiple types of goods, collecting and obtaining the types of multiple types of goods to obtain multiple type information; According to the barcodes of multiple types of goods, collecting and obtaining the batches of multiple types of goods to obtain multiple batch information; According to the multiple types of goods, obtaining multiple entity information; According to the multiple type information, obtaining a first attribute and multiple first attribute value information; According to the multiple batch information, obtaining a second attribute and multiple second attribute value information; Based on the multiple entity information, first attribute, multiple first attribute value information, second attribute and multiple second attribute value information, constructing the barcode information space based on a knowledge graph; Obtaining multiple preset angles; Based on the multiple preset angles, using the image acquisition device to collect and obtain multiple angle image information of the goods to be sorted; Taking the multiple angle image information as the image information; Inputting the image information into a pre-constructed image recognition model, including: Constructing the image recognition model; Verifying the stability of the image recognition model, if the stability meets the preset requirements, putting the image recognition model into use; Inputting the image information into the image recognition model put into use to obtain the second recognition result; Constructing the image recognition model includes: Collecting and obtaining the image information of multiple types of goods to obtain a sample image information set; Collecting and obtaining the sorting category information of the multiple types of goods to obtain a sample sorting category information set; Dividing and combining the sample image information set and the sample sorting category information set to obtain a first data set and a second data set; Based on a deep convolutional neural network, constructing an image recognition model; Verifying the stability of the image recognition model includes: Dividing and data-labeling the second data set to obtain a first stability verification set and a second stability verification set; Input the first stability verification set and the second stability verification set into the image recognition model respectively, obtain a plurality of first recognition results and a plurality of second recognition results respectively, and obtain a first sorting category set and a second sorting category set; Calculate the stability of the image recognition model according to the first sorting category set and the second sorting category set by the following formula: ； wherein, is the proportion of the i-th sorting category information in the first sorting category set, is the proportion of the i-th sorting category information in the second sorting category set, is the number of types of sorting category information, is the stability value of the image recognition model.

2. The method according to claim 1, wherein, Constructing the image recognition model includes: Dividing and data-labeling the first data set to obtain a training set, a verification set and a test set; Supervising and training the image recognition model with the training set until convergence or the accuracy rate reaches a preset threshold; Verifying and testing the image recognition model with the verification set and the test set, and if the accuracy rate meets the preset threshold, obtain the image recognition model.

3. An intelligent goods sorting system based on dual recognition, wherein, The system is used to execute the method according to any one of claims 1 to 2. The system includes a barcode scanning device and an image acquisition device, and the system further includes: A transmission module for transmitting the goods to be sorted; A barcode scanning module for acquiring the barcode information on the goods to be sorted by using the barcode scanning device during the transmission process; A judgment module for inputting the barcode information into a pre-constructed barcode information space to judge whether a first recognition result is obtained; A sorting transmission selection module for, if so, sorting the goods to be sorted according to the first recognition result, and if not, continuing to transmit the goods to be sorted; An image acquisition module for acquiring the image information of the goods to be sorted by using the image acquisition device during the transmission process; An input module for inputting the image information into a pre-constructed image recognition model to obtain a second recognition result; A sorting module for sorting the goods to be sorted according to the second recognition result.

Images