CN117333845A - Real-time detection method for small target traffic sign based on improved YOLOv5s - Google Patents
Real-time detection method for small target traffic sign based on improved YOLOv5s Download PDFInfo
- Publication number
- CN117333845A CN117333845A CN202311453323.1A CN202311453323A CN117333845A CN 117333845 A CN117333845 A CN 117333845A CN 202311453323 A CN202311453323 A CN 202311453323A CN 117333845 A CN117333845 A CN 117333845A
- Authority
- CN
- China
- Prior art keywords
- model
- traffic sign
- real
- yolov5s
- small target
- 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
- 238000011897 real-time detection Methods 0.000 title claims abstract description 33
- 238000000034 method Methods 0.000 title claims abstract description 20
- 238000001514 detection method Methods 0.000 claims abstract description 41
- 238000012549 training Methods 0.000 claims abstract description 27
- 238000012360 testing method Methods 0.000 claims abstract description 21
- 238000012795 verification Methods 0.000 claims abstract description 16
- 230000006870 function Effects 0.000 claims description 35
- 238000004364 calculation method Methods 0.000 claims description 15
- 238000011156 evaluation Methods 0.000 claims description 5
- 238000000691 measurement method Methods 0.000 claims description 5
- 238000013459 approach Methods 0.000 claims description 4
- 238000005259 measurement Methods 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 2
- 238000012544 monitoring process Methods 0.000 claims description 2
- 238000013527 convolutional neural network Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000011161 development Methods 0.000 description 3
- 230000018109 developmental process Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000001537 neural effect Effects 0.000 description 2
- 102100034112 Alkyldihydroxyacetonephosphate synthase, peroxisomal Human genes 0.000 description 1
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 description 1
- 101000799143 Homo sapiens Alkyldihydroxyacetonephosphate synthase, peroxisomal Proteins 0.000 description 1
- 238000000848 angular dependent Auger electron spectroscopy Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000001094 effect on targets Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000001902 propagating effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000001629 sign test Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
- G06V20/582—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of traffic signs
-
- 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a real-time detection method for a small target traffic sign based on improved YOLOv5s, which specifically comprises the following steps: (1) creating a traffic sign image dataset; (2) adding annotation information to the image in the dataset; (3) Constructing a small target traffic sign real-time detection model based on improved YOLOv5 s; (4) employing a new loss function; (5) training the model using the training set and the verification set; (6) And testing the model by adopting a test set, and obtaining a final small target traffic sign real-time detection model. Compared with the prior art, the real-time detection method for the small target traffic sign based on the improved YOLOv5s can effectively improve the detection precision of the small target traffic sign and greatly improve the real-time performance of traffic sign detection.
Description
Technical Field
The invention relates to the field of artificial intelligence, in particular to a real-time detection method for a small target traffic sign.
Background
With the continued development of autopilot technology, road traffic sign recognition (Traffic Sign Recognition, TSR) has become one of the key components of driving assistance systems (Advanced Driver Assistance Systems, ADAS). The TSR can efficiently and accurately identify the image captured by the camera, and provide information such as indication, warning and the like for a driver in real time. In the actual road driving process, the proportion of the traffic sign occupied by the traffic sign when the traffic sign just enters the visual field is generally less than 1%, the traffic sign belongs to small target detection, and the small target traffic sign has the problems of blurring and unobvious detailed characteristics in the image, so that the traffic sign has important significance for quickly and accurately identifying the small target traffic sign and assisting the development of a driving system.
The traditional traffic sign detection method mainly focuses on color segmentation, shape, contour and other features to perform feature extraction, and then feature classification is completed through a classifier, so that detection and identification of road traffic signs are realized. The traditional traffic sign detection algorithm extracts features by using a manual manufacturing mode, and then takes the features as the input of the algorithm, and the detection algorithm has the advantages of high dimension, large calculation amount, time and labor waste and poor instantaneity, stability and accuracy in the traffic sign detection process.
With the rapid development of deep learning, the object detection technology for improving convolutional neural models has become a current research hotspot. The convolutional neural model (Convolutional Neural Network, CNN) is used as a mainstream deep learning algorithm, and comprises a plurality of models such as R-CNN, fast-R-CNN, mask-R-CNN, alex Net, YOLO and the like. The models have good effect on target detection under different application scenes. Target detection algorithms can be divided into two categories: a two-stage detection algorithm using region candidate boxes and a single-stage detection algorithm not using region candidate boxes. The double-stage algorithm has high precision but slower detection speed, and the improved algorithm improves the speed to a certain extent, but due to the arrangement of the candidate areas, the double-stage model cannot better meet the real-time detection requirement of the small-target traffic sign in the high-speed driving process, so that the single-stage algorithm is generally adopted in the identification of the traffic sign. In the aspect of the single-stage target detection algorithm, compared with the double-stage detection algorithm, the real-time performance is improved, but the accuracy is lower. In particular, in the identification detection of small-target traffic signs, the detection precision is not high, missing detection and false detection are easy to cause, and the real-time detection is not good due to the large number of model parameters and high calculation complexity, so that the model has a certain difficulty in moving to the movable equipment.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a real-time detection method for a small target traffic sign based on improved YOLOv5s, which solves the problems of low detection precision, poor real-time performance, large model parameter quantity, high calculation complexity and inapplicability to be deployed on mobile equipment by improving a YOLOv5s target detection model and utilizing a feature enhancement and NWD measurement method.
The technical scheme provided by the invention comprises the following steps:
step 1: acquiring a traffic sign image to form a first data set;
step 2: adding annotation information to the images in the first data set to form a second data set, and dividing the second data set into a training set, a verification set and a test set;
step 3: constructing a traffic sign detection model based on improved YOLOv5 s;
step 4: adopting a new loss function;
step 5: training the traffic sign detection model based on the improved YOLOv5s by utilizing the training set and the verification set;
step 6: and testing the optimal model obtained through training by using the traffic sign test set, and obtaining a final small-target traffic sign real-time detection model.
Further, in step 1, the traffic sign image in the first data set may be captured by a digital camera, or may be collected from a network or may be obtained from a surveillance video.
Preferably, in the step 2, the training set, the verification set and the test set may be divided according to a ratio of 7:2:1.
Further, the step 3 specifically includes steps 3.1 to 3.3:
step 3.1: combining a C3 module in the Yolov5s model with ConvMixer to form a CSPCM module, and respectively replacing the C3 module of the last layer of the Yolov5s backbone network and the C3 module of the last layer of the neck network with the CSPCM module;
step 3.2: replacing the rest of C3 modules in the Yolov5s model backbone network and the neck network by using a lightweight convolution module C3_Faster, namely replacing the rest of C3 modules except the last layer of C3 modules in the backbone network and the neck network which are replaced in the step 3.1 by using C3_Faster;
step 3.3: the output layer is added with a small target detection head based on the existing 3 detection heads.
Further, in the step 4, an NWD measurement method is introduced into the CIoU loss function, the CIoU loss function of YOLOv5s is optimized by using the NWD measurement, and the optimized loss function formula is:
L＝(1-β)*(1-NWD(N a ，N b ))+β*(1-CIoU) (1)
in the formula (1), L is an optimized loss function, NWD is a normalized Wasserstein distance, N a ，N b Is composed of And->The modeled gaussian distribution, a represents the true box, b represents the predicted box, cx a ，，cy a Representing the coordinates of the center point, w, of the real frame a 、h a Representing the width and height of the real frame; cx (cx) b ，cy b Representing the coordinates of the central point of the prediction frame, w b ，h b Representing the width and height of the prediction box; beta is a weight proportionality coefficient, CIoU is a loss function in original YOLOv5s, and the calculation formula is:
in the formula (2), ρ 2 (b A ,b B ) Representing the euclidean distance between the center points of the real and predicted frames, c representing the diagonal distance of the smallest defined rectangle of the predicted and real frames, α being the weight factor, v being the aspect ratio uniformity, ioU being the overlap ratio between the real and predicted frames.
Further, the step 5 specifically includes steps 5.1 to 5.4:
step 5.1: setting the small target traffic sign real-time detection model training parameters based on the improved YOLOv5s, wherein the model training parameters specifically comprise: learning rate, momentum, weight decay, optimizer, iteration round number, batch size;
step 5.2: inputting the training set and verification set images and the corresponding labels into the small target traffic sign real-time detection model of the improved YOLOv5s, calculating the gradient of a loss function to model parameters by using a back propagation algorithm, and adjusting the model parameters by minimizing the loss function to gradually approach to an optimal solution;
step 5.3: updating model parameters using an optimizer to update them in the direction of gradient descent; until the loss functions of the training set and the verification set are not reduced, and the evaluation indexes such as the accuracy rate P, the recall rate R, mAP and the like are not improved;
step 5.4: and saving the trained model parameters as an optimal model.
Further, the step 6 specifically includes steps 6.1 to 6.3:
step 6.1: inputting the test set into the improved optimal model in the step 5;
step 6.2: calculating a model performance index: the performance indexes specifically comprise accuracy P (Precision), recall rate R, mAP, parameter quantity, calculation complexity (GFLOPs) and model size, and the specific formulas are as follows:
wherein P is the accuracy, R is the recall, mAP is the average accuracy average of all the categories, AP is the average accuracy, m is the total number of categories of traffic sign, TP represents the number of positive samples correctly identified as positive samples, FP represents the number of negative samples incorrectly identified as positive samples, and FN represents the number of positive samples incorrectly identified as negative samples;
step 6.3: and when the performance index meets the precision requirement, obtaining a small target traffic sign real-time detection model based on the improved YOLOv5 s.
Compared with the prior art, the invention has the beneficial effects that:
according to the real-time detection method of the small target traffic sign based on the improved YOLOv5s, the detection precision of the small target traffic sign is improved by introducing the CSPCM module and adding the small target detection head; the adoption of the C3_Faster network structure reduces model parameters and calculation complexity, and improves the real-time performance of traffic sign detection; an NWD measurement method is introduced into the CIoU loss function, a new loss function is adopted, larger loss caused by a small target is effectively avoided, and convergence of a model is quickened.
Drawings
FIG. 1 is a flow chart of the real-time detection method of small target traffic sign based on improved YOLOv5s of the present invention;
FIG. 2 is a schematic diagram of a model structure of a real-time detection method of a small target traffic sign based on improved YOLOv5 s;
FIG. 3 is a schematic diagram of a CSPCM structure;
FIG. 4 is a schematic diagram of a C3_Faster architecture;
Detailed Description
In order to make the technical scheme, constructional features, achieved objects and advantages of the present invention more apparent, the present invention will be further described in detail below with reference to the accompanying drawings. It should be noted that the specific embodiments described herein are only for the purpose of more clearly explaining the present invention, and are not intended to limit the present invention.
Fig. 1 is a flowchart of a method for detecting a small target traffic sign in real time based on improved YOLOv5s, which is implemented as follows:
step 1: acquiring a traffic sign image to form a first data set; the traffic sign images in the first data set can be shot by a digital camera, collected and obtained from a network or obtained from a monitoring video;
in this embodiment, in order to better evaluate the detection effect of the small target traffic sign real-time detection method based on improved YOLOv5s disclosed by the invention, a public data set TT100K (Tsinghua-Tencent 100K) traffic sign data set is adopted. With the TT100K dataset, the dataset needs to be preprocessed: firstly, counting traffic sign classes with the occurrence times smaller than 200 times in TT100K data sets, deleting labels corresponding to the traffic sign classes with the occurrence times smaller than 200 times and pictures only containing the classes, and finally manufacturing a data set containing 36 kinds of traffic signs, namely forming the first data set.
Step 2: and adding annotation information for the images in the first data set by using a labelem annotation tool. This step is skipped because the public data set TT100K taken in this embodiment has the annotation information already present. Converting the json format tag file in the first data set into the txt format tag file required by YOLOv5s to form a second data set, and dividing the second data set into a training set, a verification set and a test set according to a ratio of 7:2:1; the divided training set includes 6598 pictures, the test set includes 1889 pictures, the verification set includes 970 pictures, and in this embodiment, the size of the image in the second data set is 640 x 640.
Step 3: constructing a traffic sign detection model based on improved YOLOv5s, wherein the structure of the improved YOLOv5s model is shown in fig. 2, and the model construction process specifically comprises the steps of 3.1 to 3.3:
step 3.1: replacing the Bottleneck module in the C3 module with a ConvMixer module to form a new convolution module CSPCM; replacing the C3 module of the last layer of the backbond and the C3 module of the last layer of the Neck with a new convolution CSPCM module;
further, the structure of the ConvMixer module is shown in FIG. 3, which operates in a hybrid manner of separation space and channel dimensions, and maintains the same size and resolution throughout the model; in ConvMixer, the input is first convolved by a Depthwise Conv (Dwconv), i.e., a group number equal to the number of channels; then convolved by a Pointwise Conv (PWCONV), i.e., 1X 1; each convolution operation is followed by an activation function z' 1 And z 1+ 1 BatchNorm layer. The design can effectively reduce the number of model parameters, is beneficial to improving the real-time performance of the model, and effectively improves the feature expression capability of the model.
Step 3.2: replacing the rest of C3 modules in the Yolov5s model backbone network and the neck network by using a lightweight convolution module C3_Faster, namely replacing the rest of C3 modules except the last layer of C3 modules in the backbone network and the neck network which are replaced in the step 3.1 by using C3_Faster;
the lightweight convolution module c3_fast uses Pconv convolution; compared to deep convolution (DWConv) and group convolution (GConv), PConv can reduce the computational effort of the FLOPs, reduce computational redundancy and memory access, and fig. 4 is a c3_fast structure consisting of one PConv layer followed by 2 Conv layers, which together are shown as an inverted residual block, the middle layer has an extended number of channels, and after each middle Conv there is a normalization and activation layer.
Step 3.3: the output layer is added with a small target detection head based on the existing 3 detection heads, so that the multi-scale characteristic information can be processed more efficiently, the detector is more sensitive to the small target, and the detection performance of the small target traffic sign is improved.
Step 4: introducing an NWD measurement method into the CIoU loss function by adopting a new loss function, and combining the CIoU and the NWD measurement to provide a new loss function; the new loss function can effectively avoid larger loss caused by a small target and quicken the convergence of the model, and the specific steps comprise the steps 4.1 to 4.3:
step 4.1: modeling the boundary frame as two-dimensional Gaussian distribution, and calculating the Wasserstein distance between the predicted frame and the real frame by utilizing the Gaussian distribution corresponding to the predicted target and the actual target, wherein the calculation formula is as follows:
in the formula (1), the components are as follows,is Wasserstein distance, N a ,N b Is composed of->And the modeled gaussian distribution, a represents the true box, b represents the predicted box, cx a ，，cy a Representing the coordinates of the center point, w, of the real frame a 、h a Representing the width and height of the real frame; cx (cx) b ，cy b Representing the coordinates of the central point of the prediction frame, w b ，h b Representing the width and height of the prediction box;
step 4.2: the normalized Wasserstein distance between them is calculated as:
in the formula (2), NWD (N) a ，N b ) Is the normalized Wasserstein distance and C is a constant that is closely related to the dataset. Step 4.3: according to the proportional relation between CIoU and NWD, a new loss function is provided, and the calculation formula is as follows:
L＝(1-β)*(1-NWD(N a ，N b ))+β*(1-CIoU) (3)
in the formula (3), L is an optimized loss function, beta is a weight proportionality coefficient, CIoU is a loss function in original YOLOv5s, and a calculation formula is as follows:
in formula (4), ρ 2 (b A ，b B ) Representing the euclidean distance between the true and predicted frame center points, c representing the diagonal distance of the smallest bounding rectangle of the predicted and true frames, α being the weight factor, v being the aspect ratio uniformity, intersection overUnion (IoU) being the target detection for measuring the degree of overlap between the predicted and true bounding frames. The higher the IoU value is, the greater the overlapping degree between the predicted frame and the real frame is, the better the detection effect is, and the IoU calculation formula is as follows:
in formula (5): b is a prediction frame; b (B) GT Is a true box.
Step 5: inputting the training set and the verification set into the small target traffic sign model based on the improved YOLOv5s in the step 3 for training, wherein the training set and the verification set specifically comprise the steps 5.1 to 5.4:
step 5.1: setting training parameters of the small target traffic sign real-time detection model based on the improved YOLOv5s, wherein the model training parameters comprise: learning rate, momentum, weight decay, optimizer, iteration round number, batch size;
in this embodiment, the optimizer is SGD, the initial learning rate 1r0 is 0.01, the momentum is 0.937, the weight decay weight_decay is 0.0005, the batch size is 16, and the iteration round Epoch is 300.
Step 5.2: inputting the training set and verification set images and corresponding labels into the small target traffic sign real-time detection model of the improved YOLOv5s, and calculating gradients of the model parameters by using a back propagation algorithm. The back propagation algorithm is an efficient way to calculate the gradient, using the chain law to calculate the gradient of each parameter to the loss function. Specifically, back propagation computes the gradient of each parameter layer by propagating the loss function back from the output layer. In this process, the gradient of each parameter represents the rate of change of the loss function for that parameter, i.e. how the loss function changes with the change of that parameter. Model parameters are adjusted to gradually approach the optimal solution by minimizing the loss function.
Step 5.3: after calculating the gradients of the model parameters, an optimizer is used to update these parameters. And the optimizer updates the parameters according to the gradient information of the parameters and the opposite direction of the gradient. Parameters with larger gradients will be updated at a larger pace, while parameters with smaller gradients will be updated at a smaller pace. By continually iteratively updating the model parameters, the value of the loss function may be gradually reduced. The value of the model parameter is adjusted by minimizing the loss function, so that the model parameter gradually approaches to the optimal solution, namely, the loss function reaches the parameter value of the minimum value, the difference between the predicted result and the true value of the model is minimized, and meanwhile, the evaluation indexes such as mAP, recall rate R, accuracy rate P and the like are not improved.
Step 5.4: and saving the trained model parameters as an optimal model.
Step 6: and (3) testing the optimal model in the step (5) by adopting the test set, evaluating the test result of the test set, and meeting the precision requirement to obtain the final real-time detection model of the small target traffic sign based on the improved YOLOv5s, wherein the step (6) further comprises the steps (6.1) to (6.3):
step 6.1: inputting the test set into the optimal model in the step 5;
step 6.2: calculating a model performance index: the performance indexes comprise accuracy P, recall R, mAP, parameter quantity, calculation complexity GFLOPs and model size, and the specific calculation formulas are as follows:
wherein P is the accuracy, R is the recall, mAP is the average accuracy average of all the categories, AP is the average accuracy, m is the total number of categories of traffic sign, TP represents the number of positive samples correctly identified as positive samples, FP represents the number of negative samples incorrectly identified as positive samples, and FN represents the number of positive samples incorrectly identified as negative samples;
step 6.3: and when the performance index meets the precision requirement, obtaining a small target traffic sign real-time detection model based on the improved YOLOv5 s.
In this embodiment, in order to verify the improved model effect disclosed in the present invention, a test set is input into the small target traffic sign real-time detection model based on improved YOLOv5s and the YOLOv5s model is tested. Evaluation indexes are calculated for the two models respectively, and the evaluation index data are shown in table 1. As can be seen from Table 1The small target traffic sign real-time detection model disclosed by the invention is compared with the original YOLOv5s model in the accuracy rate P,mAP@0.5The indexes of the detection method and the detection device all obtain higher detection precision. The improved model disclosed by the invention is lower than the original YOLOv5s model in terms of performance indexes such as model parameter, model size, computational complexity GFLOPs and the like, so that the model has better light-weight index and is easier to deploy on mobile equipment.
Table 1 results of comparative experiments
The foregoing description of the embodiments of the present invention is not intended to limit the scope of the invention, but rather, it will be apparent to those skilled in the art that various modifications, equivalent substitutions, improvements, etc. can be made within the spirit and principles of the present invention.
Claims (3)
1. The real-time detection method for the small target traffic sign based on the improved YOLOv5s is characterized by comprising the following steps of:
step 1: acquiring a traffic sign image to form a first data set; the traffic sign images in the first data set can be shot by a digital camera, collected and obtained from a network or obtained from a monitoring video;
step 2: adding annotation information to the images in the first data set to form a second data set, and dividing the second data set into a training set, a verification set and a test set;
step 3: constructing a small target traffic sign real-time detection model based on improved YOLOv5s, wherein the construction of the model further comprises the steps of 3.1 to 3.3:
step 3.1: combining a C3 module in the Yolov5s model with ConvMixer to form a CSPCM module, and respectively replacing the C3 module of the last layer of the Yolov5s backbone network and the C3 module of the last layer of the neck network with the CSPCM module;
step 3.2: replacing the rest of C3 modules in the Yolov5s model backbone network and the neck network by using a lightweight convolution module C3_Faster, namely replacing the rest of C3 modules except the last layer of C3 modules in the backbone network and the neck network which are replaced in the step 3.1 by using C3_Faster;
step 3.3: the output layer is provided with a small target detection head based on the existing 3 detection heads;
step 4: the new loss function is adopted, and the specific method is as follows:
introducing an NWD measurement method into the CIoU loss function, optimizing the CIoU loss function of Yolov5s by using the NWD measurement, wherein the optimized loss function formula is as follows:
L＝(1-β)*(1-NWD(N a ,N b ))+β*(1-CIoU) (1)
NWD is normalized Wasserstein distance, N a ,N b Is composed ofAnd-> The modeled gaussian distribution, a represents the true box, b represents the predicted box, cx a ,,cy a Representing the coordinates of the center point, w, of the real frame a 、h a Representing the width and height of the real frame; cx (cx) b ,cy b Representing the coordinates of the central point of the prediction frame, w b ,h b Representing the width and height of the prediction box; beta is a weight proportionality coefficient, CIoU is a loss function in original YOLOv5s, and the CIoU calculation formula is:
in the formula (2), ρ 2 (b A ,b B ) Representing the center points of the real frames and the predicted framesEuclidean distance between, c represents the diagonal distance of the smallest defined rectangle of the predicted and real frames, α is the weight factor, v is the aspect ratio uniformity, ioU is the overlap ratio between the real and predicted frames;
step 5: training the small target traffic sign real-time detection model based on the improved YOLOv5s by using the training set and the verification set, and saving the trained model as an optimal model, and further comprising the steps of 5.1 to 5.4:
step 5.1: setting the small target traffic sign real-time detection model training parameters based on the improved YOLOv5s, wherein the model training parameters comprise: learning rate, momentum, weight decay, optimizer, iteration round number, batch size;
step 5.2: inputting the training set and verification set images and the corresponding labels into the small target traffic sign real-time detection model of the improved YOLOv5s, calculating the gradient of a loss function to model parameters by using a back propagation algorithm, and adjusting the model parameters to gradually approach an optimal solution by minimizing the loss function;
step 5.3: updating model parameters by using an optimizer SGD (generalized algorithm D), so that the model parameters are updated towards the gradient descending direction until the loss functions of the training set and the verification set are not reduced any more, and the evaluation index mAP, the recall rate R and the accuracy rate P are not improved any more;
step 5.4: saving the trained model parameters as an optimal model;
step 6: and testing the optimal model by adopting the test set, evaluating the test result of the test set, and meeting the precision requirement to obtain the final real-time detection model of the small target traffic sign based on the improved YOLOv5 s.
2. The real-time detection method of small target traffic sign based on improved YOLOv5s according to claim 1, wherein in the step 2, the training set, the verification set and the test set are divided according to a ratio of 7:2:1.
3. The real-time detection method of small target traffic sign based on improved YOLOv5s according to claim 1, wherein the step 6 further comprises steps 6.1 to 6.3:
step 6.1: inputting the test set into the optimal model in the step 5;
step 6.2: calculating a model performance index: the accuracy P, the recall R, mAP, the parameter quantity, the calculation complexity GFLOPs and the model size are calculated according to the following specific calculation formulas:
wherein P is the accuracy, R is the recall, mAP is the average accuracy average of all the categories, AP is the average accuracy, m is the total number of categories of traffic sign, TP represents the number of positive samples correctly identified as positive samples, FP represents the number of negative samples incorrectly identified as positive samples, and FN represents the number of positive samples incorrectly identified as negative samples;
step 6.3: and when the performance index meets the precision requirement, obtaining a small target traffic sign real-time detection model based on the improved YOLOv5 s.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311453323.1A CN117333845A (en) | 2023-11-03 | 2023-11-03 | Real-time detection method for small target traffic sign based on improved YOLOv5s |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311453323.1A CN117333845A (en) | 2023-11-03 | 2023-11-03 | Real-time detection method for small target traffic sign based on improved YOLOv5s |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117333845A true CN117333845A (en) | 2024-01-02 |
Family
ID=89275667
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311453323.1A Pending CN117333845A (en) | 2023-11-03 | 2023-11-03 | Real-time detection method for small target traffic sign based on improved YOLOv5s |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117333845A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118230079A (en) * | 2024-05-27 | 2024-06-21 | 中国科学院西安光学精密机械研究所 | Detection method for remote sensing small target based on improved YOLO |
| CN118570740A (en) * | 2024-08-02 | 2024-08-30 | 南京信息工程大学 | Bus stop scene vehicle pedestrian detection method based on improvement YOLOV5 |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200210826A1 (en) * | 2018-12-29 | 2020-07-02 | Northeastern University | Intelligent analysis system using magnetic flux leakage data in pipeline inner inspection |
| US20210370993A1 (en) * | 2020-05-27 | 2021-12-02 | University Of South Carolina | Computer vision based real-time pixel-level railroad track components detection system |
| CN114998831A (en) * | 2022-05-23 | 2022-09-02 | 重庆科技学院 | Fire detection method and system based on edge calculation and improved YOLOv3 |
| CN116092115A (en) * | 2022-11-28 | 2023-05-09 | 中国计量大学 | Real-time lightweight construction personnel safety dressing detection method |
| CN116259032A (en) * | 2022-11-30 | 2023-06-13 | 长春工业大学 | Road traffic sign detection and identification algorithm based on improved YOLOv5 |
| CN116434188A (en) * | 2023-03-26 | 2023-07-14 | 复旦大学 | Traffic sign detection method based on improved_yolov5s network |
| CN116469021A (en) * | 2023-04-30 | 2023-07-21 | 桂林电子科技大学 | Unmanned aerial vehicle aerial photographing target detection method based on improved YOLOv5 |
| CN116597413A (en) * | 2023-04-28 | 2023-08-15 | 哈尔滨工业大学 | Real-time traffic sign detection method based on improved YOLOv5 |
| CN116630920A (en) * | 2023-05-18 | 2023-08-22 | 天津科技大学 | Improved lane line type identification method of YOLOv5s network model |
| US11790640B1 (en) * | 2022-06-22 | 2023-10-17 | Ludong University | Method for detecting densely occluded fish based on YOLOv5 network |
| US20230351573A1 (en) * | 2021-03-17 | 2023-11-02 | Southeast University | Intelligent detection method and unmanned surface vehicle for multiple type faults of near-water bridges |
-
2023
- 2023-11-03 CN CN202311453323.1A patent/CN117333845A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200210826A1 (en) * | 2018-12-29 | 2020-07-02 | Northeastern University | Intelligent analysis system using magnetic flux leakage data in pipeline inner inspection |
| US20210370993A1 (en) * | 2020-05-27 | 2021-12-02 | University Of South Carolina | Computer vision based real-time pixel-level railroad track components detection system |
| US20230351573A1 (en) * | 2021-03-17 | 2023-11-02 | Southeast University | Intelligent detection method and unmanned surface vehicle for multiple type faults of near-water bridges |
| CN114998831A (en) * | 2022-05-23 | 2022-09-02 | 重庆科技学院 | Fire detection method and system based on edge calculation and improved YOLOv3 |
| US11790640B1 (en) * | 2022-06-22 | 2023-10-17 | Ludong University | Method for detecting densely occluded fish based on YOLOv5 network |
| CN116092115A (en) * | 2022-11-28 | 2023-05-09 | 中国计量大学 | Real-time lightweight construction personnel safety dressing detection method |
| CN116259032A (en) * | 2022-11-30 | 2023-06-13 | 长春工业大学 | Road traffic sign detection and identification algorithm based on improved YOLOv5 |
| CN116434188A (en) * | 2023-03-26 | 2023-07-14 | 复旦大学 | Traffic sign detection method based on improved_yolov5s network |
| CN116597413A (en) * | 2023-04-28 | 2023-08-15 | 哈尔滨工业大学 | Real-time traffic sign detection method based on improved YOLOv5 |
| CN116469021A (en) * | 2023-04-30 | 2023-07-21 | 桂林电子科技大学 | Unmanned aerial vehicle aerial photographing target detection method based on improved YOLOv5 |
| CN116630920A (en) * | 2023-05-18 | 2023-08-22 | 天津科技大学 | Improved lane line type identification method of YOLOv5s network model |
Non-Patent Citations (9)
| Title |
|---|
| PEILIN LIU等: "UCN-YOLOv5: Traffic Sign Object Detection Algorithm Based on Deep Learning", 《 IEEE ACCESS ( VOLUME: 11)》, 5 October 2023 (2023-10-05) * |
| WANCHANG JIANG等: "Top Invulnerability Nodes Mining in Dual-Direction Different-Weight Complex Network Based on Node Double-Level Local Structure Weighted Entropy", 《 IEEE ACCESS ( VOLUME: 7)》, 28 June 2019 (2019-06-28) * |
| XIA, KW等: "Global contextual attention augmented YOLO with ConvMixer prediction heads for PCB surface defect detection", 《SCIENTIFIC REPORTS》, 16 June 2023 (2023-06-16) * |
| 姜万昌;杨旸;隋吉生;: "基于小波变换的ARMA-RBF光功率组合预测", 光通信技术, no. 12, 15 December 2013 (2013-12-15) * |
| 崔家睿: "基于多目标追踪的交通场景异常检测", 《硕士电子期刊出版信息》, 15 July 2023 (2023-07-15) * |
| 张钟文;高宇;王静;曹登平;: "基于YOLOv3的深度学习交通标志识别系统", 建筑电气, no. 07, 25 July 2020 (2020-07-25) * |
| 殷浩东: "基于深度学习的行人跟踪算法研究", 《硕士电子期刊出版信息》, 15 March 2022 (2022-03-15) * |
| 葛明进;孙作雷;孔薇;: "基于anchor-free的交通场景目标检测技术", 计算机工程与科学, no. 04, 15 April 2020 (2020-04-15) * |
| 郑冬;李向群;许新征;: "基于轻量化SSD的车辆及行人检测网络", 南京师大学报(自然科学版), no. 01, 20 March 2019 (2019-03-20) * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118230079A (en) * | 2024-05-27 | 2024-06-21 | 中国科学院西安光学精密机械研究所 | Detection method for remote sensing small target based on improved YOLO |
| CN118570740A (en) * | 2024-08-02 | 2024-08-30 | 南京信息工程大学 | Bus stop scene vehicle pedestrian detection method based on improvement YOLOV5 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110059554B (en) | Multi-branch target detection method based on traffic scene | |
| US20230222844A1 (en) | Parking lot management and control method based on object activity prediction, and electronic device | |
| CN111898432B (en) | Pedestrian detection system and method based on improved YOLOv3 algorithm | |
| CN112016605B (en) | Target detection method based on corner alignment and boundary matching of bounding box | |
| CN117333845A (en) | Real-time detection method for small target traffic sign based on improved YOLOv5s | |
| CN111950453A (en) | Optional-shape text recognition method based on selective attention mechanism | |
| CN109583483A (en) | A kind of object detection method and system based on convolutional neural networks | |
| CN112633149B (en) | Domain-adaptive foggy-day image target detection method and device | |
| JP2022025008A (en) | License plate recognition method based on text line recognition | |
| CN112634369A (en) | Space and or graph model generation method and device, electronic equipment and storage medium | |
| CN110032952A (en) | A kind of road boundary point detecting method based on deep learning | |
| CN114360067A (en) | Dynamic gesture recognition method based on deep learning | |
| CN115170611A (en) | Complex intersection vehicle driving track analysis method, system and application | |
| CN114332921A (en) | Pedestrian detection method based on improved clustering algorithm for Faster R-CNN network | |
| CN110598758A (en) | Training modeling method, vehicle charging method, management system, and storage medium | |
| CN104599291A (en) | Structural similarity and significance analysis based infrared motion target detection method | |
| CN112949510A (en) | Human detection method based on fast R-CNN thermal infrared image | |
| CN117975418A (en) | Traffic sign detection method based on improved RT-DETR | |
| CN115690549A (en) | Target detection method for realizing multi-dimensional feature fusion based on parallel interaction architecture model | |
| Mijić et al. | Traffic sign detection using YOLOv3 | |
| CN115719475A (en) | Three-stage trackside equipment fault automatic detection method based on deep learning | |
| CN117456480B (en) | Light vehicle re-identification method based on multi-source information fusion | |
| CN113673534B (en) | RGB-D image fruit detection method based on FASTER RCNN | |
| CN117437615A (en) | Foggy day traffic sign detection method and device, storage medium and electronic equipment | |
| CN113450321B (en) | Single-stage target detection method based on edge detection |
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 |