Visual analysis method for electronic cigarette integrity

Abstract

The present invention relates to a visual analysis method for the integrity of electronic cigarettes, including: using a piece-by-piece transmission mechanism to transmit the electronic cigarette products produced by the electronic cigarette production mechanism piece by piece to the inspection station with the front end facing upward; using a gradient application mechanism to determine the number of each pixel point in the final optimization image, the gradient of each pixel value corresponding to each pixel point, the number of each pixel point in the standard pattern corresponding to the front end of the standard electronic cigarette, and the gradient of each pixel value corresponding to each pixel point, and then input into a recurrent neural network model to intelligently determine whether the front of the electronic cigarette product has appearance damage. The present invention can use a targeted visual analysis mechanism to perform a visual analysis of the integrity of the front end, i.e., the front of the electronic cigarette product, thereby providing reliable reference data for the subsequent elimination of electronic cigarette products with damaged appearance.

Description

Technical Field

[0001] The present invention relates to the field of electronic cigarette detection, and in particular to a visual analysis method for electronic cigarette integrity. Background Art

[0002] Electronic cigarettes are mainly composed of three parts: the cartridge (including the oil and the mouthpiece), the atomizer, and the cigarette rod. The cigarette rod of the electronic cigarette can be divided into two parts, the switch component and the battery component. Most electronic cigarettes use lithium-ion and secondary battery power components. Battery life depends on the type and size of the battery, the number of times it is used, and the operating environment. There are many different types of battery chargers to choose from, such as socket direct charging, car chargers, and USB interface chargers. The battery is the largest component of the electronic cigarette.

[0003] When electronic cigarette production organizations produce electronic cigarette products one by one, they need to conduct comprehensive status inspections on the appearance and performance of electronic cigarettes. In terms of appearance, the integrity of each electronic cigarette product needs to be analyzed. For example, the integrity of the front end of each electronic cigarette product needs to be analyzed. When electronic cigarette products with damaged appearance are found, they will be eliminated, thereby preventing inferior electronic cigarettes from entering the market. Summary of the invention

[0004] In order to solve the technical problems in the prior art, the present invention provides a visual analysis method for the integrity of electronic cigarettes, which can use a targeted visual analysis mechanism to perform a visual analysis of the integrity of the front end, i.e., the front side, of each newly produced electronic cigarette product when the electronic cigarette production organization produces electronic cigarette products one by one, thereby providing key information for the subsequent targeted quality inspection and processing of electronic cigarettes for discovering electronic cigarette products with damaged appearance and immediately eliminating the electronic cigarette products with damaged appearance.

[0005] According to one aspect of the present invention, a method for visually analyzing the integrity of an electronic cigarette is provided, the method comprising:

[0006] A piece-by-piece transmission mechanism is used to transmit the electronic cigarette products produced by the electronic cigarette production mechanism piece by piece to the inspection station with the front end facing upward, so as to keep the inspection station always placed with the latest transmitted electronic cigarette products;

[0007] A top-down capture mechanism is used, located directly above the detection station, to perform a top-down capture action on the electronic cigarette product that has been recently transmitted to the detection station in a top-down mode, so as to obtain and output a corresponding real-time capture picture;

[0008] Using a gamma correction device, connected to the overhead capture mechanism, for performing gamma correction processing on the received instant capture picture to obtain and output a corresponding primary optimized image;

[0009] Using a combined filtering device, connected to the gamma correction device, for performing combined filtering processing on the received primary optimized image to obtain and output a corresponding secondary optimized image;

[0010] Using an information enhancement device, connected to the combined filtering device, to perform image frequency domain enhancement processing on the received secondary optimized image to obtain and output a corresponding final optimized image;

[0011] Using a gradient application mechanism, connected to the information enhancement device, for receiving the final-level optimized image, determining the number of each pixel point in the final-level optimized image and the pixel value gradients corresponding to each pixel point, and determining the number of each pixel point in the standard pattern corresponding to the standard electronic cigarette front end and the pixel value gradients corresponding to each pixel point, the number of each pixel point in the final-level optimized image and the pixel value gradients corresponding to each pixel point and the number of each pixel point in the standard pattern and the pixel value gradients corresponding to each pixel point are synchronously input into the recurrent neural network model to obtain a digital mark of whether the standard pattern output by the recurrent neural network model exists in the final-level optimized image;

[0012] The number of each pixel point in the final optimization image and each pixel value gradient corresponding to each pixel point and the number of each pixel point in the standard pattern and each pixel value gradient corresponding to each pixel point are synchronously input into the recurrent neural network model to obtain a digital sign of whether the standard pattern output by the recurrent neural network model exists in the final optimization image, including: the recurrent neural network model is a recurrent neural network that has been trained multiple times, and the total number of training times is inversely correlated with the content repetition of the standard pattern corresponding to the front end of the standard electronic cigarette;

[0013] Among them, the recurrent neural network model is a recurrent neural network that has been trained multiple times, and the total number of training times is inversely correlated with the content repetition of the standard pattern corresponding to the standard electronic cigarette front end, including: using a numerical conversion function to represent the numerical conversion relationship in which the total number of training times is inversely correlated with the content repetition of the standard pattern corresponding to the standard electronic cigarette front end, and the more pixels with equal pixel values ​​in the standard pattern corresponding to the standard electronic cigarette front end, the higher the content repetition of the standard pattern corresponding to the standard electronic cigarette front end, and the fewer the total number of training times.

[0014] Therefore, the present invention has the following three key inventive concepts:

[0015] Inventive concept 1: Determine the number of each pixel point in the final optimization image and the pixel value gradients corresponding to each pixel point, and determine the number of each pixel point in the standard pattern corresponding to the standard electronic cigarette front end and the pixel value gradients corresponding to each pixel point, so as to provide key information for the subsequent intelligent judgment of whether the standard pattern exists in the final optimization image;

[0016] Invention concept 2: The number of each pixel point in the final optimization image and the pixel value gradients corresponding to each pixel point and the number of each pixel point in the standard pattern and the pixel value gradients corresponding to each pixel point are synchronously input into the recurrent neural network model to obtain a digital mark of whether the standard pattern output by the recurrent neural network model exists in the final optimization image;

[0017] Inventive concept three: The recurrent neural network model adopted is a recurrent neural network that has been trained multiple times, and the total number of training times is inversely correlated with the content repetition of the standard pattern corresponding to the standard electronic cigarette front end. Specifically, a numerical conversion function is used to represent the numerical conversion relationship in which the total number of training times is inversely correlated with the content repetition of the standard pattern corresponding to the standard electronic cigarette front end. The more pixels with equal pixel values ​​in the standard pattern corresponding to the standard electronic cigarette front end, the higher the content repetition of the standard pattern corresponding to the standard electronic cigarette front end, and the fewer the total number of training times, thereby achieving targeted design of recurrent neural network models for different types of electronic cigarettes. DETAILED DESCRIPTION

[0018] The implementation scheme of the electronic cigarette integrity visual analysis method of the present invention will be described in detail below.

[0019] Example 1

[0020] The following specific example is a structural block diagram of an electronic cigarette integrity visual analysis system according to an embodiment of the present invention, and the method includes:

[0021] Step S1: using a piece-by-piece transmission mechanism to transmit the electronic cigarette products produced by the electronic cigarette production mechanism piece by piece to the inspection station with the front end facing upward, so as to keep the inspection station always holding the latest transmitted electronic cigarette products;

[0022] Step S2: using a top-down capture mechanism, located directly above the detection station, to perform a top-down capture action on the electronic cigarette product that has been recently transmitted to the detection station in a top-down mode, so as to obtain and output a corresponding real-time capture picture;

[0023] For example, the overhead capture mechanism has a built-in filter, an overhead imaging lens, and a photoelectric sensor, which are used to cooperate in performing an overhead shooting action on the electronic cigarette product that has been recently transmitted to the detection station in an overhead shooting mode, so as to obtain and output a corresponding real-time captured picture;

[0024] Step S3: using a gamma correction device, connected to the overhead capture mechanism, to perform gamma correction processing on the received instant capture picture to obtain and output a corresponding primary optimized image;

[0025] Step S4: using a combined filtering device, connected to the gamma correction device, to perform combined filtering processing on the received primary optimized image to obtain and output a corresponding secondary optimized image;

[0026] Step S5: using an information enhancement device, connected to the combined filtering device, to perform image frequency domain enhancement processing on the received secondary optimized image to obtain and output a corresponding final optimized image;

[0027] Step S6: using a gradient application mechanism, connected to the information enhancement device, to receive the final optimized image, determine the number of each pixel point in the final optimized image and the pixel value gradients corresponding to each pixel point, and determine the number of each pixel point in the standard pattern corresponding to the standard electronic cigarette front end and the pixel value gradients corresponding to each pixel point, and synchronously input the number of each pixel point in the final optimized image and the pixel value gradients corresponding to each pixel point and the number of each pixel point in the standard pattern and the pixel value gradients corresponding to each pixel point into the recurrent neural network model to obtain a digital mark of whether the standard pattern output by the recurrent neural network model exists in the final optimized image;

[0028] Step S1, step S2, step S3, step S4, step S5, and step S6 are connected in sequence;

[0029] The number of each pixel point in the final optimization image and each pixel value gradient corresponding to each pixel point and the number of each pixel point in the standard pattern and each pixel value gradient corresponding to each pixel point are synchronously input into the recurrent neural network model to obtain a digital sign of whether the standard pattern output by the recurrent neural network model exists in the final optimization image, including: the recurrent neural network model is a recurrent neural network that has been trained multiple times, and the total number of training times is inversely correlated with the content repetition of the standard pattern corresponding to the front end of the standard electronic cigarette;

[0030] The recurrent neural network model is a recurrent neural network that has been trained multiple times, and the total number of training times is inversely correlated with the content repetition of the standard pattern corresponding to the standard electronic cigarette front end, including: a numerical conversion function is used to represent a numerical conversion relationship in which the total number of training times is inversely correlated with the content repetition of the standard pattern corresponding to the standard electronic cigarette front end, and the more pixels with equal pixel values ​​in the standard pattern corresponding to the standard electronic cigarette front end, the higher the content repetition of the standard pattern corresponding to the standard electronic cigarette front end, and the fewer the total number of training times;

[0031] Among them, the number of each pixel point in the final-level optimized image and the pixel value gradients corresponding to each pixel point and the number of each pixel point in the standard pattern and the pixel value gradients corresponding to each pixel point are synchronously input into the recurrent neural network model to obtain a digital mark of whether the standard pattern output by the recurrent neural network model exists in the final-level optimized image, which also includes: when the digital mark of whether the standard pattern output by the recurrent neural network model exists in the final-level optimized image is 0B01, the standard pattern does not exist in the final-level optimized image.

[0032] Example 2

[0033] Example 2 further illustrates the specific steps of the electronic cigarette integrity visual analysis method of the present invention.

[0034] The electronic cigarette integrity visual analysis method may further include:

[0035] Using a data monitoring device, including a plurality of monitoring cameras facing the gamma correction device, the combined filtering device, the information enhancement device and the gradient application mechanism respectively, for time-sharing monitoring of the real-time status of the gamma correction device, the combined filtering device, the information enhancement device and the gradient application mechanism;

[0036] Among them, the data monitoring device includes multiple monitoring cameras facing the gamma correction device, the combined filtering device, the information enhancement device and the gradient application mechanism respectively, and is used for time-sharing monitoring of the real-time status of the gamma correction device, the combined filtering device, the information enhancement device and the gradient application mechanism, including: the gamma correction device, the combined filtering device, the information enhancement device and the gradient application mechanism are respectively connected to the data monitoring device using different CSI interfaces.

[0037] In the electronic cigarette integrity visual analysis method:

[0038] The data monitoring device includes a plurality of monitoring cameras respectively facing the gamma correction device, the combined filtering device, the information enhancement device and the gradient application mechanism, and is used for time-sharing monitoring of the real-time status of the gamma correction device, the combined filtering device, the information enhancement device and the gradient application mechanism. The data monitoring device also includes: the plurality of monitoring cameras respectively facing the gamma correction device, the combined filtering device, the information enhancement device and the gradient application mechanism have the same structure.

[0039] And in the electronic cigarette integrity visual analysis method:

[0040] A data monitoring device, comprising a plurality of monitoring cameras facing the gamma correction device, the combined filtering device, the information enhancement device and the gradient application mechanism, respectively, for time-sharing monitoring of the real-time status of the gamma correction device, the combined filtering device, the information enhancement device and the gradient application mechanism, further comprising: a plurality of monitoring cameras with built-in image sensing units facing the gamma correction device, the combined filtering device, the information enhancement device and the gradient application mechanism, respectively;

[0041] Among them, a plurality of surveillance cameras respectively facing the gamma correction device, the combined filtering device, the information enhancement device and the gradient application mechanism have built-in image sensing units: the image sensing units are CMOS sensing units or CCD sensing units.

[0042] In addition, in the electronic cigarette integrity visual analysis system and method, the number of each pixel point in the final optimization image and the pixel value gradients corresponding to each pixel point and the number of each pixel point in the standard pattern and the pixel value gradients corresponding to each pixel point are synchronously input into the recurrent neural network model to obtain a digital mark of whether the standard pattern output by the recurrent neural network model exists in the final optimization image, and also includes: when the digital mark of whether the standard pattern output by the recurrent neural network model exists in the final optimization image is 0B00, the standard pattern exists in the final optimization image.

[0043] The electronic cigarette integrity visual analysis method of the present invention is used to address the technical problem in the prior art that electronic cigarette products lack a mechanism for analyzing the appearance integrity of each product. By adopting a targeted visual analysis mechanism, a visual analysis of the integrity of the front end, i.e., the front side, of the electronic cigarette products is performed, thereby achieving a step-by-step visual analysis of the integrity of the electronic cigarette products and providing reliable reference data for the subsequent elimination of electronic cigarette products with damaged appearance.

[0044] The present invention has been described in great detail above, so that after reading and understanding this specification, various changes and modifications of the present invention will become obvious to those skilled in the art. Therefore, all such changes and modifications are also included in this invention, and therefore they are within the scope of protection of the claims.

Claims

1. A visual analysis method for electronic cigarette integrity, characterized in that: The method comprises: A piece-by-piece transmission mechanism is used to transmit the electronic cigarette products produced by the electronic cigarette production mechanism piece by piece to the inspection station with the front end facing upward, so as to keep the inspection station always placed with the latest transmitted electronic cigarette products; A top-down capture mechanism is used, located directly above the detection station, to perform a top-down capture action on the electronic cigarette product that has been recently transmitted to the detection station in a top-down mode, so as to obtain and output a corresponding real-time capture picture; Using a gamma correction device, connected to the overhead capture mechanism, for performing gamma correction processing on the received instant capture picture to obtain and output a corresponding primary optimized image; Using a combined filtering device, connected to the gamma correction device, for performing combined filtering processing on the received primary optimized image to obtain and output a corresponding secondary optimized image; Using an information enhancement device, connected to the combined filtering device, to perform image frequency domain enhancement processing on the received secondary optimized image to obtain and output a corresponding final optimized image; Using a gradient application mechanism, connected to the information enhancement device, for receiving the final-level optimized image, determining the number of each pixel point in the final-level optimized image and the pixel value gradients corresponding to each pixel point, and determining the number of each pixel point in the standard pattern corresponding to the standard electronic cigarette front end and the pixel value gradients corresponding to each pixel point, the number of each pixel point in the final-level optimized image and the pixel value gradients corresponding to each pixel point and the number of each pixel point in the standard pattern and the pixel value gradients corresponding to each pixel point are synchronously input into the recurrent neural network model to obtain a digital mark of whether the standard pattern output by the recurrent neural network model exists in the final-level optimized image; The number of each pixel point in the final optimization image and each pixel value gradient corresponding to each pixel point and the number of each pixel point in the standard pattern and each pixel value gradient corresponding to each pixel point are synchronously input into the recurrent neural network model to obtain a digital sign of whether the standard pattern output by the recurrent neural network model exists in the final optimization image, including: the recurrent neural network model is a recurrent neural network that has been trained multiple times, and the total number of training times is inversely correlated with the content repetition of the standard pattern corresponding to the front end of the standard electronic cigarette; The recurrent neural network model is a recurrent neural network that has been trained multiple times, and the total number of training times is inversely correlated with the content repetition of the standard pattern corresponding to the standard electronic cigarette front end, including: a numerical conversion function is used to represent a numerical conversion relationship in which the total number of training times is inversely correlated with the content repetition of the standard pattern corresponding to the standard electronic cigarette front end, and the more pixels with equal pixel values ​​in the standard pattern corresponding to the standard electronic cigarette front end, the higher the content repetition of the standard pattern corresponding to the standard electronic cigarette front end, and the fewer the total number of training times; Using a data monitoring device, including a plurality of monitoring cameras facing the gamma correction device, the combined filtering device, the information enhancement device and the gradient application mechanism respectively, for time-sharing monitoring of the real-time status of the gamma correction device, the combined filtering device, the information enhancement device and the gradient application mechanism; The gamma correction device, the combined filtering device, the information enhancement device and the gradient application mechanism are respectively connected to the data monitoring device using different CSI interfaces.

2. The electronic cigarette integrity visual analysis method according to claim 1, characterized in that: The data monitoring device includes a plurality of monitoring cameras respectively facing the gamma correction device, the combined filtering device, the information enhancement device and the gradient application mechanism, and is used for time-sharing monitoring of the real-time status of the gamma correction device, the combined filtering device, the information enhancement device and the gradient application mechanism. The data monitoring device also includes: the plurality of monitoring cameras respectively facing the gamma correction device, the combined filtering device, the information enhancement device and the gradient application mechanism have the same structure.

3. The electronic cigarette integrity visual analysis method according to claim 2, characterized in that: A data monitoring device, comprising a plurality of monitoring cameras facing the gamma correction device, the combined filtering device, the information enhancement device and the gradient application mechanism, respectively, for time-sharing monitoring of the real-time status of the gamma correction device, the combined filtering device, the information enhancement device and the gradient application mechanism, further comprising: a plurality of monitoring cameras with built-in image sensing units facing the gamma correction device, the combined filtering device, the information enhancement device and the gradient application mechanism, respectively; Among them, a plurality of surveillance cameras respectively facing the gamma correction device, the combined filtering device, the information enhancement device and the gradient application mechanism have built-in image sensing units: the image sensing units are CMOS sensing units or CCD sensing units.