CN114723724A - Dangerous goods identification method, device, equipment and storage medium based on artificial intelligence - Google Patents
Dangerous goods identification method, device, equipment and storage medium based on artificial intelligence Download PDFInfo
- Publication number
- CN114723724A CN114723724A CN202210412390.8A CN202210412390A CN114723724A CN 114723724 A CN114723724 A CN 114723724A CN 202210412390 A CN202210412390 A CN 202210412390A CN 114723724 A CN114723724 A CN 114723724A
- Authority
- CN
- China
- Prior art keywords
- image
- dangerous goods
- liquid
- target
- liquid dangerous
- 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.)
- Withdrawn
Links
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 52
- 238000000034 method Methods 0.000 title claims abstract description 43
- 239000007788 liquid Substances 0.000 claims abstract description 259
- 238000001514 detection method Methods 0.000 claims abstract description 77
- 238000007689 inspection Methods 0.000 claims abstract description 70
- 238000000605 extraction Methods 0.000 claims abstract description 52
- 238000003709 image segmentation Methods 0.000 claims abstract description 16
- 230000011218 segmentation Effects 0.000 claims description 35
- 238000012545 processing Methods 0.000 claims description 33
- 238000011176 pooling Methods 0.000 claims description 32
- 231100001261 hazardous Toxicity 0.000 claims description 28
- 239000013598 vector Substances 0.000 claims description 22
- 230000004913 activation Effects 0.000 claims description 10
- 238000005070 sampling Methods 0.000 claims description 10
- 230000007246 mechanism Effects 0.000 claims description 9
- 230000004927 fusion Effects 0.000 claims description 8
- 230000000007 visual effect Effects 0.000 claims description 8
- 239000000383 hazardous chemical Substances 0.000 claims description 6
- 238000002372 labelling Methods 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 4
- 238000013136 deep learning model Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 239000013056 hazardous product Substances 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 238000012549 training Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 238000005096 rolling process Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000002085 persistent effect Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
Images
Classifications
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
- G06Q50/265—Personal security, identity or safety
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20112—Image segmentation details
-
- 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
- G06T2207/30112—Baggage; Luggage; Suitcase
Abstract
The invention relates to the field of artificial intelligence, and discloses a dangerous goods identification method, a dangerous goods identification device, dangerous goods identification equipment and a storage medium based on artificial intelligence, which are used for improving the accuracy of liquid dangerous goods identification. The dangerous goods identification method based on artificial intelligence comprises the following steps: acquiring an initial security inspection image to be detected, and calling an ultrahigh resolution model to preprocess the initial security inspection image to obtain a target high-definition image; performing image segmentation on the target high-definition image to obtain a plurality of image subregions, and inputting the target high-definition image into a feature extraction network to perform image feature extraction to obtain image features; inputting the image sub-regions and the image characteristics into a liquid dangerous article identification model to carry out liquid dangerous article identification to obtain a target liquid dangerous article and a liquid dangerous article category; generating regional position information according to the target liquid dangerous goods; and generating a dangerous goods detection result according to the region position information and the liquid dangerous goods category.
Description
Technical Field
The invention relates to the field of artificial intelligence, in particular to a dangerous goods identification method, a dangerous goods identification device, dangerous goods identification equipment and a dangerous goods identification storage medium based on artificial intelligence.
Background
For security, at airports, railway stations, subway stations, warehouse sorting places, and the like, security inspection of articles is required to identify dangerous articles therein. Generally, a large number of articles pass through a security inspection machine within a certain period of time, so that computer vision technology is needed to autonomously assist security inspection personnel in identifying dangerous articles and determining types and packages of the dangerous articles. At present, liquid dangerous goods are more difficult to identify than solid dangerous goods, so that the liquid dangerous goods need to be identified in a targeted manner.
The existing scheme usually adopts a security inspection instrument to scan passing articles and transmits scanning images of the articles to a visual interface, and then if liquid dangerous articles are identified manually, the identification of the liquid dangerous articles by utilizing manual experience has the problem of inaccuracy, so that the detection rate of the dangerous articles is low, namely the accuracy rate of the existing scheme is low.
Disclosure of Invention
The invention provides a dangerous goods identification method, a dangerous goods identification device, dangerous goods identification equipment and a storage medium based on artificial intelligence, which are used for improving the accuracy of liquid dangerous goods identification.
The invention provides a dangerous goods identification method based on artificial intelligence, which comprises the following steps: acquiring an initial security inspection image to be detected based on a preset liquid dangerous goods detection terminal, and calling a preset ultrahigh resolution model to perform denoising processing on the initial security inspection image to obtain a target high-definition image corresponding to the initial security inspection image; performing image segmentation on the target high-definition image based on a preset segmentation algorithm to obtain a plurality of image subregions corresponding to the target high-definition image, and inputting the target high-definition image into a preset feature extraction network to perform image feature extraction to obtain image features corresponding to the target high-definition image; inputting the image sub-regions and the image characteristics into a preset liquid dangerous article identification model for identifying a liquid dangerous article to obtain a target liquid dangerous article and a liquid dangerous article category, wherein the liquid dangerous article identification model comprises: the device comprises a pooling layer, a full-connection layer, a convolution layer and a hidden layer; generating area position information corresponding to the target liquid dangerous goods according to the target liquid dangerous goods; and generating a detection result corresponding to the target high-definition image according to the region position information and the liquid dangerous goods category to obtain a dangerous goods detection result.
Optionally, in a first implementation manner of the first aspect of the present invention, the obtaining, by the preset-based liquid hazardous article detection terminal, an initial security inspection image to be detected, and performing denoising processing on the initial security inspection image by using a preset ultrahigh resolution model to obtain a target high-definition image corresponding to the initial security inspection image includes: receiving an initial security inspection image to be detected, which is acquired by a liquid hazardous article detection terminal in real time; inputting the initial security check image into a preset ultrahigh resolution model, wherein the ultrahigh resolution model comprises: the device comprises a convolution layer, an activation layer, a residual error layer, an upper sampling layer and an output layer; and removing noise of the initial security check image through the ultrahigh resolution model to obtain a target high-definition image corresponding to the initial security check image.
Optionally, in a second implementation manner of the first aspect of the present invention, the image segmentation, performed on the target high-definition image based on a preset segmentation algorithm, to obtain a plurality of image sub-regions corresponding to the target high-definition image, and inputting the target high-definition image into a preset feature extraction network to perform image feature extraction, to obtain image features corresponding to the target high-definition image, includes: carrying out normalization processing on the target high-definition image according to a preset image size to obtain a standard image; performing pixel-level segmentation on the articles in the standard image through a preset segmentation algorithm to obtain a plurality of image subregions corresponding to the target high-definition image; inputting the target high-definition image into a preset feature extraction network, and extracting the overall features corresponding to the target high-definition image through the feature extraction network; extracting liquid features in the overall features based on a preset feature attention mechanism; and performing multi-stage feature fusion on the liquid features and the overall features to obtain image features corresponding to the target high-definition image.
Optionally, in a third implementation manner of the first aspect of the present invention, the image sub-regions and the image features are input into a preset liquid dangerous article identification model to perform liquid dangerous article identification, so as to obtain a target liquid dangerous article and a liquid dangerous article category, where the liquid dangerous article identification model includes: pooling layer, full tie layer, convolution layer and hidden layer include: inputting the plurality of image subregions and the image characteristics into a preset liquid dangerous goods identification model, wherein the liquid dangerous goods identification model comprises: the device comprises a pooling layer, a full-connection layer, a convolution layer and a hidden layer; performing downsampling processing on the image features through the pooling layer to obtain an initial feature vector; inputting the initial feature vector into the full-connection layer to perform high-dimensional feature extraction, and outputting a high-dimensional feature vector; extracting the characteristics of the plurality of image sub-regions through the convolution layer to obtain the region characteristics of each image sub-region; and inputting the regional characteristics of each image subregion and the high-dimensional characteristic vector into the hidden layer for category identification to obtain a target liquid dangerous article and a liquid dangerous article category.
Optionally, in a fourth implementation manner of the first aspect of the present invention, the generating, according to the target liquid dangerous article, area position information corresponding to the target liquid dangerous article includes: matching an image subregion corresponding to the target liquid dangerous article according to the target liquid dangerous article; acquiring coordinate information of a sub-region of the image where the target liquid dangerous article is located; and generating area position information corresponding to the target liquid dangerous goods according to the coordinate information.
Optionally, in a fifth implementation manner of the first aspect of the present invention, the generating a detection result corresponding to the target high-definition image according to the area location information and the liquid hazardous article category to obtain a hazardous article detection result includes: according to the region position information and the liquid dangerous goods category, carrying out liquid dangerous goods category labeling on the target high-definition image to obtain a labeled image; and calling a preset visual interface and outputting a dangerous article detection result according to the labeled image, wherein the dangerous article detection result comprises a labeled image with a dangerous article position and a liquid dangerous article category.
The invention provides a dangerous goods identification device based on artificial intelligence, which comprises: the system comprises an acquisition module, a detection module and a processing module, wherein the acquisition module is used for acquiring an initial security inspection image to be detected based on a preset liquid dangerous goods detection terminal, and calling a preset ultrahigh resolution model to perform denoising processing on the initial security inspection image to obtain a target high-definition image corresponding to the initial security inspection image; the segmentation module is used for carrying out image segmentation on the target high-definition image based on a preset segmentation algorithm to obtain a plurality of image sub-regions corresponding to the target high-definition image, and inputting the target high-definition image into a preset feature extraction network for image feature extraction to obtain image features corresponding to the target high-definition image; the identification module is used for inputting the image sub-regions and the image characteristics into a preset liquid dangerous article identification model to identify a liquid dangerous article, so as to obtain a target liquid dangerous article and a liquid dangerous article category, wherein the liquid dangerous article identification model comprises: a pooling layer, a full-link layer, a convolution layer and a hidden layer; the processing module is used for generating area position information corresponding to the target liquid dangerous goods according to the target liquid dangerous goods; and the generating module is used for generating a detection result corresponding to the target high-definition image according to the region position information and the liquid dangerous goods category to obtain a dangerous goods detection result.
Optionally, in a first implementation manner of the second aspect of the present invention, the obtaining module is specifically configured to: receiving an initial security inspection image to be detected, which is acquired by a liquid hazardous article detection terminal in real time; inputting the initial security check image into a preset ultrahigh resolution model, wherein the ultrahigh resolution model comprises: the device comprises a convolution layer, an activation layer, a residual error layer, an upper sampling layer and an output layer; and removing noise from the initial security inspection image through the ultrahigh resolution model to obtain a target high-definition image corresponding to the initial security inspection image.
Optionally, in a second implementation manner of the second aspect of the present invention, the dividing module is specifically configured to: carrying out normalization processing on the target high-definition image according to a preset image size to obtain a standard image; performing pixel-level segmentation on the articles in the standard image through a preset segmentation algorithm to obtain a plurality of image subregions corresponding to the target high-definition image; inputting the target high-definition image into a preset feature extraction network, and extracting the overall features corresponding to the target high-definition image through the feature extraction network; extracting liquid features in the overall features based on a preset feature attention mechanism; and performing multi-stage feature fusion on the liquid features and the overall features to obtain image features corresponding to the target high-definition image.
Optionally, in a third implementation manner of the second aspect of the present invention, the identification module is specifically configured to: inputting the plurality of image subregions and the image characteristics into a preset liquid dangerous goods identification model, wherein the liquid dangerous goods identification model comprises: the device comprises a pooling layer, a full-connection layer, a convolution layer and a hidden layer; performing downsampling processing on the image features through the pooling layer to obtain an initial feature vector; inputting the initial feature vector into the full-connection layer to perform high-dimensional feature extraction, and outputting a high-dimensional feature vector; extracting the characteristics of the plurality of image sub-regions through the convolution layer to obtain the region characteristics of each image sub-region; and inputting the regional characteristics of each image subregion and the high-dimensional characteristic vector into the hidden layer for category identification to obtain a target liquid dangerous article and a liquid dangerous article category.
Optionally, in a fourth implementation manner of the second aspect of the present invention, the processing module is specifically configured to: matching an image subregion corresponding to the target liquid dangerous article according to the target liquid dangerous article; acquiring coordinate information of a sub-region of the image where the target liquid dangerous article is located; and generating area position information corresponding to the target liquid dangerous goods according to the coordinate information.
Optionally, in a fifth implementation manner of the second aspect of the present invention, the generating module is specifically configured to: according to the region position information and the liquid dangerous goods category, carrying out liquid dangerous goods category labeling on the target high-definition image to obtain a labeled image; and calling a preset visual interface and outputting a dangerous goods detection result according to the labeled image, wherein the dangerous goods detection result comprises a labeled image with a dangerous goods position and a liquid dangerous goods category.
The third aspect of the present invention provides a dangerous goods identification device based on artificial intelligence, comprising: a memory and at least one processor, the memory having instructions stored therein; the at least one processor invokes the instructions in the memory to cause the artificial intelligence based threat identification apparatus to perform the artificial intelligence based threat identification method described above.
A fourth aspect of the present invention provides a computer-readable storage medium having stored therein instructions, which, when run on a computer, cause the computer to execute the artificial intelligence based hazardous material identification method described above.
In the technical scheme provided by the invention, an initial security inspection image to be detected is obtained based on a preset liquid dangerous goods detection terminal, and a preset ultrahigh resolution model is called to perform denoising processing on the initial security inspection image to obtain a target high-definition image corresponding to the initial security inspection image; performing image segmentation on the target high-definition image based on a preset segmentation algorithm to obtain a plurality of image subregions corresponding to the target high-definition image, and inputting the target high-definition image into a preset feature extraction network to perform image feature extraction to obtain image features corresponding to the target high-definition image; inputting the image sub-regions and the image characteristics into a preset liquid dangerous article identification model for identifying liquid dangerous articles to obtain target liquid dangerous articles and liquid dangerous article categories, wherein the liquid dangerous article identification model comprises: the device comprises a pooling layer, a full-connection layer, a convolution layer and a hidden layer; generating area position information corresponding to the target liquid dangerous goods according to the target liquid dangerous goods; and generating a detection result corresponding to the target high-definition image according to the region position information and the liquid dangerous goods category to obtain a dangerous goods detection result. According to the method, the image is subjected to noise removal through the ultrahigh resolution model to generate the high-definition image, the definition of the image is improved, the subsequent detection of the dangerous goods is facilitated, and the accuracy of dangerous goods identification is improved through constructing the deep learning model with specific identification on the liquid dangerous goods.
Drawings
FIG. 1 is a diagram of an embodiment of a dangerous goods identification method based on artificial intelligence according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of another embodiment of a dangerous goods identification method based on artificial intelligence in the embodiment of the invention;
FIG. 3 is a schematic diagram of an embodiment of a dangerous goods identification device based on artificial intelligence according to the embodiment of the invention;
FIG. 4 is a schematic diagram of another embodiment of a dangerous goods identification device based on artificial intelligence in the embodiment of the invention;
fig. 5 is a schematic diagram of an embodiment of a dangerous goods identification device based on artificial intelligence in the embodiment of the invention.
Detailed Description
The embodiment of the invention provides a dangerous goods identification method, a dangerous goods identification device, dangerous goods identification equipment and a storage medium based on artificial intelligence, which are used for improving the accuracy of liquid dangerous goods identification. The terms "first," "second," "third," "fourth," and the like in the description and in the claims, as well as in the drawings, if any, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It will be appreciated that the data so used may be interchanged under appropriate circumstances such that the embodiments described herein may be practiced otherwise than as specifically illustrated or described herein. Furthermore, the terms "comprises," "comprising," or "having," and any variations thereof, are intended to cover non-exclusive inclusions, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
For convenience of understanding, a specific flow of an embodiment of the present invention is described below, and referring to fig. 1, an embodiment of a dangerous goods identification method based on artificial intelligence in an embodiment of the present invention includes:
101. acquiring an initial security inspection image to be detected based on a preset liquid dangerous goods detection terminal, and calling a preset ultrahigh resolution model to perform denoising processing on the initial security inspection image to obtain a target high-definition image corresponding to the initial security inspection image;
it is to be understood that the execution subject of the present invention may be a hazardous article identification device based on artificial intelligence, and may also be a terminal or a server, which is not limited herein. The embodiment of the present invention is described by taking a server as an execution subject.
Specifically, since liquid dangerous goods are usually stored in containers with fixed shapes, the containers of the liquid dangerous goods are identified by constructing a deep learning model for identifying the containers of the liquid dangerous goods, the server firstly controls a preset liquid dangerous goods detection terminal to acquire images, when a luggage object of a user passes through the liquid dangerous goods detection terminal, the luggage is scanned through the liquid dangerous goods detection terminal, a scanning image is generated, and the scanning image is used as an initial security inspection image to be inspected. The server performs denoising processing on the initial security inspection image through a preset ultrahigh resolution model to generate a high-definition target high-definition image, wherein the ultrahigh resolution model can be an EDSR model which specifically comprises a convolution layer, an activation layer, a residual error layer, an upper sampling layer and an output layer. The ultrahigh resolution model adopted in the embodiment improves the resolution of the image by removing noise from the initial security inspection image, and further improves the accuracy of identification of the liquid dangerous goods.
102. Carrying out image segmentation on the target high-definition image based on a preset segmentation algorithm to obtain a plurality of image subregions corresponding to the target high-definition image, and inputting the target high-definition image into a preset feature extraction network to carry out image feature extraction to obtain image features corresponding to the target high-definition image;
specifically, the server firstly normalizes the target high-definition image to 450 × 225 (i.e., length × width), and since the pixel-level segmentation is directly completed manually with a large workload, the pixel-level segmentation can be completed by combining with the current segmentation algorithm, and a plurality of image sub-regions corresponding to the target high-definition image are obtained after the segmentation is completed. Further, the server performs pixel-level segmentation on the articles in the target high-definition image, that is, all pixels of one target high-definition image are distinguished according to corresponding article positions through a correlation algorithm, and an article region corresponding to the pixel is labeled, wherein the pixel proportion of the articles in the image sub-region is the true proportion of the articles in the image sub-region. It should be noted that, the feature extraction network may select a resnet residual error network to perform image feature extraction, and generate an image feature corresponding to the target high-definition image.
103. Inputting a plurality of image subregions and image characteristics into a preset liquid dangerous article identification model for liquid dangerous article identification to obtain a target liquid dangerous article and a liquid dangerous article category, wherein the liquid dangerous article identification model comprises: the device comprises a pooling layer, a full-connection layer, a convolution layer and a hidden layer;
specifically, a plurality of image sub-regions and image characteristics are used as input, and liquid dangerous goods identification is carried out sequentially through the pooling layer, the full-connection layer, the convolution layer and the hidden layer, so that a target liquid dangerous goods and a liquid dangerous goods category are obtained. The pooling layer is a layer in the liquid hazardous article identification model, which is used to obtain fixed-length feature vectors through a down-sampling operation. The hidden layer is provided with a directionally-linked neural network, is used for capturing the context information of the sequence and is more suitable for the task of sequence input or output. The image sub-regions and the image features are output to 1024-dimensional features after passing through a pooling layer and a full connection layer, and the convolution layer is used for extracting the features of the image sub-regions.
104. Generating area position information corresponding to the target liquid dangerous goods according to the target liquid dangerous goods;
specifically, the server matches image sub-regions corresponding to the target liquid dangerous article according to the target liquid dangerous article, that is, the server correspondingly processes the image sub-regions where the target liquid dangerous article is located one by one; the server acquires coordinate information of an image subregion where the target liquid dangerous article is located, and the server acquires the coordinate position of the image subregion corresponding to the target liquid dangerous article in the whole target high-definition image and generates the coordinate information; and the server generates area position information corresponding to the target liquid dangerous article according to the coordinate information, wherein the area position information is the position of the target liquid dangerous article.
105. And generating a detection result corresponding to the target high-definition image according to the region position information and the liquid dangerous goods category to obtain a dangerous goods detection result.
Specifically, the server generates a detection result corresponding to the target high-definition image according to the regional position information and the liquid dangerous goods category to obtain a dangerous goods detection result, visually marks the regional position information corresponding to the target liquid dangerous goods through a visual tool, and marks the category information of the dangerous goods in a marking frame.
In the embodiment of the invention, an initial security inspection image to be detected is obtained based on a preset liquid dangerous goods detection terminal, and a preset ultrahigh resolution model is called to perform denoising processing on the initial security inspection image to obtain a target high-definition image corresponding to the initial security inspection image; carrying out image segmentation on the target high-definition image based on a preset segmentation algorithm to obtain a plurality of image subregions corresponding to the target high-definition image, and inputting the target high-definition image into a preset feature extraction network to carry out image feature extraction to obtain image features corresponding to the target high-definition image; inputting a plurality of image subregions and image characteristics into a preset liquid dangerous article identification model for liquid dangerous article identification to obtain a target liquid dangerous article and a liquid dangerous article category, wherein the liquid dangerous article identification model comprises: the device comprises a pooling layer, a full-connection layer, a convolution layer and a hidden layer; generating area position information corresponding to the target liquid dangerous goods according to the target liquid dangerous goods; and generating a detection result corresponding to the target high-definition image according to the region position information and the liquid dangerous goods category to obtain a dangerous goods detection result. According to the method, the image is subjected to noise removal through the ultrahigh resolution model to generate the high-definition image, the definition of the image is improved, the subsequent detection of the dangerous goods is facilitated, and the accuracy of dangerous goods identification is improved through constructing the deep learning model with specific identification on the liquid dangerous goods.
Referring to fig. 2, another embodiment of the dangerous goods identification method based on artificial intelligence according to the embodiment of the present invention includes:
201. receiving an initial security inspection image to be detected, which is acquired by a liquid hazardous article detection terminal in real time;
specifically, a server receives an initial security inspection image to be detected acquired by a liquid hazardous article detection terminal in real time, model training needs to be carried out on an ultrahigh resolution model before the ultrahigh resolution model is input, the server firstly carries out data set division on a sample image data set to obtain a training set and a test set required by the EDSR model to be trained, wherein the training set is 50% of the whole sample image data set, and the test set is 50% of the sample image data set; the server carries out model training on the EDSR model according to the training set, outputs a prediction result, judges whether the EDSR model is trained or not according to the prediction result, and if the EDSR model is trained, the EDSR model which is trained is used as an ultrahigh resolution model.
202. Inputting the initial security check image into a preset ultrahigh resolution model, wherein the ultrahigh resolution model comprises: the device comprises a convolution layer, an activation layer, a residual error layer, an upper sampling layer and an output layer;
specifically, after the initial security inspection image is input, convolution, activation and residual scaling are performed to learn the high-frequency characteristics of the image, and then upsampling (corresponding to downsampling pooling, and upsampling is a step of increasing resolution) is performed to perform final ultra-resolution reconstruction of the image. The convolution layer is used for extracting the features of the image, the weight of the convolution kernel can be learned, the convolution operation can break through the limitation of a traditional filter, and the desired features are extracted according to the target function; and the activation layer is the activation function layer, and activates the feature matrix convolved in the previous step to form a new matrix. The residual layer is a module for enlarging the resolution of the image, and its function is to compress the input picture to a certain rule, or to a desired size. In addition, parameter sharing (one convolution kernel is used for filtering the whole image instead of a plurality of convolution kernels) reduces network parameters and improves training efficiency. According to the method, the noise is removed through the ultrahigh resolution model, the high-definition target image is generated, and the accuracy of identifying the liquid dangerous goods is improved.
203. Removing noise from the initial security inspection image through the ultrahigh resolution model to obtain a target high-definition image corresponding to the initial security inspection image;
specifically, the server inputs the initial security inspection image into the ultrahigh resolution model for image denoising, and generates a target high-definition image corresponding to the initial security inspection image.
204. Performing image segmentation on the target high-definition image based on a preset segmentation algorithm to obtain a plurality of image subregions corresponding to the target high-definition image, and inputting the target high-definition image into a preset feature extraction network to perform image feature extraction to obtain image features corresponding to the target high-definition image;
specifically, the server normalizes the target high-definition image according to a preset image size to obtain a standard image; the server performs pixel-level segmentation on the articles in the standard image through a preset segmentation algorithm to obtain a plurality of image sub-regions corresponding to the target high-definition image; the server inputs the target high-definition image into a preset feature extraction network, and extracts the overall features corresponding to the target high-definition image through the feature extraction network; the server extracts liquid features in the overall features based on a preset feature attention mechanism; and the server performs multi-stage feature fusion on the liquid features and the overall features to obtain image features corresponding to the target high-definition image. Specifically, a face overall feature map corresponding to a target high-definition image is extracted through a feature extraction network, a residual error network can be adopted as the feature extraction network in the embodiment, the residual error network comprises 5 stages in total, each stage can carry out convolution operation, and as the stages deepen, the depth of the obtained feature map is deeper, an attention mechanism is applied to feature extraction, so that the accuracy of dangerous goods identification is improved, dangerous goods are better positioned, and the feature attention mechanism can be built through a soft attention mechanism in the embodiment; and performing multi-stage feature fusion on the liquid feature map and the overall feature map to obtain a target feature map, wherein in order to prevent jitter, the feature fusion layer performs 3x3 convolution operation on all input feature maps and then performs down-sampling, so that multi-stage feature fusion is realized, and image features corresponding to a target high-definition image are generated.
205. Inputting a plurality of image subregions and image characteristics into a preset liquid dangerous article identification model for liquid dangerous article identification to obtain a target liquid dangerous article and a liquid dangerous article category, wherein the liquid dangerous article identification model comprises: the device comprises a pooling layer, a full-connection layer, a convolution layer and a hidden layer;
specifically, the server inputs a plurality of image subregions and image features into a preset liquid hazardous article identification model, wherein the liquid hazardous article identification model comprises: the device comprises a pooling layer, a full-connection layer, a convolution layer and a hidden layer; the server performs downsampling processing on the image features through the pooling layer to obtain initial feature vectors; the server inputs the initial feature vector into a full connection layer to carry out high-dimensional feature extraction, and outputs a high-dimensional feature vector; the server extracts the characteristics of the plurality of image sub-regions through the convolution layer to obtain the region characteristics of each image sub-region; the server inputs the regional characteristics and the high-dimensional characteristic vectors of each image subregion into a hidden layer for category identification to obtain a target liquid dangerous article and a liquid dangerous article category, wherein the hidden layer is provided with a directional linked neural network and is used for capturing context information of a sequence and is more suitable for a task of sequence input or output. Specifically, a plurality of image sub-regions and image characteristics are used as input, and liquid dangerous goods identification is carried out sequentially through the pooling layer, the full-connection layer, the convolution layer and the hidden layer, so that a target liquid dangerous goods and a liquid dangerous goods category are obtained. The pooling layer is a layer in the liquid hazardous article identification model, which is used to obtain fixed-length feature vectors through a down-sampling operation. The hidden layer is provided with a directionally-linked neural network, is used for capturing the context information of the sequence and is more suitable for the task of sequence input or output. The image sub-regions and the image features are output to 1024-dimensional features after passing through a pooling layer and a full connection layer, and the convolution layer is used for extracting the features of the image sub-regions. In this embodiment, 5 × 5 convolution is adopted, 1024-dimensional features of each image sub-region are used as 1024 channels, and new 1024-dimensional features are formed after respective convolution as features of the region. In order to ensure that the feature quantities of the image sub-regions before and after convolution are consistent, zero-filling processing needs to be performed on the relevant region outside the image when the image sub-regions at the edge part are convoluted.
206. Generating area position information corresponding to the target liquid dangerous goods according to the target liquid dangerous goods;
specifically, the server matches an image subregion corresponding to the target liquid dangerous article according to the target liquid dangerous article; the server acquires coordinate information of a sub-region of an image where the target liquid dangerous article is located; and the server generates area position information corresponding to the target liquid dangerous goods according to the coordinate information. Further, the server matches image sub-regions corresponding to the target liquid dangerous articles according to the target liquid dangerous articles, namely, the server correspondingly processes the image sub-regions where the target liquid dangerous articles are located one by one; the server acquires coordinate information of an image subregion where the target liquid dangerous article is located, and the server acquires the coordinate position of the image subregion corresponding to the target liquid dangerous article in the whole target high-definition image and generates the coordinate information; and the server generates area position information corresponding to the target liquid dangerous goods according to the coordinate information, wherein the area position information is the position of the target liquid dangerous goods. And displaying the identification information on the rolling image in an overlapping manner. And the dangerous goods identification is carried out according to the identified dangerous goods information, the position area of the rolling image corresponding to the suspected dangerous goods is marked, so that the display result has the rolling image and the marked dangerous goods information at the same time, the marking form comprises but is not limited to a marking frame, a character prompt and the like, the marked content is the related information of the suspected dangerous goods, and meanwhile, a corresponding sound prompt can be given.
207. And generating a detection result corresponding to the target high-definition image according to the region position information and the liquid dangerous goods category to obtain a dangerous goods detection result.
Specifically, the server performs liquid dangerous goods category labeling on the target high-definition image according to the region position information and the liquid dangerous goods category to obtain a labeled image; and the server calls a preset visual interface and outputs a dangerous goods detection result according to the marked image, wherein the dangerous goods detection result comprises the marked image with the position of the dangerous goods and the category of the liquid dangerous goods. Further, the server generates a detection result corresponding to the target high-definition image according to the region position information and the liquid dangerous goods category to obtain a dangerous goods detection result, the server visually marks the region position information corresponding to the target liquid dangerous goods through a visual tool, and the marking frame is marked with the category information of the dangerous goods.
In the embodiment of the invention, an initial security inspection image to be detected is obtained based on a preset liquid hazardous article detection terminal, and a preset ultrahigh resolution model is called to perform denoising processing on the initial security inspection image to obtain a target high-definition image corresponding to the initial security inspection image; carrying out image segmentation on the target high-definition image based on a preset segmentation algorithm to obtain a plurality of image subregions corresponding to the target high-definition image, and inputting the target high-definition image into a preset feature extraction network to carry out image feature extraction to obtain image features corresponding to the target high-definition image; inputting a plurality of image subregions and image characteristics into a preset liquid dangerous article identification model for liquid dangerous article identification to obtain a target liquid dangerous article and a liquid dangerous article category, wherein the liquid dangerous article identification model comprises: a pooling layer, a full-link layer, a convolution layer and a hidden layer; generating area position information corresponding to the target liquid dangerous goods according to the target liquid dangerous goods; and generating a detection result corresponding to the target high-definition image according to the region position information and the liquid dangerous goods category to obtain a dangerous goods detection result. According to the method, the image is subjected to noise removal through the ultrahigh resolution model to generate the high-definition image, the definition of the image is improved, the subsequent detection of the dangerous goods is facilitated, and the accuracy of dangerous goods identification is improved through constructing the deep learning model with specific identification on the liquid dangerous goods.
With reference to fig. 3, the above description is made on the method for identifying a hazardous material based on artificial intelligence in the embodiment of the present invention, and the following description is made on a hazardous material identification apparatus based on artificial intelligence in the embodiment of the present invention, where an embodiment of the hazardous material identification apparatus based on artificial intelligence in the embodiment of the present invention includes:
the acquisition module 301 is configured to acquire an initial security inspection image to be detected based on a preset liquid hazardous article detection terminal, and call a preset ultrahigh resolution model to perform denoising processing on the initial security inspection image to obtain a target high-definition image corresponding to the initial security inspection image;
the segmentation module 302 is configured to perform image segmentation on the target high-definition image based on a preset segmentation algorithm to obtain a plurality of image sub-regions corresponding to the target high-definition image, and input the target high-definition image into a preset feature extraction network to perform image feature extraction to obtain image features corresponding to the target high-definition image;
the identification module 303 is configured to input the plurality of image subregions and the image features into a preset liquid dangerous article identification model to perform liquid dangerous article identification, so as to obtain a target liquid dangerous article and a liquid dangerous article category, where the liquid dangerous article identification model includes: the device comprises a pooling layer, a full-connection layer, a convolution layer and a hidden layer;
the processing module 304 is configured to generate area position information corresponding to the target liquid dangerous article according to the target liquid dangerous article;
the generating module 305 is configured to generate a detection result corresponding to the target high-definition image according to the region location information and the liquid hazardous article category, so as to obtain a hazardous article detection result.
In the embodiment of the invention, an initial security inspection image to be detected is obtained based on a preset liquid hazardous article detection terminal, and a preset ultrahigh resolution model is called to perform denoising processing on the initial security inspection image to obtain a target high-definition image corresponding to the initial security inspection image; performing image segmentation on the target high-definition image based on a preset segmentation algorithm to obtain a plurality of image subregions corresponding to the target high-definition image, and inputting the target high-definition image into a preset feature extraction network to perform image feature extraction to obtain image features corresponding to the target high-definition image; inputting the image sub-regions and the image characteristics into a preset liquid dangerous article identification model for identifying a liquid dangerous article to obtain a target liquid dangerous article and a liquid dangerous article category, wherein the liquid dangerous article identification model comprises: the device comprises a pooling layer, a full-connection layer, a convolution layer and a hidden layer; generating area position information corresponding to the target liquid dangerous goods according to the target liquid dangerous goods; and generating a detection result corresponding to the target high-definition image according to the region position information and the liquid dangerous goods category to obtain a dangerous goods detection result. According to the method, the image is subjected to noise removal through the ultrahigh resolution model to generate the high-definition image, the definition of the image is improved, the subsequent detection of the dangerous goods is facilitated, and the accuracy of dangerous goods identification is improved through constructing the deep learning model with specific identification on the liquid dangerous goods.
Referring to fig. 4, another embodiment of the dangerous goods identification apparatus based on artificial intelligence according to the embodiment of the present invention includes:
the acquisition module 301 is configured to acquire an initial security inspection image to be detected based on a preset liquid hazardous article detection terminal, and call a preset ultrahigh resolution model to perform denoising processing on the initial security inspection image to obtain a target high-definition image corresponding to the initial security inspection image;
the segmentation module 302 is configured to perform image segmentation on the target high-definition image based on a preset segmentation algorithm to obtain a plurality of image sub-regions corresponding to the target high-definition image, and input the target high-definition image into a preset feature extraction network to perform image feature extraction to obtain image features corresponding to the target high-definition image;
the identification module 303 is configured to input the plurality of image subregions and the image features into a preset liquid dangerous article identification model to perform liquid dangerous article identification, so as to obtain a target liquid dangerous article and a liquid dangerous article category, where the liquid dangerous article identification model includes: the device comprises a pooling layer, a full-connection layer, a convolution layer and a hidden layer;
the processing module 304 is configured to generate area position information corresponding to the target liquid dangerous article according to the target liquid dangerous article;
the generating module 305 is configured to generate a detection result corresponding to the target high-definition image according to the region location information and the liquid hazardous article category, so as to obtain a hazardous article detection result.
Optionally, the obtaining module 301 is specifically configured to:
the receiving unit 3011 is configured to receive an initial security inspection image to be detected, acquired by a liquid hazardous article detection terminal in real time;
an input unit 3012, configured to input the initial security inspection image into a preset ultrahigh resolution model, where the ultrahigh resolution model includes: the device comprises a convolution layer, an activation layer, a residual error layer, an upper sampling layer and an output layer;
and the denoising unit 3013 is configured to perform noise removal on the initial security inspection image through the ultrahigh resolution model, so as to obtain a target high-definition image corresponding to the initial security inspection image.
Optionally, the segmentation module 302 is specifically configured to: carrying out normalization processing on the target high-definition image according to a preset image size to obtain a standard image; performing pixel-level segmentation on the articles in the standard image through a preset segmentation algorithm to obtain a plurality of image subregions corresponding to the target high-definition image; inputting the target high-definition image into a preset feature extraction network, and extracting the overall features corresponding to the target high-definition image through the feature extraction network; extracting liquid features in the overall features based on a preset feature attention mechanism; and performing multi-stage feature fusion on the liquid features and the overall features to obtain image features corresponding to the target high-definition image.
Optionally, the identifying module 303 is specifically configured to: inputting the plurality of image subregions and the image characteristics into a preset liquid dangerous article identification model, wherein the liquid dangerous article identification model comprises: the device comprises a pooling layer, a full-connection layer, a convolution layer and a hidden layer; performing downsampling processing on the image features through the pooling layer to obtain an initial feature vector; inputting the initial feature vector into the full-connection layer to perform high-dimensional feature extraction, and outputting a high-dimensional feature vector; extracting the characteristics of the plurality of image sub-regions through the convolution layer to obtain the region characteristics of each image sub-region; and inputting the regional characteristics of each image subregion and the high-dimensional characteristic vector into the hidden layer for category identification to obtain a target liquid dangerous article and a liquid dangerous article category.
Optionally, the processing module 304 is specifically configured to: matching an image subregion corresponding to the target liquid dangerous article according to the target liquid dangerous article; acquiring coordinate information of a sub-region of the image where the target liquid dangerous article is located; and generating area position information corresponding to the target liquid dangerous goods according to the coordinate information.
Optionally, the generating module 305 is specifically configured to: according to the region position information and the liquid dangerous goods category, carrying out liquid dangerous goods category labeling on the target high-definition image to obtain a labeled image; and calling a preset visual interface and outputting a dangerous goods detection result according to the labeled image, wherein the dangerous goods detection result comprises a labeled image with a dangerous goods position and a liquid dangerous goods category.
In the embodiment of the invention, an initial security inspection image to be detected is obtained based on a preset liquid dangerous goods detection terminal, and a preset ultrahigh resolution model is called to perform denoising processing on the initial security inspection image to obtain a target high-definition image corresponding to the initial security inspection image; performing image segmentation on the target high-definition image based on a preset segmentation algorithm to obtain a plurality of image subregions corresponding to the target high-definition image, and inputting the target high-definition image into a preset feature extraction network to perform image feature extraction to obtain image features corresponding to the target high-definition image; inputting the image sub-regions and the image characteristics into a preset liquid dangerous article identification model for identifying liquid dangerous articles to obtain target liquid dangerous articles and liquid dangerous article categories, wherein the liquid dangerous article identification model comprises: the device comprises a pooling layer, a full-connection layer, a convolution layer and a hidden layer; generating area position information corresponding to the target liquid dangerous goods according to the target liquid dangerous goods; and generating a detection result corresponding to the target high-definition image according to the region position information and the liquid dangerous goods category to obtain a dangerous goods detection result. According to the method, the image is subjected to noise removal through the ultrahigh resolution model to generate the high-definition image, the definition of the image is improved, the subsequent detection of the dangerous goods is facilitated, and the accuracy of dangerous goods identification is improved through constructing the deep learning model with specific identification on the liquid dangerous goods.
Fig. 3 and 4 above describe the hazardous material identification device based on artificial intelligence in the embodiment of the present invention in detail from the perspective of the modular functional entity, and the hazardous material identification device based on artificial intelligence in the embodiment of the present invention is described in detail from the perspective of hardware processing.
Fig. 5 is a schematic structural diagram of an artificial intelligence based hazardous article identification device 500 according to an embodiment of the present invention, where the hazardous article identification device 500 may have a relatively large difference due to different configurations or performances, and may include one or more processors (CPUs) 510 (e.g., one or more processors) and a memory 520, one or more storage media 530 (e.g., one or more mass storage devices) storing applications 533 or data 532. Memory 520 and storage media 530 may be, among other things, transient or persistent storage. The program stored on the storage medium 530 may include one or more modules (not shown), each of which may include a sequence of instructions operating on the artificial intelligence based hazardous materials identification device 500. Still further, processor 510 may be configured to communicate with storage medium 530 to execute a series of instruction operations in storage medium 530 on artificial intelligence based hazardous materials identification device 500.
Artificial intelligence based hazardous articles identification device 500 may also include one or more power supplies 540, one or more wired or wireless network interfaces 550, one or more input-output interfaces 560, and/or one or more operating systems 531, such as Windows server, Mac OS X, Unix, Linux, FreeBSD, and so on. Those skilled in the art will appreciate that the configuration of the artificial intelligence based threat identification apparatus illustrated in fig. 5 does not constitute a limitation of the artificial intelligence based threat identification apparatus, and may include more or fewer components than those illustrated, or some components in combination, or a different arrangement of components.
The invention also provides an artificial intelligence based dangerous goods identification device, which comprises a memory and a processor, wherein computer readable instructions are stored in the memory, and when being executed by the processor, the computer readable instructions cause the processor to execute the steps of the artificial intelligence based dangerous goods identification method in the embodiments.
The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium, and which may also be a volatile computer-readable storage medium, having stored therein instructions, which, when executed on a computer, cause the computer to perform the steps of the artificial intelligence based threat identification method.
Further, the computer-readable storage medium may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function, and the like; the storage data area may store data created according to the use of the blockchain node, and the like.
The block chain is a novel application mode of computer technologies such as distributed data storage, point-to-point transmission, a consensus mechanism, an encryption algorithm and the like. A block chain (Blockchain), which is essentially a decentralized database, is a string of data blocks associated by using a cryptographic method, and each data block contains information of a batch of network transactions, which is used for verifying the validity (anti-counterfeiting) of the information and generating a next block. The blockchain may include a blockchain underlying platform, a platform product service layer, an application service layer, and the like.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a read-only memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the same; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A dangerous goods identification method based on artificial intelligence is characterized by comprising the following steps:
acquiring an initial security inspection image to be detected based on a preset liquid hazardous article detection terminal, and calling a preset ultrahigh resolution model to perform denoising processing on the initial security inspection image to obtain a target high-definition image corresponding to the initial security inspection image;
performing image segmentation on the target high-definition image based on a preset segmentation algorithm to obtain a plurality of image subregions corresponding to the target high-definition image, and inputting the target high-definition image into a preset feature extraction network to perform image feature extraction to obtain image features corresponding to the target high-definition image;
inputting the image sub-regions and the image characteristics into a preset liquid dangerous article identification model for identifying a liquid dangerous article to obtain a target liquid dangerous article and a liquid dangerous article category, wherein the liquid dangerous article identification model comprises: the device comprises a pooling layer, a full-connection layer, a convolution layer and a hidden layer;
generating area position information corresponding to the target liquid dangerous goods according to the target liquid dangerous goods;
and generating a detection result corresponding to the target high-definition image according to the region position information and the liquid dangerous goods category to obtain a dangerous goods detection result.
2. The dangerous goods identification method based on artificial intelligence of claim 1, wherein the obtaining of the initial security inspection image to be detected by the detection terminal based on the preset liquid dangerous goods and the denoising of the initial security inspection image by calling the preset ultrahigh resolution model to obtain the target high-definition image corresponding to the initial security inspection image comprises:
receiving an initial security inspection image to be detected, which is acquired by a liquid hazardous article detection terminal in real time;
inputting the initial security check image into a preset ultrahigh resolution model, wherein the ultrahigh resolution model comprises: the device comprises a convolution layer, an activation layer, a residual error layer, an upper sampling layer and an output layer;
and removing noise from the initial security inspection image through the ultrahigh resolution model to obtain a target high-definition image corresponding to the initial security inspection image.
3. The artificial intelligence based dangerous goods identification method according to claim 1, wherein the image segmentation is performed on the target high-definition image based on a preset segmentation algorithm to obtain a plurality of image sub-regions corresponding to the target high-definition image, and the target high-definition image is input into a preset feature extraction network to perform image feature extraction to obtain image features corresponding to the target high-definition image, and the method comprises the following steps:
carrying out normalization processing on the target high-definition image according to a preset image size to obtain a standard image;
performing pixel-level segmentation on the articles in the standard image through a preset segmentation algorithm to obtain a plurality of image subregions corresponding to the target high-definition image;
inputting the target high-definition image into a preset feature extraction network, and extracting the overall features corresponding to the target high-definition image through the feature extraction network;
extracting liquid features in the overall features based on a preset feature attention mechanism;
and performing multi-stage feature fusion on the liquid features and the overall features to obtain image features corresponding to the target high-definition image.
4. The method for identifying dangerous goods based on artificial intelligence according to claim 1, wherein the image sub-regions and the image features are input into a preset liquid dangerous goods identification model for liquid dangerous goods identification, so as to obtain a target liquid dangerous goods and a liquid dangerous goods category, wherein the liquid dangerous goods identification model comprises: pooling layer, full tie layer, convolution layer and hidden layer include:
inputting the plurality of image subregions and the image characteristics into a preset liquid dangerous goods identification model, wherein the liquid dangerous goods identification model comprises: the device comprises a pooling layer, a full-connection layer, a convolution layer and a hidden layer;
performing downsampling processing on the image features through the pooling layer to obtain an initial feature vector;
inputting the initial feature vector into the full-connection layer to perform high-dimensional feature extraction, and outputting a high-dimensional feature vector;
extracting the characteristics of the plurality of image sub-regions through the convolution layer to obtain the region characteristics of each image sub-region;
and inputting the regional characteristics of each image subregion and the high-dimensional characteristic vector into the hidden layer for category identification to obtain a target liquid dangerous article and a liquid dangerous article category.
5. The method for identifying dangerous goods based on artificial intelligence according to claim 1, wherein the generating of the area location information corresponding to the target dangerous liquid goods according to the target dangerous liquid goods comprises:
matching an image subregion corresponding to the target liquid dangerous article according to the target liquid dangerous article;
acquiring coordinate information of a sub-region of the image where the target liquid dangerous article is located;
and generating area position information corresponding to the target liquid dangerous goods according to the coordinate information.
6. The method for identifying dangerous goods based on artificial intelligence according to any one of claims 1 to 5, wherein the generating a detection result corresponding to the target high-definition image according to the region location information and the liquid dangerous goods category to obtain a dangerous goods detection result comprises:
according to the region position information and the liquid dangerous goods category, carrying out liquid dangerous goods category labeling on the target high-definition image to obtain a labeled image;
and calling a preset visual interface and outputting a dangerous article detection result according to the labeled image, wherein the dangerous article detection result comprises a labeled image with a dangerous article position and a liquid dangerous article category.
7. An artificial intelligence based hazardous materials identification device, characterized in that, the artificial intelligence based hazardous materials identification device includes:
the system comprises an acquisition module, a detection module and a processing module, wherein the acquisition module is used for acquiring an initial security inspection image to be detected based on a preset liquid dangerous goods detection terminal, and calling a preset ultrahigh resolution model to perform denoising processing on the initial security inspection image to obtain a target high-definition image corresponding to the initial security inspection image;
the segmentation module is used for carrying out image segmentation on the target high-definition image based on a preset segmentation algorithm to obtain a plurality of image subregions corresponding to the target high-definition image, and inputting the target high-definition image into a preset feature extraction network to carry out image feature extraction to obtain image features corresponding to the target high-definition image;
the identification module is used for inputting the image sub-regions and the image characteristics into a preset liquid dangerous article identification model to identify a liquid dangerous article, so as to obtain a target liquid dangerous article and a liquid dangerous article category, wherein the liquid dangerous article identification model comprises: the device comprises a pooling layer, a full-connection layer, a convolution layer and a hidden layer;
the processing module is used for generating area position information corresponding to the target liquid dangerous goods according to the target liquid dangerous goods;
and the generating module is used for generating a detection result corresponding to the target high-definition image according to the region position information and the liquid dangerous goods category to obtain a dangerous goods detection result.
8. The hazardous article identification device based on artificial intelligence of claim 7, wherein the obtaining module is specifically configured to:
receiving an initial security inspection image to be detected, which is acquired by a liquid hazardous article detection terminal in real time;
inputting the initial security check image into a preset ultrahigh resolution model, wherein the ultrahigh resolution model comprises: the device comprises a convolution layer, an activation layer, a residual error layer, an upper sampling layer and an output layer;
and removing noise of the initial security check image through the ultrahigh resolution model to obtain a target high-definition image corresponding to the initial security check image.
9. An artificial intelligence based hazardous materials identification device, characterized in that, the artificial intelligence based hazardous materials identification device includes: a memory and at least one processor, the memory having instructions stored therein;
the at least one processor invoking the instructions in the memory to cause the artificial intelligence based threat identification apparatus to perform the artificial intelligence based threat identification method of any of claims 1-6.
10. A computer-readable storage medium having instructions stored thereon, wherein the instructions, when executed by a processor, implement the artificial intelligence based threat identification method of any of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210412390.8A CN114723724A (en) | 2022-04-19 | 2022-04-19 | Dangerous goods identification method, device, equipment and storage medium based on artificial intelligence |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210412390.8A CN114723724A (en) | 2022-04-19 | 2022-04-19 | Dangerous goods identification method, device, equipment and storage medium based on artificial intelligence |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114723724A true CN114723724A (en) | 2022-07-08 |
Family
ID=82243138
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210412390.8A Withdrawn CN114723724A (en) | 2022-04-19 | 2022-04-19 | Dangerous goods identification method, device, equipment and storage medium based on artificial intelligence |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114723724A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116188802A (en) * | 2023-04-21 | 2023-05-30 | 青岛创新奇智科技集团股份有限公司 | Data labeling method, device, equipment and storage medium |
| CN117495861A (en) * | 2024-01-02 | 2024-02-02 | 同方威视科技江苏有限公司 | Security check image checking method and device |
-
2022
- 2022-04-19 CN CN202210412390.8A patent/CN114723724A/en not_active Withdrawn
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116188802A (en) * | 2023-04-21 | 2023-05-30 | 青岛创新奇智科技集团股份有限公司 | Data labeling method, device, equipment and storage medium |
| CN117495861A (en) * | 2024-01-02 | 2024-02-02 | 同方威视科技江苏有限公司 | Security check image checking method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114723724A (en) | Dangerous goods identification method, device, equipment and storage medium based on artificial intelligence | |
| CN108648192B (en) | Method and device for detecting nodule | |
| CN112883926B (en) | Identification method and device for form medical images | |
| CN114862845B (en) | Defect detection method, device and equipment for mobile phone touch screen and storage medium | |
| CN113627411A (en) | Super-resolution-based commodity identification and price matching method and system | |
| CN112906794A (en) | Target detection method, device, storage medium and terminal | |
| CN112434555B (en) | Key value pair region identification method and device, storage medium and electronic equipment | |
| CN113792623B (en) | Security check CT target object identification method and device | |
| CN110008362A (en) | A kind of case classifying method and device | |
| CN112926601A (en) | Image recognition method, device and equipment based on deep learning and storage medium | |
| CN112288672B (en) | Tumor detection method and tumor detection device | |
| CN114792300B (en) | X-ray broken needle detection method based on multi-scale attention | |
| CN110825896A (en) | Trademark retrieval system and method | |
| CN116704324A (en) | Target detection method, system, equipment and storage medium based on underwater image | |
| CN111784561A (en) | Method, apparatus and storage medium for extracting object from image | |
| CN113139932B (en) | Deep learning defect image identification method and system based on ensemble learning | |
| CN113034420B (en) | Industrial product surface defect segmentation method and system based on frequency space domain characteristics | |
| Dixit et al. | Copy–move image forgery detection using gray-tones with texture description | |
| Singh et al. | Text Region Extraction: A Morphological Based Image Analysis Using Genetic Algorithm | |
| CN110648300A (en) | Image data synthesis method, image data synthesis device, computer equipment and storage medium | |
| CN115298691A (en) | Learning device and learning method | |
| EP4213091A1 (en) | Physics-informed anomaly detection in formed metal parts | |
| CN112988518B (en) | Experimental equipment experimental process real-time monitoring method based on image recognition technology | |
| Sharma | Traffic Sign Recognition & Detection using Transfer learning | |
| CN117152621B (en) | Building change detection method, device, electronic equipment and storage medium |
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 | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20220708 |
|
| WW01 | Invention patent application withdrawn after publication |