Data enhancement method based on depth image

Abstract

The invention provides a data enhancement method based on a depth image, which is applicable to the field of computer vision and algorithms based on depth image recognition, target detection, behaviorrecognition and the like. The method is mainly composed of a pixel coordinate to three-dimensional point cloud part, a three-dimensional point cloud space conversion part, a three-dimensional point cloud to pixel coordinate part and a minimum value filtering processing part. The step of converting a pixel coordinate system into the three-dimensional point cloud is to convert the plane pixel coordinate points in the depth image into the three-dimensional space point cloud under the world coordinate system through the conversion relationship among the pixel coordinate system, the image coordinate system, the camera coordinate system and the world coordinate system. In three-dimensional point cloud space transformation, the depth image is converted into three-dimensional space point cloud, and random translation transformation and random rotation transformation are carried out on the three-dimensional space point cloud to form new three-dimensional space point cloud. The newly generatedthree-dimensional space point cloud is projected into the depth image through the conversion relationship between the world coordinate system and the pixel coordinate system. A new depth image after data enhancement is obtained after minimum filtering processing. The data enhancement method provides a data expansion method for research based on depth images in the computer vision neighborhood. According to the method, the generalization ability of the network model can be greatly improved.

Description

Technical field: [0001] The invention proposes a data enhancement method based on a depth image, which is applicable to the field of computer vision, and algorithms such as recognition, target detection, and behavior recognition based on the depth image. Background technique: [0002] In recent years, deep learning has been widely used in the field of computer vision. The excellent performance of deep learning in the face of many problems in the field of computer vision has led more and more researchers to set foot in this research direction. The reason why deep learning can achieve such excellent performance is because the deep convolutional network itself has a strong expressive ability, and can train the required model results according to the training objectives. However, because of this, the network model itself requires a large amount of even massive data to drive model training, otherwise it may cause the model to fall into the predicament of overfitting. However, i...

AI Technical Summary

Problems solved by technology

However, the common data enhancement methods applied in the field of RGB images are not suitable for depth images.

Since the value stored in each pixel of the depth image is the depth distance from the position to the camera, directly using the data enhancement method in the RGB image will cause the image to be distorted

Images