CN110909670A - Unstructured road identification method - Google Patents
Unstructured road identification method Download PDFInfo
- Publication number
- CN110909670A CN110909670A CN201911145132.2A CN201911145132A CN110909670A CN 110909670 A CN110909670 A CN 110909670A CN 201911145132 A CN201911145132 A CN 201911145132A CN 110909670 A CN110909670 A CN 110909670A
- Authority
- CN
- China
- Prior art keywords
- image
- neural network
- road
- formula
- pheromone
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/182—Network patterns, e.g. roads or rivers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an unstructured road identification method, which mainly improves a BP neural network through an ant colony algorithm to increase the identification rate of unstructured roads. The method mainly comprises the following steps: firstly, acquiring an unstructured road image by using an automobile monocular camera; secondly, preprocessing the image and extracting characteristic values of image blocks; thirdly, the ant colony algorithm optimizes the parameters of the BP neural network, and further establishes an initial framework of the BP neural network; and (IV) taking the characteristic value of the image block as the input of the BP neural network, learning the sample by using the BP neural network, and testing the optimized BP neural network by using the test sample to obtain an image identification result. The method is used for identifying the unstructured road, so that the road area and the non-road area in the unstructured road can be quickly and accurately separated, and the problems of low identification rate and long identification time in the aspect of unstructured road identification in the existing method are solved.
Description
Technical Field
The invention relates to the field of automobile image identification, in particular to an improved unstructured road identification method.
Background
The road recognition is mainly to automatically predict the position and the category of an interested target in a picture through a computer, the improvement of the road recognition technology plays an important role in the development of the automatic driving technology, and can help the automatic driving vehicle to cope with complex road conditions and emergencies and ensure the driving safety.
The road identification is mainly divided into structured road identification and unstructured road identification, wherein the structured road refers to a highway with good structure such as an expressway and an urban trunk road, the road has clear road sign lines, the background environment of the road is single, the geometric characteristics of the road are obvious, and the detection and identification difficulty is low, so that the identification method for the road tends to be perfect. Compared with structured roads, unstructured roads have the characteristics of degraded edge lines, other coverings on the surfaces of the roads, unobvious road and non-road boundaries and the like, and are influenced by shadows, water marks and the like, so that the road areas and the non-road areas are difficult to distinguish, the detection and identification of the roads become more difficult, and unstructured road identification is also the main research direction of the current road identification technology.
At present, pattern recognition or comparison is mainly performed through a method similar to machine learning, so as to recognize the road according to the characteristic that the difference of the road part and the non-road part in the unstructured road environment is large in color. Although a certain recognition effect can be achieved, the problems of inaccurate recognition, limited application range, low recognition efficiency and the like still exist, and in order to promote further development of the unmanned technology and ensure driving safety of unmanned vehicles, a more effective improvement needs to be made on the unstructured road recognition technology so as to achieve the goal of quickly and accurately recognizing the road in a complex scene.
Disclosure of Invention
The invention aims to solve the defects in the prior art, and provides an unstructured road identification method to solve the problems of low identification rate and long identification time of the existing method in the aspect of unstructured road identification.
The technical scheme of the invention is as follows: an unstructured road identification method mainly comprises the following steps
Firstly, acquiring an unstructured road image by using an automobile monocular camera;
secondly, preprocessing the collected image, blocking the image and extracting characteristic values of the image blocks;
thirdly, optimizing parameters of the BP neural network by using an ant colony algorithm;
and (IV) using the characteristic values of the image blocks of the training sample as input to act on the BP neural network for training, and using the characteristic values in the test sample as input to carry out classification and identification on the image.
Further, the blocking and extracting process in the step two is as follows:
(1) performing ROI region selection, graying and median filtering on the acquired image, and selecting the size of an image block according to the pixel size of the preprocessed image;
(2) extracting the characteristic values of the blocks, mainly extracting the gray average value, the variance, the entropy, the consistency and the smoothness of the blocks, wherein the extraction of the characteristics can be obtained by calculating according to the following method:
1. mean value of gray scale
2. Variance (variance)
3. Entropy of the entropy
4. Consistency
5. Smoothness of the surface
Wherein M, N represents the length and width of the image, i.e. the image has M × N pixels, I (I, j) represents the gray value of the pixel in the ith row and j column, ziRepresenting the ith gray level, L being the total number of gray levels, P (z)i) Representing a grey level of z in the grey level distribution of the normalized histogramiThe probability of (c).
Further, the process of using the ant colony algorithm to perform optimization in the third step is as follows:
setting a food source node set as B, wherein the food source node set corresponds to parameters needing to be optimized in the unstructured road recognition BP neural network, namely the number of neurons of a hidden layer, the weight values of input and hidden layers of a neural network, the weight values of the hidden layer and an output layer and the threshold value of the neurons, so that the number of elements in the set B is 4; definition of dijIs the length of path ij (i, j ═ 1,2, 3.. n), τij(t) denotes the concentration of pheromones on the paths ij at time t, and the concentration of pheromones on the paths at the initial time is equal, i.e., τij(0) C; setting the number of ants in the ant colony as m; the number of neurons in the hidden layer can be determined byDetermining, wherein I is the number of input layer neurons, o is the number of output layer neurons, and rho is a constant greater than 1 and smaller than 10, and an interval of the number of hidden layer neurons can be obtained by the formula;
(1) making the cycle number N equal to 0;
(2) n is N +1 and N is not more than Nmax;
(3) M ants in the ant colony start, in the optimization process, the kth ant calculates the probability from the point i to the point j according to the formula vi and advances to the point j with higher transition probability,
in formula vi, allowkIndicating the node that the kth ant did not visit, ηij(t) is a heuristic function, ηij(t)=1/dijα is the factor of importance of pheromonesThe larger the value of the factor is, the greater the effect of the concentration of the pheromone in the transfer is, β is an importance degree factor of the heuristic function, and the larger the value is, the greater the effect of the heuristic function in the transfer is, namely, the ants can transfer to the nodes with short distance with higher probability;
(4) when the kth ant determines the node j to be accessed according to the formula vi and finishes the access of the path ij, the pheromone on the path is updated according to the formula vii,
τij(t+1)=(1-ρ)τij(t)+△τijvii
in the formula vii, (1-. rho) represents the volatilization rate of the pheromone, and △. tau.ijRepresenting the amount of change of the pheromone on path ij over the period t to t +1, representing the contribution of the kth ant to the pheromone on path ij, which can be calculated according to formula viii,
(5) repeating steps 2-4 until N ═ NmaxOr all ants converge to the same optimal path.
Further, the specific identification process in step four is as follows:
(1) substituting parameters of the BP neural network obtained by ant colony algorithm optimization into a neural network model, and taking characteristic values of image blocks in a training sample as inputs, namely 5 inputs in the established model are respectively the gray average value, the variance, the entropy, the consistency and the smoothness of the blocks, and assuming OjRepresenting the actual output of the jth output neuron, OjCan be expressed as:
yi(k)=W(k)P(k)T+γiix
in the formula, gammaiIs neuron threshold, P (k) is input of previous layer, i is inputThe number of ends, W (k), represents the weight between the output of the previous layer and the neuron of the next layer;
(2) according to the actual output OjAnd a desired output Oj TThe difference between them to calculate the error, as shown in the following equation:
(3) according to an error back propagation algorithm, errors are respectively calculated from output to input to adjust parameters, so that a network structure is optimized, and an adjustment formula is shown as the formula xi:
η is the learning rate, the learning speed is adjusted, t represents the current training times, r represents the current layer;
(4) repeatedly executing the previous steps until the network is stable and reaches a preset error range;
(5) and taking the characteristic value of the image block extracted from the test sample as the input of the BP neural network, and identifying the image in the test sample.
The invention has the beneficial effects that:
compared with the traditional method for distinguishing the road and the non-road according to the difference of the colors of the road and the non-road area, the method for distinguishing the unstructured road carries out blocking processing on the road and the non-road area, extracts the characteristics of the blocks, such as the gray average value, the variance, the entropy, the consistency, the smoothness and the like, optimizes the parameters of the BP neural network by using an ant colony algorithm, trains according to 5 feature vectors of image blocks, reduces the possibility of error recognition of a road by using a single color feature vector, improves the recognition accuracy of the unstructured road, and can quickly and accurately separate the road area and the non-road area in the unstructured road.
Drawings
FIG. 1 is a block diagram of image blocks and blocks requiring characteristic value extraction
FIG. 2 is a system flow diagram of an unstructured road identification method.
Detailed Description
The following examples further illustrate the present invention but are not to be construed as limiting the invention. Modifications and substitutions to methods, procedures, or conditions of the invention may be made without departing from the spirit of the invention.
Referring to fig. 1 and fig. 2, in order to solve the problems of low recognition rate and long recognition time in the unstructured road recognition in the prior art, the present embodiment provides an improved unstructured road recognition method, which includes the steps of:
firstly, acquiring an unstructured road image by using an automobile monocular camera;
secondly, preprocessing the acquired image, blocking the image and extracting characteristic values of the blocked image;
thirdly, optimizing parameters of the BP neural network by using an ant colony algorithm;
and (IV) taking the characteristic value of the image block of the training sample as an input, applying the characteristic value to the BP neural network for training, taking the characteristic value in the test sample as an input, and carrying out classification and identification on the image to obtain an image identification result.
According to the image blocking and characteristic value extraction process in the step two, the method comprises the following steps:
(1) the size of a unit block is related to the pixels of an image, and in order to highlight the curved edge information of a road, a rectangular template is selected to block the image, as shown in fig. 1, a picture of 540 × 360 pixels (h is 360, and w is 540) is adopted, and the block size is 90 × 60 pixels;
(2) extracting feature values of the blocks, as shown in fig. 1, extracting feature values of a road region block 1 and a non-road region block 2 after the blocks are divided, and mainly extracting a gray average value, a variance, an entropy, consistency and smoothness of the blocks; the extraction of these several features can be calculated according to the following method:
1. mean value of gray scale
2. Variance (variance)
3. Entropy of the entropy
4. Consistency
5. Smoothness of the surface
Wherein M, N represents the length and width of the image, i.e. the image has M × N pixels, I (I, j) represents the gray value of the pixel in the ith row and j column, ziRepresenting the ith gray level, L being the total number of gray levels, P (z)i) Representing a grey level of z in the grey level distribution of the normalized histogramiThe probability of (c).
The process of using the ant colony algorithm to perform optimization in the third step is as follows:
setting a food source node set as B, wherein the food source node set corresponds to parameters needing to be optimized in the unstructured road recognition BP neural network, namely the number of neurons of a hidden layer, the weight values of input and hidden layers of a neural network, the weight values of the hidden layer and an output layer and the threshold value of the neurons, so that the number of elements in the set B is 4; definition of dijIs the length of path ij (i, j ═ 1,2, 3.. n), τij(t) denotes the concentration of pheromones on the paths ij at time t, and the concentration of pheromones on the paths at the initial time is equal, i.e., τij(0) C; setting the number of ants in the ant colony as m; the number of neurons in the hidden layer can be determined byDetermining, wherein I is the number of input layer neurons, o is the number of output layer neurons, and rho is a constant greater than 1 and smaller than 10, and an interval of the number of hidden layer neurons can be obtained by the formula;
(1) make the cycle number N equal to 0
(2) N is N +1 and N is not more than Nmax;
(3) M ants in the ant colony start, in the optimizing process, the kth ant calculates the probability from the point i to the point j according to a formula vi provided in the specification, and moves to the point j with higher transition probability,
in formula vi, allowkIndicating the node that the kth ant did not visit, ηij(t) is a heuristic function, ηij(t)=1/dijβ is a heuristic function importance factor, the larger the value of which is, the larger the role of the heuristic function in the transfer is, namely, ants can transfer to nodes with short distance with higher probability;
(4) and when the kth ant determines a node j to be accessed according to the formula vi and finishes the access of the path ij, updating the pheromone according to an pheromone updating formula vii on the path.
τij(t+1)=(1-ρ)τij(t)+△τijvii
In the formula vii, (1-. rho) represents the volatilization rate of the pheromone, and △. tau.ijRepresenting the amount of change of the pheromone on path ij over the period t to t +1, representing the contribution of the kth ant to the pheromone on path ij, which can be calculated according to formula viii,
(5) repeating steps 2-4 until N ═ NmaxOr all ants converge to the same optimal path.
The specific identification process in the fourth step is as follows:
(1) substituting parameters of a BP neural network obtained by ant colony algorithm optimization into a neural network model, and taking characteristic values of image blocks in a training sample as inputs, namely 5 inputs in the established model are respectively the gray level mean value, the variance, the entropy, the consistency and the smoothness of the blocks; suppose OjRepresenting the actual output of the jth output neuron, OjCan be expressed as:
yi(k)=W(k)P(k)T+γiix
in the formula, gammaiIs neuron threshold, P (k) is input of the previous layer, i is the number of input ends, W (k) represents the weight between the output of the previous layer and the neuron of the next layer;
(2) according to the actual output OjAnd a desired output Oj TThe difference between them to calculate the error, as shown in the following equation:
(3) according to an error back propagation algorithm, the parameters are adjusted by respectively calculating errors from output to input, so that the network structure is optimized, wherein an adjustment formula is as follows:
η is the learning rate, the learning speed is adjusted, t represents the current training times, r represents the current layer;
(4) repeatedly executing the previous steps until the network is stable and reaches a preset error range;
(5) taking the characteristic value of the image block extracted from the test sample as the input of a BP neural network, and identifying the image in the test sample; since it is only necessary to determine whether the feature value of the block in the map belongs to a road or a non-road, the output of the BP neural network is only one, that is, the road region and the non-road region, and is 1 if the road region is the road region or 0 if the road region is the non-road region, so that the road region and the non-road region can be separated from each other in the image.
The foregoing illustrates and describes the principles, general features, and advantages of the present invention. However, the above description is only an example of the present invention, the technical features of the present invention are not limited thereto, and any other embodiments that can be obtained by those skilled in the art without departing from the technical solution of the present invention should be covered by the claims of the present invention.
Claims (4)
1. An unstructured road identification method is characterized by mainly comprising the following steps:
firstly, acquiring an unstructured road image by using an automobile monocular camera;
secondly, preprocessing the collected image, blocking the image and extracting characteristic values of the image blocks;
thirdly, optimizing parameters of the BP neural network by using an ant colony algorithm;
and (IV) using the characteristic values of the image blocks of the training sample as input to act on the BP neural network for training, and using the characteristic values in the test sample as input to carry out classification and identification on the image.
2. The unstructured road identification method of claim 1, wherein the partitioning and extracting process in the second step is as follows:
(1) performing ROI region selection, graying and median filtering on the acquired image, and selecting the size of an image block according to the pixel size of the preprocessed image;
(2) extracting the characteristic values of the blocks, mainly extracting the gray average value, the variance, the entropy, the consistency and the smoothness of the blocks, wherein the extraction of the characteristics can be obtained by calculating according to the following method:
1. mean value of gray scale
2. Variance (variance)
3. Entropy of the entropy
4. Consistency
5. Smoothness of the surface
Wherein M, N represents the length and width of the image, i.e. the image has M × N pixels, I (I, j) represents the gray value of the pixel in the ith row and j column, ziRepresenting the ith gray level, L being the total number of gray levels, P (z)i) Representing a grey level of z in the grey level distribution of the normalized histogramiThe probability of (c).
3. The unstructured road identification method of claim 1, wherein the optimization process using the ant colony algorithm in step three is as follows:
setting a food source node set as B, wherein the food source node set corresponds to parameters needing to be optimized in the unstructured road recognition BP neural network, namely the number of neurons of a hidden layer, the weight values of input and hidden layers of a neural network, the weight values of the hidden layer and an output layer and the threshold value of the neurons, so that the number of elements in the set B is 4; definition of dijIs the length of path ij (i, j ═ 1,2, 3.. n), τij(t) indicates the way at time tThe pheromone density on the path ij is equal to the pheromone density on each path at the initial time, i.e., τij(0) C; setting the number of ants in the ant colony as m; the number of neurons in the hidden layer can be determined byDetermining, wherein I is the number of input layer neurons, o is the number of output layer neurons, and rho is a constant greater than 1 and smaller than 10, and an interval of the number of hidden layer neurons can be obtained by the formula;
(1) making the cycle number N equal to 0;
(2) n is N +1 and N is not more than Nmax;
(3) M ants in the ant colony start, in the optimization process, the kth ant calculates the probability from the point i to the point j according to the formula vi and advances to the point j with higher transition probability,
in formula vi, allowkIndicating the node that the kth ant did not visit, ηij(t) is a heuristic function, ηij(t)=1/dijα is an pheromone importance factor, the larger the value of which is, the larger the effect of the concentration of the pheromone in the transfer is, β is an elicitation function importance factor, the larger the value is, the larger the effect of the elicitation function in the transfer is, that is, the ants can transfer to the nodes with short distance with a larger probability;
(4) when the kth ant determines the node j to be accessed according to the formula vi and finishes the access of the path ij, the pheromone on the path is updated according to the formula vii,
τij(t+1)=(1-ρ)τij(t)+△τijvii
in the formula vii, (1-. rho) represents the volatilization rate of the pheromone, and △. tau.ijRepresenting the amount of change of the pheromone on path ij over the period t to t +1, representing the contribution of the kth ant to the pheromone on path ij, which can be calculated according to formula viii,
(5) repeating steps 2-4 until N ═ NmaxOr all ants converge to the same optimal path.
4. The unstructured road identification method of claim 1, wherein the specific identification process in the fourth step is as follows:
(1) substituting parameters of the BP neural network obtained by ant colony algorithm optimization into a neural network model, and taking characteristic values of image blocks in a training sample as inputs, namely 5 inputs in the established model are respectively the gray average value, the variance, the entropy, the consistency and the smoothness of the blocks, and assuming OjRepresenting the actual output of the jth output neuron, OjCan be expressed as:
yi(k)=W(k)P(k)T+γiix
in the formula, gammaiIs neuron threshold, P (k) is input of the previous layer, i is the number of input ends, W (k) represents the weight between the output of the previous layer and the neuron of the next layer;
(2) according to the actual output OjAnd a desired output Oj TThe difference between them to calculate the error, as shown in the following equation:
(3) according to an error back propagation algorithm, errors are respectively calculated from output to input to adjust parameters, so that a network structure is optimized, and an adjustment formula is shown as the formula xi:
η is the learning rate, the learning speed is adjusted, t represents the current training times, r represents the current layer;
(4) repeatedly executing the previous steps until the network is stable and reaches a preset error range;
(5) and taking the characteristic value of the image block extracted from the test sample as the input of the BP neural network, and identifying the image in the test sample.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911145132.2A CN110909670B (en) | 2019-11-21 | 2019-11-21 | Unstructured road identification method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911145132.2A CN110909670B (en) | 2019-11-21 | 2019-11-21 | Unstructured road identification method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110909670A true CN110909670A (en) | 2020-03-24 |
CN110909670B CN110909670B (en) | 2023-06-06 |
Family
ID=69816964
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911145132.2A Active CN110909670B (en) | 2019-11-21 | 2019-11-21 | Unstructured road identification method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110909670B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112949617A (en) * | 2021-05-14 | 2021-06-11 | 江西农业大学 | Rural road type identification method, system, terminal equipment and readable storage medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103955743A (en) * | 2014-04-22 | 2014-07-30 | 江苏大学 | Ultrahigh-pressure water jet road mark line cleaning effect forecasting method and device |
US20170116477A1 (en) * | 2015-10-23 | 2017-04-27 | Nokia Technologies Oy | Integration of positional data and overhead images for lane identification |
CN109034898A (en) * | 2018-07-25 | 2018-12-18 | 河海大学常州校区 | A kind of BP neural network used car price evaluation algorithm based on improvement ant colony |
CN109726852A (en) * | 2018-11-30 | 2019-05-07 | 平安科技(深圳)有限公司 | Based on route planning method, device, terminal and the medium for improving ant group algorithm |
CN110375761A (en) * | 2019-08-07 | 2019-10-25 | 天津大学 | Automatic driving vehicle paths planning method based on enhancing ant colony optimization algorithm |
-
2019
- 2019-11-21 CN CN201911145132.2A patent/CN110909670B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103955743A (en) * | 2014-04-22 | 2014-07-30 | 江苏大学 | Ultrahigh-pressure water jet road mark line cleaning effect forecasting method and device |
US20170116477A1 (en) * | 2015-10-23 | 2017-04-27 | Nokia Technologies Oy | Integration of positional data and overhead images for lane identification |
CN109034898A (en) * | 2018-07-25 | 2018-12-18 | 河海大学常州校区 | A kind of BP neural network used car price evaluation algorithm based on improvement ant colony |
CN109726852A (en) * | 2018-11-30 | 2019-05-07 | 平安科技(深圳)有限公司 | Based on route planning method, device, terminal and the medium for improving ant group algorithm |
CN110375761A (en) * | 2019-08-07 | 2019-10-25 | 天津大学 | Automatic driving vehicle paths planning method based on enhancing ant colony optimization algorithm |
Non-Patent Citations (2)
Title |
---|
张乾燕;梁量;: "Gabor滤波和蚁群优化BP神经网络的人脸识别" * |
张曦;胡根生;梁栋;鲍文霞;: "基于时频特征的高分辨率遥感图像道路提取" * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112949617A (en) * | 2021-05-14 | 2021-06-11 | 江西农业大学 | Rural road type identification method, system, terminal equipment and readable storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN110909670B (en) | 2023-06-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109977812B (en) | Vehicle-mounted video target detection method based on deep learning | |
CN111444821B (en) | Automatic identification method for urban road signs | |
CN110796168B (en) | Vehicle detection method based on improved YOLOv3 | |
CN111274976B (en) | Lane detection method and system based on multi-level fusion of vision and laser radar | |
CN107609525B (en) | Remote sensing image target detection method for constructing convolutional neural network based on pruning strategy | |
CN111797716B (en) | Single target tracking method based on Siamese network | |
CN108921083B (en) | Illegal mobile vendor identification method based on deep learning target detection | |
CN112464911A (en) | Improved YOLOv 3-tiny-based traffic sign detection and identification method | |
CN111079602A (en) | Vehicle fine granularity identification method and device based on multi-scale regional feature constraint | |
CN111259796A (en) | Lane line detection method based on image geometric features | |
CN114627437B (en) | Traffic target identification method and system | |
CN113033687A (en) | Target detection and identification method under rain and snow weather condition | |
CN114581486A (en) | Template updating target tracking algorithm based on full convolution twin network multilayer characteristics | |
CN111274964B (en) | Detection method for analyzing water surface pollutants based on visual saliency of unmanned aerial vehicle | |
CN113065431A (en) | Human body violation prediction method based on hidden Markov model and recurrent neural network | |
CN116246169A (en) | SAH-Unet-based high-resolution remote sensing image impervious surface extraction method | |
CN117765480B (en) | Method and system for early warning migration of wild animals along road | |
CN115311449A (en) | Weak supervision image target positioning analysis system based on class reactivation mapping chart | |
CN114549909A (en) | Pseudo label remote sensing image scene classification method based on self-adaptive threshold | |
CN111260658A (en) | Novel depth reinforcement learning algorithm for image segmentation | |
CN113095164A (en) | Lane line detection and positioning method based on reinforcement learning and mark point characterization | |
CN110909670A (en) | Unstructured road identification method | |
CN113129336A (en) | End-to-end multi-vehicle tracking method, system and computer readable medium | |
CN117351702A (en) | Intelligent traffic management method based on adjustment of traffic flow | |
CN112085101A (en) | High-performance and high-reliability environment fusion sensing method and system |
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 | ||
GR01 | Patent grant |