CN111291779A - Vehicle information identification method and system, memory and processor - Google Patents
Vehicle information identification method and system, memory and processor Download PDFInfo
- Publication number
- CN111291779A CN111291779A CN201811500432.3A CN201811500432A CN111291779A CN 111291779 A CN111291779 A CN 111291779A CN 201811500432 A CN201811500432 A CN 201811500432A CN 111291779 A CN111291779 A CN 111291779A
- Authority
- CN
- China
- Prior art keywords
- vehicle information
- data set
- vehicle
- information data
- image
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 61
- 238000012549 training Methods 0.000 claims abstract description 32
- 238000012545 processing Methods 0.000 claims description 31
- 238000007781 pre-processing Methods 0.000 claims description 17
- 238000003860 storage Methods 0.000 claims description 6
- 230000003068 static effect Effects 0.000 claims description 4
- 238000001514 detection method Methods 0.000 abstract description 9
- 230000000694 effects Effects 0.000 abstract description 6
- 238000012360 testing method Methods 0.000 abstract description 3
- 238000005070 sampling Methods 0.000 description 5
- 241001455214 Acinonyx jubatus Species 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 241000557133 Hunnemannia fumariifolia Species 0.000 description 1
- 240000002836 Ipomoea tricolor Species 0.000 description 1
- 241000533950 Leucojum Species 0.000 description 1
- 240000002853 Nelumbo nucifera Species 0.000 description 1
- 235000006508 Nelumbo nucifera Nutrition 0.000 description 1
- 235000006510 Nelumbo pentapetala Nutrition 0.000 description 1
- 241000282376 Panthera tigris Species 0.000 description 1
- 241000270295 Serpentes Species 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000007499 fusion processing Methods 0.000 description 1
- 239000009731 jinlong Substances 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000000843 powder Substances 0.000 description 1
- OLBCVFGFOZPWHH-UHFFFAOYSA-N propofol Chemical compound CC(C)C1=CC=CC(C(C)C)=C1O OLBCVFGFOZPWHH-UHFFFAOYSA-N 0.000 description 1
- 229960004134 propofol Drugs 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 239000010193 shuanglong Substances 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 239000009137 wuling Substances 0.000 description 1
- -1 xiali Substances 0.000 description 1
Images
Classifications
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a vehicle information identification method, a vehicle information identification system, a memory and a processor. The vehicle information identification method comprises the steps of establishing a vehicle information data set; training the vehicle information data set based on the yolov3 network to obtain a multi-attribute network model capable of identifying the vehicle information; and performing multi-attribute prediction on the vehicle information of the image to be recognized by using the obtained multi-attribute network model so as to recognize the vehicle structural information in the image to be recognized. On the basis of yolov3, set up the network parameter that influences network model performance for convergence speed is faster, network model's effect is more excellent, directly resolves vehicle information itself into every independent class, realizes end-to-end input and output, makes the network output be the structured information of every vehicle, thereby vehicle information discernment testing process consuming time fewly, and recognition efficiency is high, generally is applicable to the vehicle detection field.
Description
[ technical field ] A method for producing a semiconductor device
The invention relates to the technical field of information identification, in particular to a vehicle information identification method, a vehicle information identification system, a memory and a processor.
[ background of the invention ]
At present, a video monitoring vehicle information system is used as an important component of intelligent security and intelligent traffic in the application of the internet of things facing urban public safety comprehensive management, and faces a great challenge of deep application. The main application bottleneck is how to rapidly and efficiently provide vehicle structural information in the video.
In a video monitoring vehicle information system implemented by a conventional method, although a convolutional neural network method is applied, different input features including shallow features, deep features, local features and the like can be extracted and classified according to feature information. At present, the method has a good effect on image classification, but has an improved space for the real-time performance and accuracy of fine-grained classification, can not uniformly extract vehicle information, needs to analyze and fuse information for many times, and is complex in process and long in time consumption.
[ summary of the invention ]
The technical problem to be solved by the invention is to provide a vehicle information identification method, a system, a memory and a processor, which can directly analyze the information of a vehicle into each independent class on the basis of yolov3, realize end-to-end input and output, and enable the network output to be the structural information of the vehicle, so that the time consumption of the vehicle information identification detection process is less, and the identification efficiency is high.
To solve the foregoing technical problem, in one aspect, an embodiment of the present invention provides a vehicle information identification method, including: establishing a vehicle information data set; setting network parameters based on yolov 3; training the vehicle information data set based on the yolov3 network to obtain a multi-attribute network model capable of identifying the vehicle information; and performing multi-attribute prediction on the vehicle information of the image to be recognized by using the obtained multi-attribute network model so as to recognize the vehicle structural information in the image to be recognized.
Preferably, the creating the vehicle information data set includes: preprocessing the vehicle information data set; and performing data enhancement processing on the vehicle information data set.
Preferably, the setting is based on yolov3 network parameters including: setting a class number defined for the vehicle information data set based on a class parameter of the yolov3 network; the filter parameters based on yolov3 network were set to 3 × (class parameters + 5).
Preferably, the training of the vehicle information data set based on the yolov3 network to obtain the multi-attribute network model capable of identifying the vehicle information comprises: carrying out category calibration on the vehicle information data set by using a marking tool to generate a calibration file in an xml format; converting the calibration file in the xml format into a txt file, and generating the txt file for storing the path of each image in the vehicle information data set; and setting and changing network super parameters according to the categories in the vehicle information data set.
Preferably, the performing multi-attribute prediction on the vehicle information of the image to be recognized by using the obtained multi-attribute network model includes: and inputting the video resources or the static images containing the vehicle information into the trained network model, and outputting the structural information of the vehicle information at one time.
Preferably, the preprocessing the vehicle information data set includes: and counting the vehicle information data sets, and storing the vehicle information data sets into different folders according to different categories.
Preferably, the data enhancement processing of the vehicle information data set includes: and turning the images of the vehicle information data set, and saving the corresponding turned images under the same folders of the images of the vehicle information data set in an additional name.
Preferably, the data enhancement processing of the vehicle information data set includes: and carrying out mirror image processing on the images of the vehicle information data set, and saving the corresponding mirror image images under the same folders of the images of the vehicle information data set in an additional name mode.
Preferably, the angle of the turning process is-15 to 15 °.
Preferably, the converting the calibration file in the xml format into a corresponding txt format file and generating a txt file storing a path on which each image in the vehicle information data set is located includes: reading a calibration file in an xml format to obtain the width and the height of an image containing the vehicle data set, and the width and the height and the category of the vehicle contained in the image of the vehicle data set; normalizing the width and height of the vehicle contained in the image of the vehicle data set; calculating vehicle center point coordinates contained in an image of the vehicle data set; storing the category serial number, the central point coordinate and the width and height information of the vehicle in the image of the vehicle data set as a txt text file; generating a txt text file storing the path of the vehicle information data set.
Preferably, the txt text file is saved in a tab folder.
Preferably, the normalizing the width and height of the vehicle included in the image of the vehicle data set means: and converting the aspect ratio of the vehicle contained in the image of the vehicle data set into a numerical value between 0 and 1.
In another aspect, an embodiment of the present invention provides a storage medium including a stored program, wherein the program is executed to perform the above-described vehicle information identification method.
In another aspect, an embodiment of the present invention provides a processor, configured to execute a program, where the program executes the vehicle information identification method described above.
In another aspect, an embodiment of the present invention provides a vehicle information identification system, including: the system comprises a vehicle information data set setting module, a yolov 3-based network parameter setting module, a yolov 3-based network training module and a yolov 3-based network prediction module which are electrically connected; the vehicle information data set setting module is used for establishing a vehicle information data set; the yolov 3-based network parameter sending module is used for setting yolov 3-based network parameters; the yolov 3-based network training module is used for training the vehicle information data set to obtain a multi-attribute network model capable of identifying the vehicle information; the yolov 3-based network prediction module is used for performing multi-attribute prediction on the vehicle information of the image to be recognized by using the obtained multi-attribute network model so as to recognize the vehicle structural information in the image to be recognized.
Preferably, the vehicle information data set setting module further comprises a data preprocessing module and a data enhancement processing module, wherein the data preprocessing module is used for preprocessing the vehicle information data set; the data enhancement processing module is used for carrying out data enhancement processing on the vehicle information data set.
Preferably, the yolov 3-based network parameter setting module sets the number of categories defined for the vehicle information dataset based on the category parameters in the yolov3 network; set filter parameters based on yolov3 network parameters to 3 × (class parameters + 5).
Preferably, the data preprocessing module counts the vehicle information data sets and stores the vehicle information data sets in different folders according to different categories.
Preferably, the data enhancement processing module performs flipping processing on the image of the vehicle information data set, and saves the corresponding flipped image in a folder with the same image of the vehicle information data set.
Preferably, the data enhancement processing module performs mirror image processing on the image of the vehicle information data set, and saves the corresponding mirror image under a folder with the same image of the vehicle information data set.
Compared with the prior art, the technical scheme has the following advantages: on the basis of yolov3, set up the network parameter that influences network model performance for convergence speed is faster, network model's effect is more excellent, directly resolves vehicle information itself into every independent class, realizes end-to-end input and output, makes the network output be the structural information of vehicle, thereby vehicle information discernment testing process consuming time is few, and the recognition efficiency is high, generally is applicable to the vehicle detection field.
[ description of the drawings ]
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without inventive labor.
Fig. 1 is a flowchart of a vehicle information recognition method of the present invention.
Fig. 2 is a flowchart of a preferred embodiment of the vehicle information recognition method of the present invention.
Fig. 3 is a structural view of a vehicle information recognition system of the present invention.
[ detailed description ] embodiments
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by one of ordinary skill in the art from the embodiments given herein without making any creative effort, shall fall within the scope of the present invention.
yolov3 describes the input, structure, and output of the network primarily from three points.
(1) Network input: the size of the net input is typically 320 x 320, 416 x 416, 608 x 608. This size must be an integer multiple of 32, yolov3 has 5 lower hits, each with a step size of 2, so the maximum stride of the network (stride refers to the input size of the layer divided by the output) is 2^5 ^ 32.
(2) The network structure is as follows: typically, a darknet-53 is trained, which is mainly for two purposes: a. the network structure can have good classification results in ImageNet, thereby showing that the network can learn good characteristics (design a new network structure, which is equivalent to adjusting parameters), and b. Darknet-53 As opposed to ResNet-152 and ResNet-101, Darknet-53 not only provides a much better classification accuracy, but also is much faster to compute than ResNet-152 and ResNet-101, and has fewer network layers.
Darknet-53 adopts a layer-hopping connection mode of ResNet, and the performance is completely better than that of two deep networks of ResNet-152 and ResNet-101, because: differences in the basic elements of the network; the fewer the number of network layers, the fewer the parameters, and the less the amount of computation required.
The yolov3 network used the first 52 layers of darknet-53 (no fully connected layers). yolov3 is a full convolution network, largely using residual jump layer connections. In previous work, the sampling was generally performed by down-sampling using max-posing or average-posing with a size of 2 x 2 and a step size (stride) of 2. But in this network structure a convolution with step size 2 is used for down-sampling. Meanwhile, the network uses up-sampling and route operation, and 3 times of detection are carried out in one network structure.
Benefits of the structure using residuals: whether the depth model can be normally converged is a key point of the depth model, the structure of the residual error can ensure that the network structure can still be converged under the condition of very deep, and the model can be trained; the deeper the network, the better the expressed characteristics, and the improved classification + detection effects; 1-1 convolution in residual error greatly reduces channel of each convolution by using the idea of network in network, on one hand, reduces parameter quantity (the larger the parameter quantity is, the larger the stored model is), on the other hand, reduces calculation quantity to a certain extent.
(3) And (3) network output:
a. firstly, the size of the network output characteristic layer is determined. For example, if the input is 320 × 320, the output is 320/32 × 10, and thus the output is a feature map (feature map) of 10 × 10, where 10 × 10 — 100 cells exist; similarly, when the input is 416 × 416, the output feature layer is 13 × 13, and 13 × 13 is 169 cells; when the input is 608 × 608, the output feature map size is 19 × 19, and the cell has 19 × 19 — 361 cells. The output feature layer is doubled for each up-sample.
Determination of Anchor box. This prior box is different from the manual setting of the previous fast-Rcnn and SSD, both in yolov2 and yolov3, which employ k-means clustering of objects in the image.
Each cell in feature map predicts 3 bounding boxes (bounding boxes), each of which predicts three things: (1) the position of each box (4 values, center coordinates tx and ty, height bh and width bw of the box), (2) one object prediction, (3) N categories, coco dataset 80 category, voc20 category. Thus for the coco dataset, at network input 416 × 416, the output size of the network is 13 × 13(3 × 4+1+80)) -43095
yolov3 uses sigmoid function for center coordinate prediction. This results in an output value between 0 and 1. In market situations, YOLO does not predict the exact coordinates of the bounding box center. It predicts that: an offset associated with the upper left corner of the grid cell of the prediction target; and normalized using the cell size in feature map.
When the input image is 416 x 416, if the prediction of the center is (0.4, 0.7), the relative coordinates of the second cell on the 13x13 feature map are (1.4, 1.7), the specific location x coordinate also needs to be 1.4 times the width of the cell, and the y coordinate is 1.7 times the height of the cell.
Width bw and height bh of Bounding box
The predictions bw and bh from yolov3 are normalized using the height and width of the image, the predictions bx and by of the box are (0.3, 0.8), then the actual width and height of the 13x13 feature map is (13x0.3, 13x 0.8).
d. Three times of detection, each time corresponding to different feelings, the feeling of 32 times of down sampling is the largest, and the method is suitable for detecting a large target, so that when the input is 416 × 416, three anchors box of each cell are (116, 90); (156, 198); (373, 326). 16 times fit into a general size object, and the anchor box is (30, 61); (62, 45); (59, 119). The 8-fold minimum sensitivity is suitable for detecting small targets, so the anchor box is (10, 13); (16, 30); (33, 23). Therefore, when the input is 416 × 416, there are 10647 total propofol boxes in practice (52 × 52+26 +13 × 3).
Example one
Fig. 1 is a flowchart of a vehicle information recognition method of the present invention. As shown in fig. 1, a vehicle information identification method includes the steps of:
s11, establishing a vehicle information data set;
s12, setting the network parameters based on yolov 3;
s13, training the vehicle information data set based on the yolov3 network to obtain a multi-attribute network model capable of identifying vehicle information;
and S14, performing multi-attribute prediction on the vehicle information of the image to be recognized by using the obtained multi-attribute network model so as to recognize the vehicle structural information in the image to be recognized.
The method comprises the steps of collecting a certain number of vehicle information samples, establishing a vehicle information data set, setting network parameters based on yolov3, and training the collected vehicle information samples. And obtaining a multi-attribute network model capable of identifying the vehicle attributes according to the training result. And identifying the structural information of the vehicle information of the image to be identified at one time, such as attributes of a vehicle type, a afternoon lot, a vehicle body color and the like.
By adopting the vehicle information identification method, the structured information of each vehicle can be identified at one time by a large amount of vehicle information through network training and an end-to-end mode, and the identification speed and the efficiency are high.
Example two
Fig. 2 is a flowchart of a preferred embodiment of the vehicle information recognition method of the present invention. As shown in fig. 2, in a specific implementation, the step S11 of creating the vehicle information data set includes: preprocessing the vehicle information data set; and carrying out data enhancement processing on the vehicle information data set. The vehicle information data set may be a video from a different vehicle or may be a collection of still images containing a different vehicle. Since the number of vehicle samples used for training is small and the distribution is not uniform, data enhancement processing needs to be performed on the vehicle information data set, so that the amount of the vehicle samples used for training can be increased, and the training result is more reliable.
And preprocessing the vehicle information data set, such as counting the vehicle information data set, and storing the vehicle information data set into different folders according to different categories. And performing data enhancement processing on the vehicle information data set, for example, sequentially turning over certain images of the vehicle information data set, and saving the corresponding turned-over images in a folder with the same images of the vehicle information data set. In order to ensure the image inversion rationality, the inversion process angle may be set to-15 ° to 15 °. Or carrying out mirror image processing on certain images of the vehicle information data set, and saving the corresponding mirror image under the same folder of the certain images of the vehicle information data set. The angle transformation is more rodlike for the samples with less image quantity, so that the effects of balanced and diversified stems are achieved. The flipping or mirroring process is generally performed according to the number of test bars and the specific size of the vehicle information data set image.
The input process of network training comprises the following steps:
because the method is realized based on a supervised learning method, a corresponding label needs to be made on the sample.
The first step is as follows: all samples in the folder vehicle information dataset image were calibrated with the calibration tool labelImg. All categories of vehicle structural information such as vehicle types, noon copies, vehicle body colors and the like are independently attributed to one category, and the categories are mutually independent. After the sample is calibrated, a corresponding calibration file in the xml format is generated.
The second step is that: and converting the generated calibration files in the xml format into txt files in a one-to-one correspondence manner. The method comprises the steps of reading information in a calibration file in an xml format, such as the width and the height of an image of a vehicle data set, the width and the height of a vehicle contained in the image of the vehicle data set and the category of the vehicle contained in the image of the vehicle data set, normalizing the width and the height of the vehicle contained in the image of the vehicle data set, such as setting the width and the height ratio of the vehicle contained in the image of the vehicle data set to be between 0 and 1, calculating the coordinates of a center point of the vehicle, and storing the coordinates of the center point and the width as a txt file according to the category serial number, the coordinates of the vehicle and the width and the height of the vehicle contained in the image of the vehicle data set. And converting the format of the xml format file into a txt format file, and realizing the result through software codes.
The third step: a txt text file is generated that stores the paths of all samples as input for network training.
Setting network parameters:
the parameters based on the detection layer in the yolov3 network model need to be defined according to the number of categories of network input. If the class number changes, the number of the corresponding filters of the network detection layer is correspondingly changed. In addition, the learning rate has a profound effect on the performance of the network model.
The first step is as follows: the class parameter in the network is equal to the number of classes defined by all samples.
The second step is that: the filters (filters) in the network parameters are calculated according to a formula filter (filters) × (class parameters + 5). This setup is according to yolov3 based network recommendations.
The third step: according to experimental summary, the learning rate directly affects the performance of the network model. The super-parameter learning rate is adjusted through multiple experiments, for example, the learning rate value can be continuously tried and can be: 0.001, 0.0001, 0.00001 and the like, and finally selecting a proper numerical value according to training requirements, so that the network training can be converged quickly, and the network model has better performance.
And (3) attribute prediction:
and performing multi-attribute prediction on the vehicle information of the image to be recognized by using the obtained multi-attribute network model, inputting any vehicle video into the multi-attribute network model, and outputting structural information such as the vehicle type, the afternoon number, the vehicle body color and the like of the vehicle to be predicted.
By adopting the vehicle information identification method, the information of the vehicle is directly analyzed into each independent class on the basis of yolov3, end-to-end input and output are realized, and network output is the structural information of the vehicle, so that the time consumption of the vehicle information identification and detection process is less, and the identification efficiency is high.
EXAMPLE III
The embodiment of the invention also provides a storage medium, which comprises a stored program, wherein the program executes the vehicle information identification method flow when running.
Alternatively, in the present embodiment, the storage medium described above may be configured to store program codes for executing the following vehicle information identification method flow:
s11, establishing a vehicle information data set;
s12, setting the network parameters based on yolov 3;
s13, training the vehicle information data set based on the yolov3 network to obtain a multi-attribute network model capable of identifying vehicle information;
and S14, performing multi-attribute prediction on the vehicle information of the image to be recognized by using the obtained multi-attribute network model so as to recognize the vehicle structural information in the image to be recognized.
Optionally, in this embodiment, the storage medium may include, but is not limited to: various media capable of storing program codes, such as a usb disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic disk, or an optical disk.
Therefore, by adopting the storage medium, the program running speed of the built-in vehicle information identification method flow is higher, and the identification of all vehicle information is completed quickly and efficiently.
Example four
Embodiments of the present invention also provide a processor, configured to run a program, where the program executes to perform the steps in the vehicle information identification method.
Optionally, in this embodiment, the program is configured to perform the following steps:
s11, establishing a vehicle information data set;
s12, setting the network parameters based on yolov 3;
s13, training the vehicle information data set based on the yolov3 network to obtain a multi-attribute network model capable of identifying vehicle information;
and S14, performing multi-attribute prediction on the vehicle information of the image to be recognized by using the obtained multi-attribute network model so as to recognize the vehicle structural information in the image to be recognized.
Optionally, for a specific example in this embodiment, reference may be made to the above-described embodiment and examples described in the specific implementation, and details of this embodiment are not described herein again.
Therefore, by adopting the processor, the program running speed of the built-in vehicle information identification method flow is higher, and the identification of all vehicle information is completed quickly and efficiently.
EXAMPLE five
Fig. 3 is a structural view of a vehicle information recognition system of the present invention. As shown in fig. 3, a vehicle information recognition system includes: the system comprises a vehicle information data set setting module, a yolov 3-based network parameter setting module, a yolov 3-based network training module and a yolov 3-based network prediction module which are electrically connected;
the vehicle information data set setting module is used for establishing a vehicle information data set;
the yolov 3-based network parameter setting module is used for setting yolov 3-based network parameters;
the yolov 3-based network training module is used for training a vehicle information data set to obtain a multi-attribute network model capable of identifying vehicle information;
the yolov 3-based network prediction module is used for performing multi-attribute prediction on the vehicle information of the image to be recognized by using the obtained multi-attribute network model so as to recognize the vehicle structural information in the image to be recognized.
In specific implementation, the vehicle information data set setting module further comprises a data preprocessing module and a data enhancement processing module, wherein the data preprocessing module is used for preprocessing the vehicle information data set, and storing the vehicle information data set into different folders according to different categories. The data enhancement processing module is used for performing data enhancement processing on the vehicle information data set, for example, turning over an image of the vehicle information data set, and saving the corresponding turned-over image in a folder with the same picture of the vehicle information data set. In consideration of the effectiveness of the data enhancement process, the angle of the flip process may be set to-15 ° to 15 ° as necessary. The data enhancement processing module can also perform mirror image processing on the images of the vehicle information data set, and the corresponding mirror images are saved under the folders with the same images of the vehicle information data set in an additional name. If the number of the vehicle information data sets is n (n is greater than 1), half of the data needs to be subjected to inversion processing, and the other half of the data needs to be subjected to mirror image processing, the number of the finally formed vehicle data sets is 2n, so that the number of samples for training can be increased, and the subsequent network training precision is higher.
The yolov 3-based network parameter setting module sets the number of categories defined for all samples based on the class parameters in the yolov3 network; setting the filter parameters in the network parameters based on yolov3 to be 3x (class parameters + 5); according to the results of multiple experiments, the learning rate parameter based on the yolov3 network parameters is set.
Carrying out category calibration on a vehicle information data set by using a marking tool in the training process of the vehicle information data set based on a yolov3 network training module to generate a calibration file in an xml format; converting the calibration file in the xml format into a corresponding txt file, and generating a txt text file for storing each image path in the vehicle information data set; and setting and changing network super parameters according to the categories in the vehicle information data set. The converting of the calibration file in the xml format into a corresponding txt file and the generating of the txt text file storing each image path in the vehicle information data set include: reading a calibration file in an xml format to obtain the width and the height of an image containing the vehicle data set, and the width and the height and the category of the vehicle contained in the image of the vehicle data set; normalizing the width and height of the vehicle contained in the image of the vehicle data set; calculating vehicle center point coordinates contained in an image of the vehicle data set; storing the txt text file according to the category serial number, the coordinates of the center point of the vehicle and the width and height information of the vehicle contained in the image of the vehicle data set; a txt text file is generated that stores the paths of all samples. The txt text file is saved in a tag folder. The normalization process of the width and height of the vehicle included in the image of the vehicle data set means: and converting the aspect ratio of the vehicle contained in the image of the vehicle data set into a numerical value between 0 and 1.
Based on yolov3 network prediction module, in prediction, a large amount of video resources or static images containing vehicle information are input into a trained network model, and structural information of the vehicle information, such as vehicle type, noon, vehicle body color and the like, is output at one time. The vehicle type is as domestic: north steam, biyadi, chang river, great wall, east wind, southeast, flying saucer, futian, fudi, hafei, space flight, red flag, huapu, huaxiang, ji ao, jili, rui wind, jiangling, golden cup, cheetah, south steam, qirui, huzhong, shenlong, wanfeng, shang fei, dao, shui guan, dicyclo, wuling, xiali, jinlong, yi, iverko, china, zhongshun, zhongxing, li , sight, legend, cheetah, and luo, etc.; germany: BMW, Audi, Porsche, Benz, Volkswagen; british: lauses, bentley, tigers, morgan, cobra, lotus, etc.; italy: feitin, farley, lambertian, mosalatty, blue flag, and the like; france: reynolds, logo, snowflake, etc.; switzerland: shenbao, Volvo, etc.; in the United states: general, claisler, ford, etc.; in Japan: toyota, Honda, Isuzu, Mazda, Suzuki, Nissan, Fuji, Mitsubishi, etc.; korea: modern, Shuanglong, Dayu, etc. The afternoon shares are as follows: a-2010; b-2011; c-2012; d-2013; e-2014; f-2015; g-2016; h-2017; j-2018; k-2019; l-2020; m-2021; n-2022; p-2023; r-2024; s-2025; t-2026; v-2027; w-2028; x-2029; y-2030; 1-2031; 2-2032; 3-2033; 4-2034; 5-2035; 6-2036; 7-2037; 8-2038; 9-2039; a-2040 and the like. The color of the car body is as follows: a: white B: ash C: yellow D: powder E: and (3) red F: purple G: green H: blue I: brown J: black Z: and others. For example, a large amount of video resources containing vehicle information or static images are input into a trained network model, wherein the structural information of one-time output of the vehicle information by one image is as follows: and B, A, the vehicle structural information represented by the image is: the vehicle type is made in Chery, the noon B means that the vehicle is produced in 2011 noon, and the body color A means that the body color of the vehicle is white.
In the existing design, for the identification of vehicle information, the vehicle information such as vehicle type, afternoon, body color and the like is available to a deep learning-based method, but basically processed separately, which consumes much time and has larger data volume for network training. By adopting the vehicle information identification method, the system, the memory and the processor, the input vehicle video information is processed at one time, no fusion process is generated in the middle, and various attribute value sets of the vehicle information, such as vehicle type identification, afternoon service identification, vehicle body color identification and the like, are output at one time. Based on the yolov3 network model, the parts of vehicle information such as vehicle type, noon, vehicle body color and the like are different categories, the network model is trained end to end, and the trained network model is used for predicting and outputting the structural information of the vehicle such as vehicle type, noon, vehicle body color and the like at one time. Compared with the prior art, the method has the advantages of less time consumption and high recognition efficiency, and can better meet the application requirements of actual scenes.
The above embodiments of the present invention are described in detail, and the principle and the implementation of the present invention are explained by applying specific embodiments, and the above description of the embodiments is only used to help understanding the method of the present invention and the core idea thereof; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.
Claims (20)
1. A vehicle information identification method characterized by comprising:
establishing a vehicle information data set;
setting network parameters based on yolov 3;
training the vehicle information data set based on the yolov3 network to obtain a multi-attribute network model capable of identifying the vehicle information;
and performing multi-attribute prediction on the vehicle information of the image to be recognized by using the obtained multi-attribute network model so as to recognize the vehicle structural information in the image to be recognized.
2. The vehicle information identification method according to claim 1, wherein the creating a vehicle information data set includes:
preprocessing the vehicle information data set;
and performing data enhancement processing on the vehicle information data set.
3. The vehicle information identification method according to claim 1, wherein the setting based on yolov3 network parameters includes:
setting a class number defined for the vehicle information data set based on a class parameter of the yolov3 network;
the filter parameters based on yolov3 network were set to 3 × (class parameters + 5).
4. The vehicle information identification method according to claim 1, wherein the training of the vehicle information data set based on the yolov3 network to obtain a multi-attribute network model that can identify the vehicle information comprises:
carrying out category calibration on the vehicle information data set by using a marking tool to generate a calibration file in an xml format;
converting the calibration file in the xml format into a txt file, and generating the txt file for storing the path of each image in the vehicle information data set;
and determining and changing network super parameters according to the categories in the vehicle information data set.
5. The vehicle information identification method according to claim 1, wherein the performing multi-attribute prediction on the vehicle information of the image to be identified by using the obtained multi-attribute network model comprises: and inputting the video resources or the static images containing the vehicle information into a trained network model, and outputting the structural information of the vehicle information at one time.
6. The vehicle information identification method according to claim 2, wherein the preprocessing the vehicle information data set includes:
and counting the vehicle information data sets, and storing the vehicle information data sets into different folders according to different categories.
7. The vehicle information identification method according to claim 2, wherein the data-enhancement processing of the vehicle information data set includes:
and turning the images of the vehicle information data set, and saving the corresponding turned images under the same folders of the images of the vehicle information data set in an additional name.
8. The vehicle information identification method according to claim 2, wherein the data-enhancement processing of the vehicle information data set includes:
and carrying out mirror image processing on the images of the vehicle information data set, and saving the corresponding mirror image images under the same folders of the images of the vehicle information data set in an additional name mode.
9. The vehicle information identification method according to claim 7, characterized in that an angle of the flip process is-15 ° to 15 °.
10. The vehicle information identification method of claim 4, wherein converting the calibration file in the xml format to a txt file, and generating the txt file storing a path along which each image in the vehicle information data set is located comprises:
reading a calibration file in an xml format to obtain the width and the height of an image containing the vehicle data set, and the width and the height and the category of the vehicle contained in the image of the vehicle data set;
normalizing the width and height of the vehicle contained in the image of the vehicle data set;
calculating vehicle center point coordinates contained in an image of the vehicle data set;
storing the txt text file according to the category serial number, the center point coordinate and the width and height of the vehicle contained in the image of the vehicle data set;
generating a txt text file storing the path of the vehicle information data set.
11. The vehicle information identification method of claim 10, wherein the txt text file is saved in a tag folder.
12. The vehicle information identification method according to claim 10, wherein the normalizing the width and height of the vehicle included in the image of the vehicle data set is performed by:
and converting the aspect ratio of the vehicle contained in the image of the vehicle data set into a numerical value between 0 and 1.
13. A storage medium characterized by comprising a stored program, wherein the program executes the vehicle information identification method according to any one of claims 1 to 12 when running.
14. A processor, characterized in that the processor is configured to execute a program, wherein the program executes the vehicle information identification method according to any one of claims 1 to 12.
15. A vehicle information identification system characterized by comprising: the system comprises a vehicle information data set setting module, a yolov 3-based network parameter setting module, a yolov 3-based network training module and a yolov 3-based network prediction module which are electrically connected;
the vehicle information data set setting module is used for establishing a vehicle information data set;
the yolov 3-based network parameter setting module is used for setting yolov 3-based network parameters;
the yolov 3-based network training module is used for training the vehicle information data set to obtain a multi-attribute network model capable of identifying the vehicle information;
the yolov 3-based network prediction module is used for performing multi-attribute prediction on the vehicle information of the image to be recognized by using the obtained multi-attribute network model so as to recognize the vehicle structural information in the image to be recognized.
16. The vehicle information identification system according to claim 15, wherein the vehicle information data set setting module further includes a data preprocessing module and a data enhancement processing module,
the data preprocessing module is used for preprocessing the vehicle information data set;
the data enhancement processing module is used for carrying out data enhancement processing on the vehicle information data set.
17. The vehicle information identification system of claim 15, wherein the yolov 3-based network parameter setting module sets a number of categories defined for the vehicle information dataset based on category parameters in a yolov3 network; the filter parameters based on yolov3 network parameters were set to 3 × (class parameters + 5).
18. The vehicle information identification system of claim 16, wherein the data preprocessing module counts the vehicle information data sets and stores the vehicle information data sets in different folders according to different categories.
19. The vehicle information identification system according to claim 16, wherein the data enhancement processing module performs a flipping process on the image of the vehicle information data set, and saves a corresponding flipped image under a folder with the same image of the vehicle information data set.
20. The vehicle information identification system according to claim 16, wherein the data enhancement processing module performs mirroring on the images of the vehicle information data set, and saves the corresponding mirrored images under a folder with the same images of the vehicle information data set.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811500432.3A CN111291779B (en) | 2018-12-07 | 2018-12-07 | Vehicle information identification method, system, memory and processor |
| PCT/CN2019/112504 WO2020114120A1 (en) | 2018-12-07 | 2019-10-22 | Method for identifying vehicle information, system, memory device, and processor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811500432.3A CN111291779B (en) | 2018-12-07 | 2018-12-07 | Vehicle information identification method, system, memory and processor |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111291779A true CN111291779A (en) | 2020-06-16 |
| CN111291779B CN111291779B (en) | 2024-09-13 |
Family
ID=70974906
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811500432.3A Active CN111291779B (en) | 2018-12-07 | 2018-12-07 | Vehicle information identification method, system, memory and processor |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN111291779B (en) |
| WO (1) | WO2020114120A1 (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112347693A (en) * | 2020-10-26 | 2021-02-09 | 上海感探号信息科技有限公司 | Vehicle operation dynamic mirror image simulation method, device and system |
| CN113191426A (en) * | 2021-04-28 | 2021-07-30 | 深圳市捷顺科技实业股份有限公司 | Vehicle identification model creation method, vehicle identification method and related components |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112906497A (en) * | 2021-01-29 | 2021-06-04 | 中国海洋大学 | Embedded safety helmet detection method and equipment |
| CN113657180A (en) * | 2021-07-23 | 2021-11-16 | 浙江大华技术股份有限公司 | Vehicle identification method, server and computer readable storage medium |
| CN115984840A (en) * | 2022-12-20 | 2023-04-18 | 航天科技控股集团股份有限公司 | Identity information identification system and method for oil truck for security system |
| CN115984274B (en) * | 2023-03-20 | 2023-05-30 | 菲特（天津）检测技术有限公司 | Vehicle appearance detection model, construction method and detection method |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106599869A (en) * | 2016-12-22 | 2017-04-26 | 安徽大学 | Vehicle attribute identification method based on multi-task convolutional neural network |
| CN106874840A (en) * | 2016-12-30 | 2017-06-20 | 东软集团股份有限公司 | Vehicle information recognition method and device |
| CN108537117A (en) * | 2018-03-06 | 2018-09-14 | 哈尔滨思派科技有限公司 | A kind of occupant detection method and system based on deep learning |
| WO2018166116A1 (en) * | 2017-03-13 | 2018-09-20 | 平安科技（深圳）有限公司 | Car damage recognition method, electronic apparatus and computer-readable storage medium |
| CN108764148A (en) * | 2018-05-30 | 2018-11-06 | 东北大学 | Multizone real-time action detection method based on monitor video |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107578007A (en) * | 2017-09-01 | 2018-01-12 | 杭州电子科技大学 | A kind of deep learning face identification method based on multi-feature fusion |
| CN108875754B (en) * | 2018-05-07 | 2022-04-05 | 华侨大学 | Vehicle re-identification method based on multi-depth feature fusion network |
-
2018
- 2018-12-07 CN CN201811500432.3A patent/CN111291779B/en active Active
-
2019
- 2019-10-22 WO PCT/CN2019/112504 patent/WO2020114120A1/en active Application Filing
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106599869A (en) * | 2016-12-22 | 2017-04-26 | 安徽大学 | Vehicle attribute identification method based on multi-task convolutional neural network |
| CN106874840A (en) * | 2016-12-30 | 2017-06-20 | 东软集团股份有限公司 | Vehicle information recognition method and device |
| WO2018166116A1 (en) * | 2017-03-13 | 2018-09-20 | 平安科技（深圳）有限公司 | Car damage recognition method, electronic apparatus and computer-readable storage medium |
| CN108537117A (en) * | 2018-03-06 | 2018-09-14 | 哈尔滨思派科技有限公司 | A kind of occupant detection method and system based on deep learning |
| CN108764148A (en) * | 2018-05-30 | 2018-11-06 | 东北大学 | Multizone real-time action detection method based on monitor video |
Non-Patent Citations (1)
| Title |
|---|
| 刘舒娴;林伟;陆培民;: "基于卷积神经网络的车型识别研究", 有线电视技术, no. 12 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112347693A (en) * | 2020-10-26 | 2021-02-09 | 上海感探号信息科技有限公司 | Vehicle operation dynamic mirror image simulation method, device and system |
| CN112347693B (en) * | 2020-10-26 | 2023-12-22 | 上海感探号信息科技有限公司 | Vehicle running dynamic mirror image simulation method, device and system |
| CN113191426A (en) * | 2021-04-28 | 2021-07-30 | 深圳市捷顺科技实业股份有限公司 | Vehicle identification model creation method, vehicle identification method and related components |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2020114120A1 (en) | 2020-06-11 |
| CN111291779B (en) | 2024-09-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111291779A (en) | Vehicle information identification method and system, memory and processor | |
| Liu et al. | Classifying urban land use by integrating remote sensing and social media data | |
| Williams et al. | Visualization and analysis of eddies in a global ocean simulation | |
| CN108596108A (en) | Method for detecting change of remote sensing image of taking photo by plane based on the study of triple semantic relation | |
| CN113516228B (en) | Network anomaly detection method based on deep neural network | |
| Zhao et al. | Accurate and efficient vehicle detection framework based on SSD algorithm | |
| CN114998744B (en) | Agricultural machinery track field dividing method and device based on motion and vision dual-feature fusion | |
| CN110378398A (en) | A kind of deep learning network improvement method based on the jump fusion of Analysis On Multi-scale Features figure | |
| Losch et al. | Semantic bottlenecks: Quantifying and improving inspectability of deep representations | |
| Hui et al. | Detail texture detection based on Yolov4‐tiny combined with attention mechanism and bicubic interpolation | |
| Wu et al. | Traffic sign detection based on SSD combined with receptive field module and path aggregation network | |
| CN115909280A (en) | Traffic sign recognition algorithm based on multi-head attention mechanism | |
| Ding et al. | Efficient vanishing point detection method in unstructured road environments based on dark channel prior | |
| Baluyan et al. | Novel approach for rooftop detection using support vector machine | |
| Meng et al. | A block object detection method based on feature fusion networks for autonomous vehicles | |
| Sun et al. | A method for extracting small water bodies based on DEM and remote sensing images | |
| Qin et al. | Dense sampling and detail enhancement network: Improved small object detection based on dense sampling and detail enhancement | |
| Chen et al. | Object counting in remote sensing via selective spatial‐frequency pyramid network | |
| CN116206203B (en) | Oil spill detection method based on SAR and Dual-EndNet | |
| Shi et al. | Anchor free remote sensing detector based on solving discrete polar coordinate equation | |
| CN115456957B (en) | Method for detecting change of remote sensing image by full-scale feature aggregation | |
| Nie et al. | Optimization of the economic and trade management legal model based on the support vector machine algorithm and logistic regression algorithm | |
| Yang et al. | Recognition of taxi violations based on semantic segmentation of PSPNet and improved YOLOv3 | |
| Jänicke et al. | Automatic detection and visualization of distinctive structures in 3d unsteady multi‐fields | |
| Xu et al. | Ship images detection and classification based on convolutional neural network with multiple feature regions |
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 | ||
| GR01 | Patent grant |