Unlock instant, AI-driven research and patent intelligence for your innovation.

Road target detection method and device, electronic equipment and storage medium

A target detection and target technology, applied in the direction of instruments, calculations, character and pattern recognition, etc., can solve the problems of low accuracy and low efficiency of road targets

Pending Publication Date: 2020-04-24
熊猫汽车(上海)有限公司
View PDF5 Cites 17 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0007] Embodiments of the present invention provide a road target detection method, device, electronic equipment, and storage medium to solve the problem that the existing model training process for road target detection is inefficient and the accuracy of road target detection is also low The problem

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 more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Road target detection method and device, electronic equipment and storage medium
  • Road target detection method and device, electronic equipment and storage medium
  • Road target detection method and device, electronic equipment and storage medium

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0063] figure 1 A schematic diagram of the road target detection process provided by the embodiment of the present invention, the process includes the following steps:

[0064] S101: For each sample image in the training set, input the sample image and corresponding label information into the target detection model; wherein the label information records the coordinate information and category of the real frame of the target.

[0065] The model training method provided by the embodiment of the present invention is applied to an electronic device, and the electronic device may be a device such as a PC or a tablet computer, or may be a server.

[0066] A training set for training the model is pre-stored in the electronic device, and each sample image in the training set has corresponding label information.

[0067] Specifically, a txt file can be used to record the label information. The label information includes the coordinate information and category of the target real frame....

Embodiment 2

[0081] In order to avoid the model overfitting phenomenon caused by too few sample images, on the basis of the above-mentioned embodiments, in the embodiment of the present invention, for each sample image in the training set, the sample image and the corresponding label information Before inputting the target detection model, the method also includes:

[0082] Perform sample enhancement processing on the sample images in the training set to generate new sample images; wherein, the sample enhancement processing includes randomly increasing or reducing the size of the sample images, performing random probability horizontal flipping on the sample images, and performing random probability horizontal flip on the sample images. The brightness is randomly adjusted, the chroma of the sample image is randomly adjusted, and the contrast of the sample image is randomly adjusted.

[0083] In the embodiment of the present invention, the sample images in the training set are enriched by pe...

Embodiment 3

[0089] In the process of training the model, the anchor box needs to be determined in advance, and the target detection model calculates the predicted category and offset of the predetermined anchor box, adjusts the position of the anchor box, and outputs the predicted box of the sample image.

[0090] In the embodiment of the present invention, the process of pre-determining the anchor frame includes:

[0091] The number of anchor boxes is set in advance, and the Kmeans clustering algorithm is used to cluster the real boxes of the sample images in the training set to obtain the anchor boxes of the target detection model, where the distance between the real box and the cluster center box in the clustering process is expressed as d=1-IoU.

[0092] In the embodiment of the present invention, the Kmeans clustering algorithm is used to cluster the real frames of the sample images in the training set, and the number of preset anchor frames is the K value in the Kmeans clustering al...

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

No PUM Login to View More

Abstract

The invention discloses a road target detection method and a device, electronic equipment and a storage medium. The method comprises the steps of inputting a to-be-detected image into the target detection model to detect a road target; in model training, calculating a loss value by adopting LrIOU =-In (rIOU); the original loss function calculation which is respectively carried out on the four coordinate offsets is replaced; the problem that the frame is inaccurate is solved; rIoU=0.5*(IoU+U / C), it can be known that the range of the rIOU is larger than 0 and smaller than or equal to 1, the problem that when a prediction frame and a real frame are not overlapped, the IOU is always 0, and a model cannot be optimized is solved, the loss function gradient is gradually increased along with reduction of the rIOU, a loss function is more reasonable, and the convergence speed of coordinate regression is increased. Therefore, the model training process for road target detection provided by the embodiment of the invention is relatively high in efficiency, and the accuracy of road target detection is relatively high.

Description

technical field [0001] The present invention relates to the technical field of road target detection, in particular to a road target detection method, device, electronic equipment and storage medium. Background technique [0002] Target detection is one of the hottest directions in the field of machine vision in recent years. Road target detection can be applied in many real-life scenarios, such as unmanned driving, security, etc. Perform classification recognition. [0003] Road object detection methods based on deep learning models in the prior art are generally divided into two categories, namely second-order detection algorithms and first-order detection algorithms. The second-order detection algorithm, which divides the detection process into two stages, first generates a series of candidate regions, then classifies the candidate regions and fine-tunes the positions of the candidate frames, so it has a higher average precision (mAP, Mean Average Precision) indicator. ...

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
no application Login to View More
IPC IPC(8): G06K9/62
CPCG06V2201/07G06F18/23213G06F18/214
Inventor 陈海波
Owner 熊猫汽车(上海)有限公司