CN111598807B - Automobile part detection data sharing system and method based on block chain - Google Patents
Automobile part detection data sharing system and method based on block chain Download PDFInfo
- Publication number
- CN111598807B CN111598807B CN202010416005.8A CN202010416005A CN111598807B CN 111598807 B CN111598807 B CN 111598807B CN 202010416005 A CN202010416005 A CN 202010416005A CN 111598807 B CN111598807 B CN 111598807B
- Authority
- CN
- China
- Prior art keywords
- image
- automobile part
- automobile
- data
- parts
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 77
- 238000000034 method Methods 0.000 title claims abstract description 24
- 238000004364 calculation method Methods 0.000 claims description 73
- 238000013528 artificial neural network Methods 0.000 claims description 51
- 238000005457 optimization Methods 0.000 claims description 30
- 238000003709 image segmentation Methods 0.000 claims description 18
- 238000010586 diagram Methods 0.000 claims description 15
- 238000012549 training Methods 0.000 claims description 15
- 238000012545 processing Methods 0.000 claims description 9
- 230000006798 recombination Effects 0.000 claims description 8
- 238000005215 recombination Methods 0.000 claims description 8
- 238000007477 logistic regression Methods 0.000 claims description 4
- 238000004088 simulation Methods 0.000 claims description 4
- 238000004040 coloring Methods 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 238000007689 inspection Methods 0.000 claims 1
- 238000013523 data management Methods 0.000 abstract description 3
- 230000001419 dependent effect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/06—Protocols specially adapted for file transfer, e.g. file transfer protocol [FTP]
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1097—Protocols in which an application is distributed across nodes in the network for distributed storage of data in networks, e.g. transport arrangements for network file system [NFS], storage area networks [SAN] or network attached storage [NAS]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30164—Workpiece; Machine component
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L9/00—Cryptographic mechanisms or cryptographic arrangements for secret or secure communications; Network security protocols
- H04L9/50—Cryptographic mechanisms or cryptographic arrangements for secret or secure communications; Network security protocols using hash chains, e.g. blockchains or hash trees
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
A block chain-based automobile part detection data sharing system and method comprises the following steps: the detection unit is used for acquiring detection data of automobile parts and uploading the detection data of the automobile parts to the corresponding distributed server nodes, and the distributed server nodes are used for transmitting the received detection data of the automobile parts to other distributed server nodes on the block chain so as to store the detection data of the automobile parts acquired by the detection unit into all the distributed server nodes on the block chain, so that the detection data of the automobile parts are associated in one system, the additional cost of data management is reduced, and the sharing of the detection data of the automobile parts of all users on the server is realized.
Description
Technical Field
The invention relates to the field of automobile part production, in particular to an automobile part detection data sharing system and method based on a block chain.
Background
The automobile is an electromechanical hybrid complex system consisting of tens of thousands of parts, the automobile parts are used as the basis of the automobile industry and are necessary factors for supporting the continuous and healthy development of the automobile industry, and in order to meet high requirements on product quality and safety, the automobile parts need to be subjected to various detections.
However, there are a plurality of manufacturers of the same automobile parts, and the detection data of the automobile parts are stored in respective independent systems, and the technical architectures, communication protocols, data storage formats, and the like of the systems are different, which seriously affects the efficiency of interconnection.
Moreover, automobile parts produced by manufacturers are often unqualified, and personnel are arranged to detect the automobile parts by naked eyes, so that a large amount of manpower is needed, and the naked eyes can easily judge errors, so that the unqualified automobile parts flow into the market.
Disclosure of Invention
The invention aims to solve the problems found in the background art, and provides a block chain-based automobile part detection data sharing system and method.
In order to achieve the purpose, the invention adopts the following technical scheme: a block chain-based automobile part detection data sharing system comprises detection units and corresponding distributed server nodes
The detection unit is used for acquiring detection data of automobile parts and uploading the detection data of the automobile parts to the corresponding distributed server nodes;
the distributed server nodes are used for transmitting the received detection data of the automobile parts to other distributed server nodes on the block chain, so that the detection data of the automobile parts acquired by the detection unit are stored in all the distributed server nodes on the block chain.
Further, each automobile part possesses a unique digital signature, and the digital signature is used for signing the detection data of the automobile part.
Further, when the automobile part is conveyed to the detection unit, the original image of the automobile part is obtained, and whether the automobile part is qualified is judged according to the obtained original image of the automobile part, which is specifically as follows:
constructing an automobile part image denoising network model, wherein the automobile part image denoising network model consists of an automobile part image denoising neural network and an automobile part image attention network, training the automobile part image denoising network model, acquiring an automobile part clear image to form an automobile part training image data set, correcting the automobile part training image data set, fusing white noise into the automobile part clear image through white noise simulation reality noise to obtain a noisy automobile part image, inputting the noisy automobile part image into a deconvolution layer of the automobile part image denoising neural network, inputting the calculated data into a dense connection convolution network consisting of convolution layers, and inputting the calculated data of the dense connection convolution network into the deconvolution layer to obtain an automobile part noise layout image, eliminating the noise layout diagram of the automobile parts in the noisy automobile parts image to obtain an initial de-noised automobile parts image, inputting the initial de-noised automobile parts image into a deconvolution network of an automobile parts image attention network, inputting calculation data into a dense connection convolution network formed by convolution layers, fusing the calculation data of the dense connection convolution network with the calculation data of the deconvolution network to obtain fused calculation data, then inputting the fused calculation data into the deconvolution network again to obtain final calculation data, and then performing logistic regression calculation on the final calculation data to obtain an automobile parts noise ratio diagram; fusing the noise layout drawing of the automobile parts and the noise ratio drawing of the automobile parts to obtain a noise scatter image of the automobile parts, eliminating the noise scatter image of the automobile parts in the noise image of the automobile parts to obtain a secondary de-noising image of the automobile parts, the method comprises the steps of conducting mean square error processing on an initial de-noised automobile part image and a secondary de-noised automobile part image, learning the corresponding relation between the de-noised automobile part image and the secondary de-noised automobile part image for multiple times, correcting parameters of an automobile part image de-noising network model until the automobile part image de-noising network model is robust, obtaining the learned automobile part image de-noising network model, obtaining an original automobile part image, inputting the automobile part original image into the trained automobile part image de-noising network model, and obtaining an automobile part de-noising image matched with the automobile part original image.
Further, constructing an automobile part de-noising image segmentation model, wherein the automobile part de-noising image segmentation model consists of an automobile part input neural network and an automobile part output neural network, the automobile part input neural network comprises a reduced image layer and a parameter quantity optimization convolutional layer, the automobile part input neural network is used for acquiring multi-scale data of the automobile part de-noising image and reducing the resolution of the calculated characteristic image, the reduced image layer consists of two parts, one part of the two parts completes maximum pooling calculation, the other part of the two parts completes convolution calculation, the calculation data of the two parts are fused, the fused calculation data is combined with the data of the two parts through recombination calculation to obtain the calculation data of the reduced image layer, and the parameter quantity optimization convolutional layer adopts asymmetric kernel convolution to correct parameter quantity, firstly, the loading of a parameter quantity optimization convolutional layer is divided into two convolutional groups with the same channel quantity, one convolutional group is used for collecting loaded original characteristic data, the convolutional loading of the other convolutional group is respectively the superposition of the original loading and the loading of the previous group, all the convolutional groups are connected by adopting a cascade structure, the calculation data of all the convolutional groups are fused, the fused output data are recombined and calculated to fuse the data of all the convolutional groups to obtain a final calculation result, the automobile part output neural network is constructed by a parameter quantity optimization convolutional layer and a deconvolution layer, the automobile part output neural network is used for decoding the calculation result of the automobile part input neural network and amplifying and restoring the reduced output characteristic graph to the initial characteristic graph resolution, and the parameter quantity optimization convolutional layer adopts asymmetric kernel convolution to correct the parameter quantity, firstly, the loading of a parameter optimization convolution layer is divided into two convolution groups with the same channel quantity, one convolution group is used for collecting loaded original characteristic data, the convolution loading of the other convolution group is respectively the superposition of the original loading and the loading of the previous group, the convolution modes adopted by the convolution groups are all asymmetric kernel convolution, all the convolution groups are connected in a cascade mode, then the calculation results of all the convolution groups are fused, the fused calculation results are fused with the data of all the convolution groups through recombination calculation to obtain the final calculation result, the deconvolution layer applies a convolution kernel with a preset size and a convolution form with a preset step length to amplify and load the characteristic image resolution and reduce the channel quantity, an automobile part training image data set is adopted to train an automobile part de-noising image segmentation model to obtain the corresponding model weight, and the model weight is loaded to obtain a learned automobile part de-noising image segmentation model, after the dimensionless data is input into the learned automobile part denoising image segmentation model, obtaining an automobile part multi-scale characteristic image containing automobile part multi-scale characteristic data by inputting a neural network into the automobile part, decoding the automobile part multi-scale characteristic image into corresponding label budget data by the output neural network of the automobile part, amplifying the characteristic image resolution of the automobile part multi-scale characteristic image to the characteristic image resolution of the automobile part denoising image, and coloring the characteristic image calculated by the automobile part denoising image segmentation model according to the budgeted label to obtain an automobile part target area.
Further, the automobile part target area is rectangular, the position of the automobile part target area is selected by two position coordinates of the lower left corner of the automobile part target area and the upper right corner of the rectangular automobile part target area, the two position coordinates are a pair, coordinate groups of all pairs are recorded in an array, position data of all automobile part target areas are confirmed, characteristic points are measured, the automobile part de-noised image is used as a comparison image, a qualified automobile part image is selected from an image library as an automobile part sample image, the comparison image is converted into an automobile part gray-scale image, and the contrast of the automobile part gray-scale image is finely adjusted by utilizing an image histogram to obtain a corrected comparison image; the method comprises the steps of selecting an initial structure matching critical value according to an automobile part target area, setting the initial structure matching critical value as a set value, calculating the structure similarity of each automobile part target area of an automobile part sample image and a corrected comparison image, comparing the structure similarity of each automobile part target area with the structure matching critical value respectively, matching the current automobile part target area with the corresponding automobile part target area of the automobile part sample image if the structure similarity is larger than the structure matching critical value, processing comparison data of all the automobile part target areas, and if the automobile part sample image is matched with all the automobile part target areas of the corrected comparison image, the automobile part is qualified, otherwise, the automobile part is unqualified.
A block chain-based automobile part detection data sharing method comprises the following steps:
the detection data of the automobile parts are obtained through a detection unit, and the detection data of the automobile parts are uploaded to the corresponding distributed server nodes;
and transmitting the received detection data of the automobile parts to other distributed server nodes on the block chain through the distributed server nodes, so as to store the detection data of the automobile parts acquired by the detection unit into all the distributed server nodes on the block chain.
Compared with the prior art, the invention has the advantages that:
1. the detection data of the automobile parts are sent to the distributed server nodes through the detection unit, and then are stored in all the distributed server nodes on the block chain through the distributed server nodes, so that the detection data of the automobile parts are associated in one system, the additional cost of data management is reduced, and the sharing of the detection data of the automobile parts of all users on the server is realized.
2. The method combines the automobile part image denoising neural network with the automobile part image attention system structure, applies attention to automobile part image denoising, and improves the denoising effect of the automobile part image by endowing different proportions to the noise at different positions of the automobile part image so as to improve the subsequent processing quality of the automobile part image.
3. The parameter quantity of the parameter quantity optimization convolutional layer in the automobile part input neural network is corrected by using the grouping convolution and the asymmetric convolution, so that the parameter quantity of the parameter quantity optimization convolutional layer in the automobile part input neural network is greatly reduced, the operation speed of the parameter quantity optimization convolutional layer in the automobile part input neural network is improved, all convolution groups are connected in a cascade mode, the convolution view of the parameter quantity optimization convolutional layer in the automobile part input neural network is controlled by an expansion rate, the capability of acquiring the multi-scale characteristic data of the automobile part image by the parameter quantity optimization convolutional layer in the automobile part input neural network is given, the speed of acquiring the multi-scale characteristic of the automobile part image is improved by the grouping convolution mode, and the data exchange among channels is increased by the automobile part input neural network reduction image layer with recombination operation, the method is suitable for the grouping operation of the parameter quantity optimization convolution layer in the subsequent automobile part input neural network, reduces the loss of characteristic data and improves the accuracy rate of extracting the target area of the automobile part.
4. The automobile part de-noising image is converted into the automobile part gray-scale image, and the image histogram is used for finely adjusting the contrast of the automobile part gray-scale image, so that the detection of the feature points in the later process is very dependent on the gray value of the automobile part image, and the accuracy of identification can be interfered if the image is not converted.
Drawings
Fig. 1 is a schematic diagram of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example (b):
referring to fig. 1, the present embodiment provides a block chain-based automobile part detection data sharing system, including: a base detection unit and corresponding distributed server nodes;
the detection unit is used for acquiring detection data of automobile parts and uploading the detection data of the automobile parts to the corresponding distributed server nodes;
the distributed server nodes are used for transmitting the received detection data of the automobile parts to other distributed server nodes on the block chain so as to store the detection data of the automobile parts acquired by the detection unit into all the distributed server nodes on the block chain;
each automobile part is provided with a unique digital signature, and the digital signature is used for signing the detection data of the automobile part.
The detection data of the automobile parts are sent to the distributed server nodes through the detection unit, and then are stored in all the distributed server nodes on the block chain through the distributed server nodes, so that the detection data of the automobile parts are associated in one system, the additional cost of data management is reduced, and the sharing of the detection data of the automobile parts of all users on the server is realized.
When the automobile part is conveyed to the detection unit, the original image of the automobile part is obtained, and whether the automobile part is qualified or not is judged according to the obtained original image of the automobile part, which is specifically as follows:
firstly, constructing an automobile part image denoising network model, wherein the automobile part image denoising network model is composed of an automobile part image denoising neural network and an automobile part image attention network, training the automobile part image denoising network model, acquiring an automobile part clear image to construct an automobile part training image data set, correcting the automobile part training image data set, fusing white noise into the automobile part clear image through white noise simulation reality noise to obtain a noisy automobile part image, inputting the noisy automobile part image into a deconvolution layer of the automobile part image denoising neural network, inputting calculation data into a dense connection convolution network composed of convolution layers, and inputting the calculation data of the dense connection convolution network into the deconvolution layer to obtain an automobile part noise layout image, eliminating the noise layout diagram of the automobile parts in the noisy automobile parts image to obtain an initial de-noised automobile parts image, inputting the initial de-noised automobile parts image into a deconvolution network of an automobile parts image attention network, inputting calculation data into a dense connection convolution network formed by convolution layers, fusing the calculation data of the dense connection convolution network with the calculation data of the deconvolution network to obtain fused calculation data, then inputting the fused calculation data into the deconvolution network again to obtain final calculation data, and then performing logistic regression calculation on the final calculation data to obtain an automobile parts noise ratio diagram; fusing the noise layout drawing of the automobile parts and the noise ratio drawing of the automobile parts to obtain a noise scatter image of the automobile parts, eliminating the noise scatter image of the automobile parts in the noise image of the automobile parts to obtain a secondary de-noising image of the automobile parts, the method comprises the steps of conducting mean square error processing on an initial de-noised automobile part image and a secondary de-noised automobile part image, learning the corresponding relation between the de-noised automobile part image and the secondary de-noised automobile part image for multiple times, correcting parameters of an automobile part image de-noising network model until the automobile part image de-noising network model is robust, obtaining the learned automobile part image de-noising network model, obtaining an original automobile part image, inputting the automobile part original image into the trained automobile part image de-noising network model, and obtaining an automobile part de-noising image matched with the automobile part original image.
The method combines the automobile part image denoising neural network with the automobile part image attention system structure, applies attention to automobile part image denoising, and improves the denoising effect of the automobile part image by endowing different proportions to the noise at different positions of the automobile part image so as to improve the subsequent processing quality of the automobile part image.
Then, constructing an automobile part de-noising image segmentation model, wherein the automobile part de-noising image segmentation model consists of an automobile part input neural network and an automobile part output neural network, the automobile part input neural network comprises a reduced image layer and a parameter quantity optimization convolutional layer, the automobile part input neural network is used for acquiring multi-scale data of the automobile part de-noising image and reducing the resolution of the calculated characteristic image, the reduced image layer consists of two parts, one part of the two parts is used for completing maximum pooling calculation, the other part of the two parts is used for completing convolution calculation, the calculation data of the two parts are used for fusion, the fused calculation data is combined with the data of the two parts through recombination calculation to obtain the calculation data of the reduced image layer, and the parameter quantity optimization convolutional layer adopts asymmetric kernel convolution correction parameters, firstly, the loading of a parameter quantity optimization convolutional layer is divided into two convolutional groups with the same channel quantity, one convolutional group is used for collecting loaded original characteristic data, the convolutional loading of the other convolutional group is respectively the superposition of the original loading and the loading of the previous group, all the convolutional groups are connected by adopting a cascade structure, the calculation data of all the convolutional groups are fused, the fused output data are recombined and calculated to fuse the data of all the convolutional groups to obtain a final calculation result, the automobile part output neural network is constructed by a parameter quantity optimization convolutional layer and a deconvolution layer, the automobile part output neural network is used for decoding the calculation result of the automobile part input neural network and amplifying and restoring the reduced output characteristic graph to the initial characteristic graph resolution, and the parameter quantity optimization convolutional layer adopts asymmetric kernel convolution to correct the parameter quantity, firstly, the loading of a parameter optimization convolution layer is divided into two convolution groups with the same channel quantity, one convolution group is used for collecting loaded original characteristic data, the convolution loading of the other convolution group is respectively the superposition of the original loading and the loading of the previous group, the convolution modes adopted by the convolution groups are all asymmetric kernel convolution, all the convolution groups are connected in a cascade mode, then the calculation results of all the convolution groups are fused, the fused calculation results are fused with the data of all the convolution groups through recombination calculation to obtain the final calculation result, the deconvolution layer applies a convolution kernel with a preset size and a convolution form with a preset step length to amplify and load the characteristic image resolution and reduce the channel quantity, an automobile part training image data set is adopted to train an automobile part de-noising image segmentation model to obtain the corresponding model weight, and the model weight is loaded to obtain a learned automobile part de-noising image segmentation model, after the dimensionless data are input into the learned automobile part de-noising image segmentation model, obtaining an automobile part multi-scale characteristic image containing automobile part multi-scale characteristic data by inputting a neural network into the automobile part, decoding the automobile part multi-scale characteristic image into corresponding label budget data by the automobile part output neural network, amplifying the characteristic diagram resolution of the automobile part multi-scale characteristic image to the characteristic diagram resolution of the automobile part de-noising image, coloring the characteristic diagram calculated by the automobile part de-noising image segmentation model according to the budgeted label to obtain an automobile part target area, wherein the automobile part target area is rectangular, and the lower left corner of the automobile part target area and the upper right corner of the rectangular automobile part target area are located at two positions Selecting the position of an automobile part target area, wherein two position coordinates are a pair, recording all pairs of coordinate sets in an array, confirming the position data of all automobile part target areas, measuring and calculating characteristic points, taking an automobile part de-noised image as a comparison image, selecting a qualified automobile part image from an image library as an automobile part sample image, converting the comparison image into an automobile part gray image, and finely adjusting the contrast of the automobile part gray image by utilizing an image histogram to obtain a corrected comparison image; the method comprises the steps of selecting an initial structure matching critical value according to an automobile part target area, setting the initial structure matching critical value as a set value, calculating the structure similarity of each automobile part target area of an automobile part sample image and a corrected comparison image, comparing the structure similarity of each automobile part target area with the structure matching critical value respectively, matching the current automobile part target area with the corresponding automobile part target area of the automobile part sample image if the structure similarity is larger than the structure matching critical value, processing comparison data of all the automobile part target areas, and if the automobile part sample image is matched with all the automobile part target areas of the corrected comparison image, the automobile part is qualified, otherwise, the automobile part is unqualified.
The parameter quantity of the parameter quantity optimization convolutional layer in the automobile part input neural network is corrected by using the grouping convolution and the asymmetric convolution, so that the parameter quantity of the parameter quantity optimization convolutional layer in the automobile part input neural network is greatly reduced, the operation speed of the parameter quantity optimization convolutional layer in the automobile part input neural network is improved, all convolution groups are connected in a cascade mode, the convolution view of the parameter quantity optimization convolutional layer in the automobile part input neural network is controlled by an expansion rate, the capability of acquiring the multi-scale characteristic data of the automobile part image by the parameter quantity optimization convolutional layer in the automobile part input neural network is given, the speed of acquiring the multi-scale characteristic of the automobile part image is improved by the grouping convolution mode, and the data exchange among channels is increased by the automobile part input neural network reduction image layer with recombination operation, the method is suitable for the grouping operation of the parameter quantity optimization convolution layer in the subsequent automobile part input neural network, reduces the loss of characteristic data and improves the accuracy rate of extracting the target area of the automobile part. In addition, the automobile part de-noising image is converted into the automobile part gray-scale image, and the image histogram is used for finely adjusting the contrast of the automobile part gray-scale image, so that the detection of the feature points in the later process is very dependent on the gray-scale value of the automobile part image, and the accuracy of identification is interfered if the image is not converted.
The embodiment also provides a block chain-based automobile part detection data sharing method, which specifically comprises the following steps:
the detection data of the automobile parts are obtained through a detection unit, and the detection data of the automobile parts are uploaded to the corresponding distributed server nodes;
and transmitting the received detection data of the automobile parts to other distributed server nodes on the block chain through the distributed server nodes, so as to store the detection data of the automobile parts acquired by the detection unit into all the distributed server nodes on the block chain.
Claims (4)
1. A block chain-based automobile part detection data sharing system is characterized by comprising a detection unit and corresponding distributed server nodes,
the detection unit is used for acquiring detection data of automobile parts and uploading the detection data of the automobile parts to the corresponding distributed server nodes;
the distributed server nodes are used for transmitting the received detection data of the automobile parts to other distributed server nodes on the block chain so as to store the detection data of the automobile parts acquired by the detection unit into all the distributed server nodes on the block chain;
when the automobile parts are conveyed to the detection unit, original images of the automobile parts are obtained, and whether the automobile parts are qualified or not is judged according to the obtained original images of the automobile parts; the method for judging whether the automobile parts are qualified or not according to the acquired original images of the automobile parts specifically comprises the following steps:
constructing an automobile part image denoising network model, wherein the automobile part image denoising network model consists of an automobile part image denoising neural network and an automobile part image attention network, training the automobile part image denoising network model, acquiring an automobile part clear image to form an automobile part training image data set, correcting the automobile part training image data set, fusing white noise into the automobile part clear image through white noise simulation reality noise to obtain a noisy automobile part image, inputting the noisy automobile part image into a deconvolution layer of the automobile part image denoising neural network, inputting the calculated data into a dense connection convolution network consisting of convolution layers, and inputting the calculated data of the dense connection convolution network into the deconvolution layer to obtain an automobile part noise layout image, eliminating the noise layout diagram of the automobile parts in the noisy automobile parts image to obtain an initial de-noised automobile parts image, inputting the initial de-noised automobile parts image into a deconvolution network of an automobile parts image attention network, inputting calculation data into a dense connection convolution network formed by convolution layers, fusing the calculation data of the dense connection convolution network with the calculation data of the deconvolution network to obtain fused calculation data, then inputting the fused calculation data into the deconvolution network again to obtain final calculation data, and then performing logistic regression calculation on the final calculation data to obtain an automobile parts noise ratio diagram; fusing the noise layout drawing of the automobile parts and the noise ratio drawing of the automobile parts to obtain a noise scatter image of the automobile parts, eliminating the noise scatter image of the automobile parts in the noise image of the automobile parts to obtain a secondary de-noising image of the automobile parts, the method comprises the steps of conducting mean square error processing on an initial de-noised automobile part image and a secondary de-noised automobile part image, learning the corresponding relation between the de-noised automobile part image and the secondary de-noised automobile part image for multiple times, correcting parameters of an automobile part image de-noising network model until the automobile part image de-noising network model is robust, obtaining the learned automobile part image de-noising network model, obtaining an original automobile part image, inputting the automobile part original image into the trained automobile part image de-noising network model, and obtaining an automobile part de-noising image matched with the automobile part original image.
2. The system according to claim 1, wherein each of the vehicle components has a unique digital signature, and the digital signature is used to sign the detection data of the vehicle component.
3. The system for sharing data of parts inspection of automobile based on block chain as claimed in claim 1, further comprising the steps of:
constructing an automobile part de-noising image segmentation model, wherein the automobile part de-noising image segmentation model consists of an automobile part input neural network and an automobile part output neural network, the automobile part input neural network comprises a reduced image layer and a parameter quantity optimization convolutional layer, the automobile part input neural network is used for acquiring multi-scale data of the automobile part de-noising image and reducing the resolution of a calculated characteristic image, the reduced image layer consists of two parts, one part of the reduced image layer is used for completing maximum pooling calculation, the other part of the reduced image layer is used for completing convolution calculation, the calculation data of the two parts are fused, the fused calculation data is combined with the data of the two parts through recombination calculation to obtain the calculation data of the reduced image layer, and the parameter quantity optimization convolutional layer adopts asymmetric kernel convolution to correct the parameter quantity, firstly, the loading of a parameter quantity optimization convolutional layer is divided into two convolutional groups with the same channel quantity, one convolutional group is used for collecting loaded original characteristic data, the convolutional loading of the other convolutional group is respectively the superposition of the original loading and the loading of the previous group, all the convolutional groups are connected by adopting a cascade structure, the calculation data of all the convolutional groups are fused, the fused output data are recombined and calculated to fuse the data of all the convolutional groups to obtain a final calculation result, the automobile part output neural network is constructed by a parameter quantity optimization convolutional layer and a deconvolution layer, the automobile part output neural network is used for decoding the calculation result of the automobile part input neural network and amplifying and restoring the reduced output characteristic graph to the initial characteristic graph resolution, and the parameter quantity optimization convolutional layer adopts asymmetric kernel convolution to correct the parameter quantity, firstly, the loading of a parameter optimization convolution layer is divided into two convolution groups with the same channel quantity, one convolution group is used for collecting loaded original characteristic data, the convolution loading of the other convolution group is respectively the superposition of the original loading and the loading of the previous group, the convolution modes adopted by the convolution groups are all asymmetric kernel convolution, all the convolution groups are connected in a cascade mode, then the calculation results of all the convolution groups are fused, the fused calculation results are fused with the data of all the convolution groups through recombination calculation to obtain the final calculation result, the deconvolution layer applies a convolution kernel with a preset size and a convolution form with a preset step length to amplify and load the characteristic image resolution and reduce the channel quantity, an automobile part training image data set is adopted to train an automobile part de-noising image segmentation model to obtain the corresponding model weight, and the model weight is loaded to obtain a learned automobile part de-noising image segmentation model, after the dimensionless data are input into the learned automobile part de-noising image segmentation model, obtaining an automobile part multi-scale characteristic image containing automobile part multi-scale characteristic data by inputting a neural network into the automobile part, decoding the automobile part multi-scale characteristic image into corresponding label budget data by the automobile part output neural network, amplifying the characteristic diagram resolution of the automobile part multi-scale characteristic image to the characteristic diagram resolution of the automobile part de-noising image, coloring the characteristic diagram calculated by the automobile part de-noising image segmentation model according to the budgeted label to obtain an automobile part target area, wherein the automobile part target area is rectangular, and the lower left corner of the automobile part target area and the upper right corner of the rectangular automobile part target area are located at two positions Selecting the position of an automobile part target area, wherein two position coordinates are a pair, recording all pairs of coordinate sets in an array, confirming the position data of all automobile part target areas, measuring and calculating characteristic points, taking an automobile part de-noised image as a comparison image, selecting a qualified automobile part image from an image library as an automobile part sample image, converting the comparison image into an automobile part gray image, and finely adjusting the contrast of the automobile part gray image by utilizing an image histogram to obtain a corrected comparison image; the method comprises the steps of selecting an initial structure matching critical value according to an automobile part target area, setting the initial structure matching critical value as a set value, calculating the structure similarity of each automobile part target area of an automobile part sample image and a corrected comparison image, comparing the structure similarity of each automobile part target area with the structure matching critical value respectively, matching the current automobile part target area with the corresponding automobile part target area of the automobile part sample image if the structure similarity is larger than the structure matching critical value, processing comparison data of all the automobile part target areas, and if the automobile part sample image is matched with all the automobile part target areas of the corrected comparison image, the automobile part is qualified, otherwise, the automobile part is unqualified.
4. A block chain-based automobile part detection data sharing method is characterized by comprising the following steps:
the method comprises the steps that detection data of automobile parts are obtained through a detection unit, and the detection data of the automobile parts are uploaded to corresponding distributed server nodes;
transmitting the received detection data of the automobile parts to other distributed server nodes on the block chain through the distributed server nodes, so as to store the detection data of the automobile parts acquired by the detection unit into all the distributed server nodes on the block chain;
when the automobile parts are conveyed to the detection unit, original images of the automobile parts are obtained, and whether the automobile parts are qualified or not is judged according to the obtained original images of the automobile parts; the method for judging whether the automobile parts are qualified or not according to the acquired original images of the automobile parts specifically comprises the following steps:
constructing an automobile part image denoising network model, wherein the automobile part image denoising network model consists of an automobile part image denoising neural network and an automobile part image attention network, training the automobile part image denoising network model, acquiring an automobile part clear image to form an automobile part training image data set, correcting the automobile part training image data set, fusing white noise into the automobile part clear image through white noise simulation reality noise to obtain a noisy automobile part image, inputting the noisy automobile part image into a deconvolution layer of the automobile part image denoising neural network, inputting the calculated data into a dense connection convolution network consisting of convolution layers, and inputting the calculated data of the dense connection convolution network into the deconvolution layer to obtain an automobile part noise layout image, eliminating the noise layout diagram of the automobile parts in the noisy automobile parts image to obtain an initial de-noised automobile parts image, inputting the initial de-noised automobile parts image into a deconvolution network of an automobile parts image attention network, inputting calculation data into a dense connection convolution network formed by convolution layers, fusing the calculation data of the dense connection convolution network with the calculation data of the deconvolution network to obtain fused calculation data, then inputting the fused calculation data into the deconvolution network again to obtain final calculation data, and then performing logistic regression calculation on the final calculation data to obtain an automobile parts noise ratio diagram; fusing the noise layout drawing of the automobile parts and the noise ratio drawing of the automobile parts to obtain a noise scatter image of the automobile parts, eliminating the noise scatter image of the automobile parts in the noise image of the automobile parts to obtain a secondary de-noising image of the automobile parts, the method comprises the steps of conducting mean square error processing on an initial de-noised automobile part image and a secondary de-noised automobile part image, learning the corresponding relation between the de-noised automobile part image and the secondary de-noised automobile part image for multiple times, correcting parameters of an automobile part image de-noising network model until the automobile part image de-noising network model is robust, obtaining the learned automobile part image de-noising network model, obtaining an original automobile part image, inputting the automobile part original image into the trained automobile part image de-noising network model, and obtaining an automobile part de-noising image matched with the automobile part original image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010416005.8A CN111598807B (en) | 2020-05-16 | 2020-05-16 | Automobile part detection data sharing system and method based on block chain |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010416005.8A CN111598807B (en) | 2020-05-16 | 2020-05-16 | Automobile part detection data sharing system and method based on block chain |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111598807A CN111598807A (en) | 2020-08-28 |
| CN111598807B true CN111598807B (en) | 2021-06-15 |
Family
ID=72182520
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010416005.8A Active CN111598807B (en) | 2020-05-16 | 2020-05-16 | Automobile part detection data sharing system and method based on block chain |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111598807B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112214549B (en) * | 2020-10-10 | 2024-06-04 | 中育数据(广州)科技有限公司 | File feature code generation method and device and electronic equipment |
| CN113012153A (en) * | 2021-04-30 | 2021-06-22 | 武汉纺织大学 | Aluminum profile flaw detection method |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108197947A (en) * | 2017-12-27 | 2018-06-22 | 深圳市轱辘车联数据技术有限公司 | A kind of vehicle accessory data processing method and device for block chain |
| WO2019133922A1 (en) * | 2017-12-29 | 2019-07-04 | Flir Systems, Inc. | Point cloud denoising systems and methods |
| CN109615589A (en) * | 2018-10-31 | 2019-04-12 | 北京达佳互联信息技术有限公司 | Remove the method, apparatus and terminal device of picture noise |
| CN110108714A (en) * | 2019-04-28 | 2019-08-09 | 浙江博拉自动化科技有限公司 | A kind of auto parts and components appearance detection system and device |
| CN110428255B (en) * | 2019-07-05 | 2023-03-10 | 深圳壹账通智能科技有限公司 | Vehicle data management method and device based on block chain |
| CN110490817A (en) * | 2019-07-22 | 2019-11-22 | 武汉大学 | A kind of image noise suppression method based on mask study |
| CN110597785B (en) * | 2019-08-27 | 2022-11-11 | 武汉科技大学 | Automobile part remanufacturing sharing system based on block chain and private cloud |
| CN110766632A (en) * | 2019-10-22 | 2020-02-07 | 广东启迪图卫科技股份有限公司 | Image denoising method based on channel attention mechanism and characteristic pyramid |
-
2020
- 2020-05-16 CN CN202010416005.8A patent/CN111598807B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN111598807A (en) | 2020-08-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109584248B (en) | Infrared target instance segmentation method based on feature fusion and dense connection network | |
| CN112561910B (en) | Industrial surface defect detection method based on multi-scale feature fusion | |
| CN111699499B (en) | Inspection, image recognition, recognizer generation system, and learning data generation device | |
| CN109829891B (en) | Magnetic shoe surface defect detection method based on dense generation of antagonistic neural network | |
| Guan et al. | A steel surface defect recognition algorithm based on improved deep learning network model using feature visualization and quality evaluation | |
| CN110570396A (en) | industrial product defect detection method based on deep learning | |
| CN111832484A (en) | Loop detection method based on convolution perception hash algorithm | |
| CN110969606B (en) | Texture surface defect detection method and system | |
| CN111598807B (en) | Automobile part detection data sharing system and method based on block chain | |
| US12125193B2 (en) | Method for detecting defect in products and electronic device using method | |
| CN107341508B (en) | Fast food picture identification method and system | |
| CN113822284A (en) | RGBD image semantic segmentation method based on boundary attention | |
| CN111079539A (en) | Video abnormal behavior detection method based on abnormal tracking | |
| CN111325224A (en) | Computer-readable storage medium, input data checking method, and computing device | |
| CN113780423A (en) | Single-stage target detection neural network based on multi-scale fusion and industrial product surface defect detection model | |
| CN113111875A (en) | Seamless steel rail weld defect identification device and method based on deep learning | |
| CN113763364A (en) | Image defect detection method based on convolutional neural network | |
| CN117911350A (en) | PCB surface defect detection method based on improvement YOLOv-tiny | |
| CN110751191A (en) | Image classification method and system | |
| CN108615401B (en) | Deep learning-based indoor non-uniform light parking space condition identification method | |
| CN112364974A (en) | Improved YOLOv3 algorithm based on activation function | |
| CN115908142A (en) | Contact net tiny part damage testing method based on visual recognition | |
| JP7059889B2 (en) | Learning device, image generator, learning method, and learning program | |
| CN115330729A (en) | Multi-scale feature attention-fused light-weight strip steel surface defect detection method | |
| CN113591997B (en) | Assembly feature graph connection relation classification method based on graph learning convolutional neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20210525 Address after: Room 356, building A3, No.3, Qingxu street, Tianhe District, Guangzhou, Guangdong 510630 (office only) Applicant after: Guangzhou shuhuitong Information Technology Co.,Ltd. Address before: 430000 room 5-2-2-b-1, European Garden community, No.7 Xudong Road, Hongshan District, Wuhan City, Hubei Province Applicant before: Wuhan bronze Technology Information Consulting Co., Ltd |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |