An Image Feature Extraction Method Based on Convolutional Autoencoder Model
A technology of image feature extraction and convolutional self-encoding, which is applied in the field of image processing of self-driving vehicles, can solve problems such as mutual influence, whether there are others, information is not fully utilized, and a large number of training samples, so as to reduce losses, image acquisition and Ease of handling and reduced workload
Active Publication Date: 2021-06-08
TSINGHUA UNIV
View PDF4 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
This method requires a large number of training samples, and requires a lot of manpower for calibration. At the same time, because the relatively important information is artificially screened and extracted, some information that may be useful for subsequent decision-making is not fully utilized, and the output information is only Including physical quantities such as distance and location, but more advanced semantic information that is not intuitively understood, such as whether there is mutual influence between surrounding road participants, whether there are other road participants' behaviors, etc., will affect subsequent decision-making. , The control process puts forward higher requirements
In addition, the noise reduction self-encoder can also be used to extract image features, Figure 7 is a schematic diagram of the training process of the traditional denoising autoencoder model. Its training set uses images containing irrelevant features as input, and uses images without noise as labels to train the denoising autoencoder model. Noise-free images are difficult to obtain, and The extracted feature part needs to be completely consistent with the original image, which also brings great difficulty to the acquisition of the training set
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0034] In the following, with reference to the drawings and embodiments, the image feature extraction method for the autoencoder model will be described in detail through the embodiment of extracting the self-driving vehicle.
[0035] This method is a method based on the autoencoder model, and the method architecture is as follows figure 1 As shown, in the embodiment, the convolutional autoencoder model formed by combining the convolutional neural network model and the autoencoder model is used to realize the extraction of vehicle features in the road image.
[0036] This research will be carried out based on the method of deep learning, and the implementation steps of this method are summarized as follows:
[0037] 1. Acquisition of surrounding images containing vehicles;
[0038] 2. Acquisition of surrounding images that have nothing to do with the vehicle;
[0039] 3. Write the convolutional autoencoder model code;
[0040] 4. Train the convolutional autoencoder model; ...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention discloses a method for extracting image features based on a convolutional self-encoder model. The method comprises: step 1, collecting images containing features to be extracted to form a data set I; step 2, collecting images not containing features to be extracted The picture constitutes the data set II; step 3, use the data set I and data set II to train the self-encoder model at the same time; step 4, input the image to be extracted, and the output of the encoder part of the auto-encoder model is extracted features. Compared with the current mainstream image feature extraction methods, this method does not require manual calibration, which can reduce the workload of manual calibration while ensuring the reliability of feature extraction, and at the same time make the output features contain more advanced semantic features.
Description
technical field [0001] The invention relates to the technical field of image processing of self-driving vehicles, in particular to an image feature extraction method based on a convolutional autoencoder. Background technique [0002] At present, in the image processing technology of self-driving vehicles, the image feature extraction technology based on deep learning mainly uses convolutional neural networks, combined with classifiers, for image recognition of vehicles, pedestrians, cyclists, and road signs. It is necessary to prepare a large number of training samples, including the input image and the true value of the final output, and train through the method of supervised learning. This method requires a large number of training samples, and requires a lot of manpower for calibration. At the same time, because the relatively important information is artificially screened and extracted, some information that may be useful for subsequent decision-making is not fully utili...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & AuthorityPatents(China)
IPC IPC(8): G06K9/00G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06V20/56G06N3/045G06F18/2413
Inventor罗禹贡王庭晗李克强余大蒙刘金鑫杨殿阁王建强连小珉郑四发李升波
OwnerTSINGHUA UNIV