TWI782358B - Inventory management method and system - Google Patents

Abstract

The present invention is an inventory management method and system, which includes using a photographic device to shoot a target object to generate an image, and then using an image recognition module to analyze the image to determine the target category and target outline of the target object, and then use the regression module to The area of the target contour is substituted into the corresponding linear regression equation to obtain the target weight of the target object, and finally the target weight is transmitted to the management server, and the target weight is displayed by the interface module in it, so as to achieve the monitoring and optimization of the inventory of the object The purpose of cargo transportation arrangement route.

Description

Inventory management method and system

The invention belongs to the field of environmental protection, in particular to a stock management method and system for estimating waste stock by images.

With the gradual emphasis on environmental protection in our country, the control over the removal and disposal of industrial waste is becoming more and more stringent. There is now a "Waste Disposal Law" that stipulates that the removal and disposal of industrial waste should be carried out by local environmental protection agencies. Bureaus, or private waste removal and disposal organizations approved by the competent authority, are responsible for removal and disposal.

For this reason, general business organizations are used to storing waste in one place, and then manually judge the waste stock, and after the stock reaches a certain threshold, they will notify the waste removal and processing agency to transport the waste to the designated place for disposal. deal with.

However, there is an information gap between waste removal and treatment agencies and business groups. Before receiving the notification, waste removal and treatment agencies cannot know the real-time waste inventory of business groups, so that they cannot efficiently arrange waste transportation schedules. , resulting in a waste of transportation resources, so how to fill up the information gap between waste removal and treatment organizations and business groups is still an urgent problem to be solved.

In view of the various shortcomings derived from the above-mentioned prior art, the inventor of this case is eager to improve and innovate, and after years of painstaking research, he finally successfully developed the inventory management method and system of the present invention.

In order to solve the problems of the above-mentioned prior art, the present invention provides an inventory management method and system, the purpose of which is to: 1. Provide a method and system for automatic identification of waste types; 2. Provide a method and system for real-time monitoring of waste inventory; 3 .Provide an information that helps to optimize the waste disposal process.

The present invention provides a stock management method, which includes using a photographic device to shoot a target object to generate an image, and then using a mask-type regional convolutional neural network (Region Based Convolutional Neural Networks, R-CNN, Mask-RCNN) as an image recognition module ) to analyze the image to determine the target category and target contour of the target object, and then use the regression module to substitute the area of the target contour as a variable into the corresponding linear regression equation to obtain the target weight of the target object, and finally transmit the target weight To manage the server, and use the interface module to display the target weight, so as to achieve the purpose of monitoring the stock of the target object.

Wherein, the target contour is the contour of the target object, or the contour formed by stacking a plurality of target objects.

Wherein, before the regression module substitutes the area of the target contour into the variable linear regression equation, it further performs binarization analysis on the target object according to the characteristic value of the target object.

Wherein, the feature value is the color of the target object on the image.

The present invention also provides an inventory management system, including photographic equipment, a data server, a computing server, and a management server. The inventory management system of the present invention can run the inventory management method of the present invention to achieve the purpose of monitoring the inventory of target objects.

Wherein, the computing server includes an image recognition module and a regression module.

The inventory management system and method as mentioned above can provide the waste inventory of relevant business groups of waste removal and treatment institutions, so as to eliminate the information gap between waste removal and treatment institutions and business groups, and provide information that helps to optimize waste transportation. Arranged waste inventory information.

1: Photographic equipment

2: Data server

3: Operation server

31: Image recognition module

32: Regression Mod

4: Manage Server

41:Interface module

5: Image

6: waste circuit board

7: Discarded cartons

8: Target object

81:Target outline

82:Secondary image

821: The first segmented two-quality image

822:Second segmented two-quality image

823: The third split two-quality image

824: The fourth split two-quality image

9: Target object

91:Target outline

92:Secondary image

93:Advanced two-quality image

S201-S204: Steps

S301-S305: Steps

S401-S404: Steps

Figure 1 is a schematic diagram of the inventory management system of the present invention; Figure 2 is a step diagram of the inventory management method of the present invention; Figure 3 is a diagram of the image recognition steps of the present invention; Figure 4 is a diagram of the weight calculation steps of the present invention; Figure 5 is a diagram of the present invention Figure 6 is a schematic diagram of the image recognition results of the embodiment of the present invention; Figure 7 is a schematic diagram of the binary image of the target object in the embodiment of the present invention; Figure 8 is a schematic diagram of the segmentation of the binary image of the target object in the embodiment of the present invention.

Please refer to FIG. 1 , which is a schematic diagram of the inventory management system of the present invention. The inventory management system of the present invention includes a photographic device 1 , a data server 2 , a computing server 3 and a management server 4 .

Wherein, the photographing device 1 is a photographing device erected in the waste storage area, and the photographing device 1 photographs the area and the wastes therein around the clock to generate continuous images.

Wherein, the data server 2 comprises a storage element and a transmission element, and the transmission element connects the photographic equipment 1 with the data server 2, and receives images, and the data server 2 stores the received images in the storage element for further use. The calculation server 3 or the user reads.

Wherein, the computing server 3 is connected to the data server 2, and the computing server 3 includes an image recognition module 31 and a regression module 32, and in one embodiment of the present invention, the image recognition module 31 is a trained Mask Region Based Convolutional Neural Networks (Mask RCNN).

Wherein, when the calculation server 3 reads the image from the data server 2, the image recognition module 31 recognizes the waste as the target object, and obtains the type and outline of the waste. The regression module 32 selects a specific linear regression equation from the storage element according to the category, and substitutes the area of the contour into the linear regression equation to calculate the weight of the waste, and then uses the transmission element to transmit the weight of the waste. to the management server4.

Wherein, the management server 4 includes an interface module 41 through which the user can monitor the weight of the waste, and use the weight of the waste as an index for judging the waste stock, so as to achieve the purpose of waste stock management.

Please refer to FIG. 2 , which is a step diagram of the inventory management method of the present invention. The inventory management method of the present invention includes:

S201: Provide an image containing a target object and a background object;

S202: Use the image recognition module 31 to determine the target category and target outline according to the feature values of the target object in the image;

S203: Use the regression module 32 to select an appropriate linear regression equation according to the target category, and use the area of the target outline as a variable to calculate the target weight.

S204: Upload the target weight to the management server 4 for the user to monitor the target weight.

Please refer to FIG. 3 , which is a diagram of the image recognition steps of the present invention. The judging method S202 of the image recognition module 31 of the present invention includes: S301: Input the image into a feature pyramid network (Feature Pyramid Network, FPN) with a residual learning network (Residual Network, ResNet) as the backbone, and generate a feature pyramid ( Feature Pyramid); S302: Input the feature pyramid to the Region Proposal Network (RPN), perform binarization and frame regression (Bounding Box Regression), and generate a Region Of Interest (ROI); S303: Use the ROIAlign (Region Of Interest Alignment) method to extract the feature value of the ROI, and then proceed to step S304 and step S305 at the same time; S304: classify the ROI with Faster-RCNN, and judge the target category; S305: use the fully convolutional network (Fully Convolutional Networks, FCN) accurately segment the ROI to generate the target contour.

Wherein, before the image recognition module 31 performs the judgment method S202, the present invention further trains the image recognition module 31 with labeled images until the correct rate of the image recognition module 31 in the target category and target outline reaches 85%. until.

Please refer to Fig. 4, which is a diagram of the weight calculation steps of the present invention. The weight calculation S203 of the regression module 32 of the present invention includes: S401: According to the target category of the target object, select the corresponding linear regression equation from the database; S402: Binarize the target object of the target outline to generate a binary image ; S402: According to the number of variables of the linear regression equation, cut the binary image to generate a plurality of partial contours; S403: Use the area of each partial contour in the cut binary image as a variable, substitute into the linear regression equation, and calculate the target The weight of the object. Among them, the formula of the linear regression equation is: weight = α ₀ x ₀ + α ₁ x ₁ +…+ α _K x _K + β weight: the weight of the target object; α: the coefficient set corresponding to the target category; x: variable; β: constant.

Among them, before the regression module 32 performs the weight calculation S203, the present invention further adjusts the coefficient α, the variable x and the constant β of the linear regression equation based on the actual weighing result until the image recognition module 31 is in the target category and Until the correct rate of the target contour reaches 85%, the adjusted linear regression equation is stored in the data server 2 .

Please refer to FIG. 5 , which is a schematic diagram of an image of an embodiment of the present invention. In the embodiment of the present invention, the present invention uses the photography equipment 1 to photograph the waste storage area, and generates an image 5, and in the waste storage area, there are waste circuit boards 6 and waste cartons 7, wherein the target object 8 is A waste pile formed by a plurality of waste circuit boards 6 .

Please refer to FIG. 6 , which is a schematic diagram of an image recognition result according to an embodiment of the present invention. After the regression module 32 of the present invention reads the image 5, it proceeds to the judgment method S202, and judges that the target category of the target object 8 is a waste circuit board 6, while the target category of the waste carton 7 is unknown, so only the target category of the target object 8 The target contour 81 is generated around, and then the regression module 32 is used to perform weight calculation S203.

Among them, the regression module 32 of the present invention selects the circuit board linear regression equation according to the result that the target category is the discarded circuit board 6, the formula is: weight _PCB = α _{P 0} x _{P 0} + α _{P 1} x _{P 1} + α _{P 2} x _{P 2} + α _{P 3} x _{P 3} + β _P weight _PCB : the weight of the target object whose target category is scrap circuit board; α _P : the coefficient set for the target category of scrap circuit board; x _P : the target category of scrap circuit variable for the board; β _P : constant for the target category of discarded boards.

Please refer to FIGS. 7-8 , which are schematic diagrams of the two-quality image of the target object and its cutting according to the embodiment of the present invention. The regression module 32 of the present invention binarizes the target object 8 in the target outline 81 to generate a binary image 82, and because the circuit board linear regression equation has 4 variables, the regression module 32 binarizes the binary image 82 is cut into a total of four images, the first split binary image 821, the second split binary image 822, the third split binary image 823, and the fourth split binary image 824, and then use the bright area of each split binary image is a variable, which is substituted into the board linear regression equation to calculate the target weight.

Wherein, the area of the bright part of the segmented binary image is the partial contour of this embodiment.

In the embodiment of the present invention, the calculation server 3 transmits the target weight to the management server 4, and the business organization or waste removal and processing organization can check the target weight through the interface module 41, so as to achieve the purpose of managing the waste inventory Purpose.

In the stock management method and system of the present invention, the image in the business group is sent back to the waste removal and processing organization in real time through the photographic equipment, and the image recognition module installed in the computing server is used to identify the type of waste, and the regression model The group calculates the weight of waste to achieve the purpose of monitoring waste stock.

In addition, the inventory management method and system of the present invention also includes a management server that provides an interface module to facilitate waste removal and disposal organizations to check the waste inventory, so as to eliminate the information gap between business groups and removal organizations.

Furthermore, the waste removal and processing organization can use the above waste inventory information to arrange waste transportation routes, recover waste in the most efficient way, and reduce transportation costs.

The above is only a preferred embodiment of the present invention, and is not used to limit the implementation scope of the present invention; if it does not depart from the spirit and scope of the present invention, any modification or equivalent replacement of the present invention shall be covered by the protection of the patent scope of the present invention in the range.

S201-S204: Steps

Claims (10)

A stock management method: providing an image, the image contains a target object; using an image recognition module, according to the feature value of the target object, and judging the target category and target outline of the target object, and eliminating the inability to determine the target category using a regression module, select a linear regression equation according to the target category, and then cut the target contour according to the number of variables in the linear regression equation to generate a plurality of partial contours, and use the area of each partial contour As a variable, calculate the target weight of the target object; transmit the target weight to a management server, and use the target weight to represent the stock of the target object; wherein, the linear regression equation is:

Weight : target weight; class : target category; k : number of variables; a _i : coefficient of the i -th variable; x _i : i -th variable; β : constant. The inventory management method as described in Claim 1, wherein the image recognition module is a trained Region Based Convolutional Neural Networks (R-CNN) module. The inventory management method according to Claim 1, wherein the target outline is the outline of the target object or the overlapping outlines of two or more target objects. The inventory management method as described in Claim 1, wherein the regression module further performs binarization on the target object according to the characteristic value. The inventory management method as described in claim 1 or 4, wherein the feature value is the color of the target object on the image. An inventory management system, comprising: a photographic device used to capture an image of an area, and there is a target object in the image; a data server connected to the photographic device and used to store the image; a computing server The server is connected with the data server, and the calculation server includes: an image recognition module, which is used to judge the target type and target outline of the target object according to the characteristic value of the target object, and eliminate the inability to judge the target object The target profile of the target category; a regression module is used to select a linear regression equation according to the target category, and then cut the target profile according to the number of variables in the linear regression equation to generate a plurality of partial profiles, and each of the The area of the partial outline is variable, and the target weight of the target object is calculated; a management server, connected with the calculation server, is used to receive the target weight, and provide an interface module to manage the inventory of the target object, wherein The stock of the target object is represented by the target weight; wherein, the linear regression equation is:

Weight : target weight; class : target category; k : number of variables; a _i : coefficient of the i -th variable; x _i : i -th variable; β : constant. The inventory management system as described in Claim 1, wherein the image recognition module includes a trained Region Based Convolutional Neural Networks (R-CNN) module. The inventory management system according to claim 1, wherein the target outline is the outline of the target object or the overlapping outlines of two or more target objects. The inventory management method as described in Claim 1, wherein the regression module performs binarization on the target object according to the characteristic value. The inventory management system as described in Claim 6 or 9, wherein the feature value is the color of the target object on the image.

Images