CN111985463A - White spirit steaming and steam detecting method based on convolutional neural network - Google Patents
White spirit steaming and steam detecting method based on convolutional neural network Download PDFInfo
- Publication number
- CN111985463A CN111985463A CN202010787213.9A CN202010787213A CN111985463A CN 111985463 A CN111985463 A CN 111985463A CN 202010787213 A CN202010787213 A CN 202010787213A CN 111985463 A CN111985463 A CN 111985463A
- Authority
- CN
- China
- Prior art keywords
- neural network
- steam
- convolutional neural
- white spirit
- infrared
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 27
- 238000010025 steaming Methods 0.000 title claims abstract description 14
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 13
- 238000011068 loading method Methods 0.000 claims abstract description 21
- 239000000463 material Substances 0.000 claims abstract description 21
- 238000001514 detection method Methods 0.000 claims abstract description 19
- 238000012549 training Methods 0.000 claims abstract description 19
- 238000000605 extraction Methods 0.000 claims abstract description 13
- 230000001502 supplementing effect Effects 0.000 claims abstract description 7
- 238000007781 pre-processing Methods 0.000 claims abstract description 5
- 238000004821 distillation Methods 0.000 claims description 6
- 230000008707 rearrangement Effects 0.000 claims description 6
- 238000002372 labelling Methods 0.000 claims description 4
- 238000013528 artificial neural network Methods 0.000 claims description 2
- 230000009467 reduction Effects 0.000 claims description 2
- 239000013589 supplement Substances 0.000 claims description 2
- 230000008901 benefit Effects 0.000 abstract description 3
- 238000003825 pressing Methods 0.000 abstract description 3
- 239000010410 layer Substances 0.000 description 15
- 230000006870 function Effects 0.000 description 14
- 230000008569 process Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 7
- 238000012360 testing method Methods 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 4
- 238000012840 feeding operation Methods 0.000 description 4
- 230000007480 spreading Effects 0.000 description 4
- 238000003892 spreading Methods 0.000 description 4
- 230000017105 transposition Effects 0.000 description 4
- 238000013124 brewing process Methods 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000013526 transfer learning Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- XLOMVQKBTHCTTD-UHFFFAOYSA-N Zinc monoxide Chemical compound [Zn]=O XLOMVQKBTHCTTD-UHFFFAOYSA-N 0.000 description 1
- 230000003321 amplification Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000013078 crystal Substances 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000000855 fermentation Methods 0.000 description 1
- 230000004151 fermentation Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000004064 recycling Methods 0.000 description 1
- 238000010563 solid-state fermentation Methods 0.000 description 1
- 235000015096 spirit Nutrition 0.000 description 1
- 239000002344 surface layer Substances 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- 235000014692 zinc oxide Nutrition 0.000 description 1
- 239000011787 zinc oxide Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/10—Image acquisition
- G06V10/12—Details of acquisition arrangements; Constructional details thereof
- G06V10/14—Optical characteristics of the device performing the acquisition or on the illumination arrangements
- G06V10/143—Sensing or illuminating at different wavelengths
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1628—Programme controls characterised by the control loop
- B25J9/163—Programme controls characterised by the control loop learning, adaptive, model based, rule based expert control
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B65—CONVEYING; PACKING; STORING; HANDLING THIN OR FILAMENTARY MATERIAL
- B65G—TRANSPORT OR STORAGE DEVICES, e.g. CONVEYORS FOR LOADING OR TIPPING, SHOP CONVEYOR SYSTEMS OR PNEUMATIC TUBE CONVEYORS
- B65G69/00—Auxiliary measures taken, or devices used, in connection with loading or unloading
- B65G69/04—Spreading out the materials conveyed over the whole surface to be loaded; Trimming heaps of loose materials
- B65G69/0425—Spreading out the materials conveyed over the whole surface to be loaded; Trimming heaps of loose materials with vibrating or shaking means
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- Mechanical Engineering (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Robotics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of intelligent identification, in particular to a white spirit steaming and steam detecting method based on a convolutional neural network, which comprises the following steps: the method comprises the following steps: the method comprises the steps of firstly obtaining an infrared-like image of a retort barrel through a thermal infrared camera, finishing preprocessing of the infrared-like image, and then judging the conditions of retort loading, partial region material supplementing and whole material laying based on a Deconv-SSD model. The method has the advantages of strong generalization capability, accurate detection and classification, accurate acquisition of the feeding position and the like by using the strong feature extraction capability of the convolutional neural network and training through a large number of infrared-like images of various specific conditions, can judge three conditions of the liquor on the steamer in real time, greatly reduces the phenomena of steam pressing and steam leakage, and improves the yield and the quality of the liquor.
Description
Technical Field
The invention relates to the technical field of intelligent identification, in particular to a white spirit steaming and steam detecting method based on a convolutional neural network.
Background
Chinese white spirit is one of six distilled spirits in the world, wherein the solid state fermentation brewing process is a unique process, is a crystal of the intelligence of all Chinese workers and people, and the brewing experience of 'aroma generation by fermentation and aroma extraction by distillation' is thousands of years, wherein in the brewing process of the white spirit, the distillation process is particularly important, the yield and the quality of the white spirit mainly depend on the distillation process, and the steaming is one of the most important processes for distillation, so that the 'steam detection and steaming, thin-layer feeding' are required to be realized under specific conditions, accurate steam detection can be realized, various conditions can be identified, accurate steaming and feeding are realized, the phenomena of 'steam pressing' and 'steam leakage' are reduced to the greatest extent, and the yield and the quality of the white spirit can be effectively improved. In the process of loading into the steamer, when the liquor steam front is about to overflow the surface of the fermented grains, a layer of thin cold material is laid, so that the optimal fat dissolving condition is realized, the traditional steaming process adopts manual steaming by experience, the mode needs to consume a large amount of human resources, and has no uniform standard, different working personnel have different judgment results, and the phenomena of steam pressing and steam leakage can be caused; some researchers adopt a mode of combining digital image processing and a traditional machine learning algorithm to realize retort-loading steam detection, but the methods have the defect of poor generalization capability due to the adoption of a manually designed characteristic extraction algorithm, and the phenomenon that the results of retort-loading steam detection are different in different seasons and different factories can be caused; in addition, the retort-loading steam-detecting method of the type lacks judgment on the special condition of retort loading.
Disclosure of Invention
In order to solve the problems, the invention provides a white spirit steaming-on steam detection method based on a convolutional neural network, which has strong generalization ability, strong detection capability and low error rate.
In order to achieve the purpose, the invention adopts the technical scheme that:
a white spirit steaming and steam detecting method based on a convolutional neural network comprises the following steps: the method comprises the steps of firstly obtaining an infrared-like image of a retort barrel through a thermal infrared camera, judging retort loading, partial region material supplementing and whole material paving conditions based on a Deconv-SSD model after preprocessing the infrared-like image, specifically, obtaining the infrared-like image of the retort barrel through the thermal infrared camera, preprocessing the image, extracting characteristics by adopting a Deconv-net convolution neural network, and judging retort loading, partial region material supplementing and whole material paving conditions on the basis of an improved SSD algorithm. The whole steamer barrel is mainly operated under the two conditions of steamer feeding and whole material spreading, and the operation is realized by combining feeding equipment; for part of feeding conditions, the precise feeding is realized by accurately detecting the two-dimensional position needing feeding and combining feeding equipment. The image preprocessing comprises median filtering of the collected thermal infrared image, so that noise is effectively suppressed, image blurring is avoided, and the effects of retaining image details and reducing image boundary distortion are achieved; in addition, the method also comprises the processes of cutting, translation and brightness change of the thermal infrared image, so that the data amplification effect is realized.
Further, in the minimum unit of feature extraction, the Deconv-SSD model first uses point-by-point grouping convolution to perform dimensionality reduction on the input feature map to reduce the computation amount of deep separable convolution, then uses channel rearrangement to perform cross rearrangement on the multi-channel feature map, and finally uses the combination of the deep separable convolution and the point-by-point grouping convolution to complete feature extraction, and performs feature concatenation with the input feature map, so that there is a jump connection between each group of minimum units, and the input of each group of minimum units is the union of all previous layer outputs.
Further, the target detection loss function of the Deconv-SSD model is represented by equation (1), and is composed of a weighted sum of a position loss function and a classification loss function:
wherein: m represents the sum of positive samples; x represents the matching result of the default frame and the real frame, wherein x is 0 to represent failure, and x is 1 to represent success; c is the confidence coefficient of the softmax function for judging each category respectively; l isloc(x, L, g) is the position loss function, Lconf(x, c) a classification loss function, i is the search box sequence number; j is the real box sequence number; p is a class serial number, and p ═ 0 represents the background;
whether the IOU of the ith search box and the jth category box is larger than a threshold value, larger than 1 and not larger than 0 is represented, and at the moment, the object category in the real box is p; is a prediction box; g is a real frame; α is the weight of the position loss;
the prediction probability of the category p corresponding to the ith search box is represented as:
furthermore, the Deconv-SSD model adopts quasi-infrared images of various specific conditions (different plants, different seasons, different light rays and the like) to perform labeling supervision training, labels of data are obtained according to judgment of liquor distillation engineers with abundant experience, and labeling is completed according to a unified standard.
The invention has the following beneficial effects:
the method has the advantages of strong generalization capability, accurate detection and classification, accurate acquisition of the feeding position and the like, can judge three conditions of steaming on white spirit in real time, greatly reduces the phenomena of 'steam pressure' and 'steam leakage', and improves the yield and quality of the white spirit.
Drawings
Fig. 1 is a feature extraction minimum unit in the embodiment of the present invention.
Fig. 2 is a schematic diagram of transposed convolution and feature fusion in an embodiment of the present invention.
Fig. 3 is a Deconv-SSD model framework in an embodiment of the invention.
Fig. 4 is a flowchart illustrating the operation of the present invention.
Fig. 5 is a system layout diagram of an embodiment of the present invention.
FIG. 6 is a retort loading condition classification in an embodiment of the present invention;
in the figure: (a) feeding operation; (b) waiting for feeding into a steamer; (c) and paving the whole layer.
FIG. 7 is a graph of model training loss functions in an embodiment of the present invention.
FIG. 8 is a graph of the average IOU of the training model in an embodiment of the present invention
FIG. 9 shows the results of the test conducted in the experimental example of the present invention.
Detailed Description
The present invention will be described in detail with reference to specific examples. The following examples will assist those skilled in the art in further understanding the invention, but are not intended to limit the invention in any way. It should be noted that variations and modifications can be made by persons skilled in the art without departing from the spirit of the invention. All falling within the scope of the present invention.
The invention provides a Deconv-SSD model by taking advantage of a Densenet network structure, and firstly uses point-by-point grouping convolution to reduce the dimension of an input feature map in a minimum unit of feature extraction so as to reduce the calculated amount of deep separable convolution. And then, carrying out cross rearrangement on the multi-channel feature map by utilizing channel rearrangement, finally completing feature extraction by utilizing the combination of depth separable convolution and point-by-point grouping convolution, and carrying out feature splicing with the input feature map. Thus, there is a jump connection between each group of minimum units, and the input of each group of minimum units is the union of all the previous layer outputs. The feature graph learned by the minimum unit can be directly transmitted to all the following minimum units as input to realize feature recycling, the improved feature extraction minimum unit is shown in figure 1, the improved network structure low-layer high-resolution feature graph has more global information and stronger fitting capability, and meanwhile, the high-layer feature fitting capability is unchanged, so that the problem of overfitting cannot be caused. The improved network structure low-layer high-resolution characteristic diagram has more global information and stronger fitting capability, and meanwhile, the fitting capability of the high-layer characteristic is unchanged, so that the problem of overfitting is avoided. Meanwhile, in order to reduce the omission ratio of small targets, the method introduces the transposition convolution and feature fusion. As shown in fig. 2.
The SSD algorithm is found to have better detection capability for medium-sized and large-sized targets and weaker detection capability for small-sized targets by consulting documents, so the invention improves the low-layer characteristic diagram to avoid increasing the calculation amount by introducing excessive transposition convolution. The image resolution of the input convolution model is set to a scale of 300 × 300, and in addition, feature maps having resolutions of 19 × 19 and 10 × 10 are subjected to transposed convolution, respectively. After the low-resolution features are transposed and convolved, the feature graph merging can be completed only if the feature graph resolution is consistent with the high-resolution feature graph. Combining the actual feature diagram resolution condition of the SSD network structure, the invention adopts the convolution kernel of 2 × 2 and the transposition convolution parameter of 2 step length for the feature diagram with 19 × 19 resolution, and adopts the convolution kernel of 3 × 3 and 2 step length for the transposition convolution parameter of the feature diagram with 10 × 10 resolution, thereby realizing the purpose of edge extension. The minimum unit for light weight feature extraction is used in the feature extraction part, so that the problem of increased calculation amount caused by introducing a transposed convolution structure can be solved. Fig. 3 is an improved model proposed in connection with the present invention based on the SSD model.
Deconv-SSD modelThe target detection loss function of (2) is expressed by the formula (1), and is composed of position loss and classification loss weighted summation, and training and testing in the experimental process of the invention are carried out according to the function. The invention needs to accurately judge the retort loading area, and 2 times of weight is set, wherein: m represents the sum of positive samples; x represents the matching result of the default frame and the real frame, wherein x is 0 to represent failure, and x is 1 to represent success; c is the confidence coefficient of the softmax function for judging each category respectively; l isloc(x, L, g) is the position loss function, Lconf(x, c) classification loss functions, respectively expressed as:
wherein: i is the search box sequence number; j is the real box sequence number; p is a class serial number, and p ═ 0 represents the background;
whether the IOU of the ith search box and the jth category box is larger than a threshold value, larger than 1 and not larger than 0 is represented, and at the moment, the object category in the real box is p; l is a prediction box; g is a real frame; α is the weight of the position loss;
the prediction probability of the category p corresponding to the ith search box is represented as:
the operation process of the whole algorithm is shown in fig. 4, the infrared-like image is firstly obtained on the surface of the whole fermented grain, after pretreatment, classified judgment of judgment conditions of retort loading, partial region material supplement and whole material paving conditions is carried out based on a Deconv-SSD model, so that different steam detection retort loading conditions are judged in real time, and accurate retort loading operation is realized according to the conditions.
As shown in figure 5, the thermal images of the surface of the fermented grains are captured by the thermal infrared imager, and then the thermal images are used for judging the retort loading condition through a convolutional neural network model of a server side.
FIG. 6 is a classification of retort loading cases, in which (a) is a feeding operation, (b) is a waiting retort loading operation, and (c) is a full-layer spreading operation, in which infrared rays are represented by a white hot area and image enhancement is represented by a dark state, indicating that wine steam does not reach surface fermented grains; the material supplementing operation is a region with a small amount of white heat shown by a heat map, which indicates that the wine steam in a local region rises faster, and the phenomenon of steam leakage can be effectively avoided after the material supplementing operation is carried out in time; the whole layer of spreading operation shows that a large amount of white hot areas exist, and the wine steam reaches fermented grains on the surface layer.
Examples of the experiments
In the experiment, a thermal infrared image of the retort in the winery is made to serve as a training data set of a model, the data set is shown in table 1, the data set comprises three parts, namely a training set, a verification set and a test set, the three parts are distributed according to the proportion of 8: 1, and weight parameter optimization training of the model and model verification and testing of a certain training stage are mainly completed. The invention has three categories for liquor steaming and steam detection, namely waiting for steaming, material spreading in the whole layer and material supplementing operation, training is carried out by adopting a supervision and learning mode, and acquired images are classified and marked with the assistance of a brewer.
TABLE 1 data set Allocation
The invention uses a transfer learning method under a Deconv-net convolution neural model, after improving part of structural layers, the data set manufactured by the invention is used for training, the experiment sets batch training data to be 32, the weight attenuation value (Decay) to be 0.0005, the training momentum is configured to be 0.9, the whole training process is completed by 40000 steps in 200 epochs, the experiment adopts a small batch gradient descent method to optimize parameters, the learning rate of the initial training is 0.0001, after 7000 times of training, the learning rate is set to be 0.00003, and when the training reaches 11000 times, the learning rate is set to be 0.00001. Fig. 7 is a graph of variation of loss function values in model training, and it can be seen that, after the transfer learning is adopted, the loss function of the model drops very quickly, and after the model iterates to 40000 times, the model basically stays close to 0.5, wherein the lower the loss value, the closer the result predicted by the model is to the real result. FIG. 8 is a graph of the average IOU predicted by the model, showing the intersection ratio of the predicted target position to the true target position, with a larger ratio indicating a better prediction of the effectiveness of the model; it can be seen from the figure that the IOU value of the model prediction effect is close to 1 after 40000 iterations.
The trained model is verified by adopting the test data set, the average accuracy of the model reaches 86.7%, the time spent on detecting one thermal infrared image is only 0.2s, the real-time effect is basically achieved, and the field deployment of a factory is facilitated. FIG. 8 shows the calculation output result of the liquor steam-detecting retort-loading algorithm, and therefore the method can accurately realize the identification and classification of three types of retort-loading conditions and can calculate the position of a steam-detecting area. It is worth noting that for the two conditions of whole layer material laying and waiting for feeding into a steamer, the algorithm does not need to output position information, because the two conditions need material laying operation on the whole fermented grain surface, and the calculation of a specific position is not needed; for the feeding operation, the algorithm can calculate its specific position, fig. 9 is the judgment of the feeding region, where there are two regions requiring feeding operation a and B, their position information is (x1, y1, w1, h1) and (x2, y2, w2, h2), respectively, for the a feeding region, x1, y1 are the coordinates of the a point at the upper left corner of the target box, w1 and h1 are the lengths of ac and bc, i.e. the width and height of the target; meanwhile, the whole paving area is established by adopting a uniform two-dimensional coordinate system.
The foregoing description of specific embodiments of the present invention has been presented. It is to be understood that the present invention is not limited to the specific embodiments described above, and that various changes or modifications may be made by one skilled in the art within the scope of the appended claims without departing from the spirit of the invention. The embodiments and features of the embodiments of the present application may be combined with each other arbitrarily without conflict.
Claims (5)
1. A white spirit steaming and steam detecting method based on a convolutional neural network is characterized by comprising the following steps: the method comprises the following steps: the method comprises the steps of firstly obtaining an infrared-like image of a retort barrel through a thermal infrared camera, finishing preprocessing of the infrared-like image, and then judging the conditions of retort loading, partial region material supplementing and whole material laying based on a Deconv-SSD model.
2. The white spirit steaming-on steam detection method based on the convolutional neural network as claimed in claim 1, characterized in that: the characteristics are extracted by adopting a Deconv-net convolution neural network, the judgment of the situations of retort loading, partial region material supplement and whole material paving is realized on the improved SSD algorithm, the whole retort barrel is mainly operated under the two situations of retort loading and whole material paving, and the operation is realized by combining feeding equipment; for part of feeding conditions, the precise feeding is realized by accurately detecting the two-dimensional position needing feeding and combining feeding equipment.
3. The white spirit steaming-on steam detection method based on the convolutional neural network as claimed in claim 1, characterized in that: in the minimum unit of feature extraction, the Deconv-SSD model firstly uses point-by-point grouping convolution to perform dimensionality reduction on an input feature map so as to reduce the calculated amount of deep separable convolution, then uses channel rearrangement to perform cross rearrangement on a multi-channel feature map, and finally uses the combination of the deep separable convolution and the point-by-point grouping convolution to complete feature extraction and performs feature splicing with the input feature map, so that jump connection exists between each group of minimum units, and the input of each group of minimum units is the union of all previous layer outputs.
4. The white spirit steaming-on steam detection method based on the convolutional neural network as claimed in claim 1, characterized in that: the target detection loss function of the Deconv-SSD model is represented by the formula (1) and is composed of a position loss function and a classification loss function through weighted summation:
wherein: m represents the sum of positive samples; x represents the matching result of the default frame and the real frame, wherein x is 0 to represent failure, and x is 1 to represent success; c is the confidence coefficient of the softmax function for judging each category respectively; l isloc(x, L, g) is the position loss function, Lconf(x, c) a classification loss function, i is the search box sequence number; j is the real box sequence number; p is a class serial number, and p ═ 0 represents the background;
whether the IOU of the ith search box and the jth category box is larger than a threshold value, larger than 1 and not larger than 0 is represented, and at the moment, the object category in the real box is p; l is a prediction box; g is a real frame; α is the weight of the position loss;
the prediction probability of the category p corresponding to the ith search box is represented as:
5. the white spirit steaming-on steam detection method based on the convolutional neural network as claimed in claim 1, characterized in that: the Deconv-SSD model adopts infrared-like images of various specific conditions to perform labeling supervision training, labels of data are obtained according to judgment of liquor distillation engineers with abundant experience, and labeling is completed according to unified standards.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010787213.9A CN111985463A (en) | 2020-08-07 | 2020-08-07 | White spirit steaming and steam detecting method based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010787213.9A CN111985463A (en) | 2020-08-07 | 2020-08-07 | White spirit steaming and steam detecting method based on convolutional neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111985463A true CN111985463A (en) | 2020-11-24 |
Family
ID=73445257
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010787213.9A Pending CN111985463A (en) | 2020-08-07 | 2020-08-07 | White spirit steaming and steam detecting method based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111985463A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114170173A (en) * | 2021-11-30 | 2022-03-11 | 中科九创智能科技(北京)有限公司 | Detection method, detection module and detection system of retort loading robot |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108288075A (en) * | 2018-02-02 | 2018-07-17 | 沈阳工业大学 | A kind of lightweight small target detecting method improving SSD |
| CN110070072A (en) * | 2019-05-05 | 2019-07-30 | 厦门美图之家科技有限公司 | A method of generating object detection model |
| CN110084253A (en) * | 2019-05-05 | 2019-08-02 | 厦门美图之家科技有限公司 | A method of generating object detection model |
| CN110084312A (en) * | 2019-05-05 | 2019-08-02 | 厦门美图之家科技有限公司 | A method of generating object detection model |
| CN111260630A (en) * | 2020-01-16 | 2020-06-09 | 高新兴科技集团股份有限公司 | Improved lightweight small target detection method |
-
2020
- 2020-08-07 CN CN202010787213.9A patent/CN111985463A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108288075A (en) * | 2018-02-02 | 2018-07-17 | 沈阳工业大学 | A kind of lightweight small target detecting method improving SSD |
| CN110070072A (en) * | 2019-05-05 | 2019-07-30 | 厦门美图之家科技有限公司 | A method of generating object detection model |
| CN110084253A (en) * | 2019-05-05 | 2019-08-02 | 厦门美图之家科技有限公司 | A method of generating object detection model |
| CN110084312A (en) * | 2019-05-05 | 2019-08-02 | 厦门美图之家科技有限公司 | A method of generating object detection model |
| CN111260630A (en) * | 2020-01-16 | 2020-06-09 | 高新兴科技集团股份有限公司 | Improved lightweight small target detection method |
Non-Patent Citations (1)
| Title |
|---|
| 田万春 等: "基于支持向量机的白酒上甑探汽方法研究", 《食品与机械》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114170173A (en) * | 2021-11-30 | 2022-03-11 | 中科九创智能科技(北京)有限公司 | Detection method, detection module and detection system of retort loading robot |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109934200B (en) | RGB color remote sensing image cloud detection method and system based on improved M-Net | |
| CN112215819B (en) | Airport pavement crack detection method based on depth feature fusion | |
| CN104050471B (en) | Natural scene character detection method and system | |
| CN111179217A (en) | Attention mechanism-based remote sensing image multi-scale target detection method | |
| CN105825511A (en) | Image background definition detection method based on deep learning | |
| CN104517122A (en) | Image target recognition method based on optimized convolution architecture | |
| CN110766020A (en) | System and method for detecting and identifying multi-language natural scene text | |
| CN109635726B (en) | Landslide identification method based on combination of symmetric deep network and multi-scale pooling | |
| CN107545571A (en) | A kind of image detecting method and device | |
| CN111046928B (en) | Single-stage real-time universal target detector and method with accurate positioning | |
| CN105405138A (en) | Water surface target tracking method based on saliency detection | |
| CN113297988A (en) | Object attitude estimation method based on domain migration and depth completion | |
| CN115527234A (en) | Infrared image cage dead chicken identification method based on improved YOLOv5 model | |
| CN115410081A (en) | Multi-scale aggregated cloud and cloud shadow identification method, system, equipment and storage medium | |
| CN114926826A (en) | Scene text detection system | |
| CN111985463A (en) | White spirit steaming and steam detecting method based on convolutional neural network | |
| Dong et al. | Field-matching attention network for object detection | |
| CN112884741B (en) | Printing apparent defect detection method based on image similarity comparison | |
| CN116805360B (en) | Obvious target detection method based on double-flow gating progressive optimization network | |
| CN111368637B (en) | Transfer robot target identification method based on multi-mask convolutional neural network | |
| Geng et al. | DPSA: dense pixelwise spatial attention network for hatching egg fertility detection | |
| CN117218101A (en) | Composite wind power blade defect detection method based on semantic segmentation | |
| CN116503354A (en) | Method and device for detecting and evaluating hot spots of photovoltaic cells based on multi-mode fusion | |
| CN116309398A (en) | Printed circuit board small target defect detection method based on multi-channel feature fusion learning | |
| CN114529766A (en) | Heterogeneous source SAR target identification method based on domain adaptation |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20201124 |
|
| RJ01 | Rejection of invention patent application after publication |