A method for simulating and generating dirty images based on pixel point fusion

Abstract

The present invention relates to a method for simulating and generating a dirty image based on pixel point fusion, and the method includes: building an image data acquisition system; the image data acquisition system acquires a source dirty image for simulating and generating a dirty image; obtaining a target scene image for simulating and generating dirt; performing pixel point fusion on the source dirty image and the target scene image to simulate and generate dirty image data collected when the surface of a camera optical window is dirty. Compared with the prior art, the method of the present invention can conveniently and quickly simulate and generate a dirty image.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and more particularly to a method for simulating and generating a dirty image based on pixel point fusion. Background Art

[0002] In the field of intelligent driving, cameras are widely used as a key sensor in the perception field. However, since cameras are widely arranged outside the vehicle body, the optical window is easily covered with dirt, which may cause a serious decline or even complete loss of the perception function. At the same time, the image data collected by cameras are often used as deep learning data sets, and many data sets need to appropriately add dirty image data to enhance the robustness during the supervised learning process of the algorithm model. Meanwhile, the development of a cleaning system for the surface of the vehicle-mounted camera optical window is extremely urgent, and the analysis and recognition of the pollution state of the camera optical window surface are extremely important. Simulating and generating dirty images is conducive to quantitatively analyzing the impact of dirt on image data and shortening the development cycle. Summary of the Invention

[0003] An object of the present invention is to provide a method for simulating and generating a dirty image based on pixel point fusion to overcome the defects existing in the above-mentioned prior art.

[0004] The object of the present invention can be achieved by the following technical solutions:

[0005] A method for simulating and generating a dirty image based on pixel point fusion, the method comprising:

[0006] Constructing an image data acquisition system;

[0007] The image data acquisition system acquires a source dirty image for simulating and generating a dirty image;

[0008] Obtaining a target scene image for simulating and generating dirt;

[0009] Performing pixel point fusion on the source dirty image and the target scene image to simulate and generate dirty image data collected when the surface of the camera optical window is dirty.

[0010] Preferably, the image data acquisition system includes:

[0011] A light box: providing an acquisition environment with a constant light intensity;

[0012] A camera: used for acquiring images;

[0013] A specimen: used for being arranged in front of the camera optical window when acquiring the source dirty image, and the specimen is provided with dirt, and the specimen with dirt is used for simulating the dirt on the surface of the camera optical window;

[0014] Background image: The background when the camera captures images.

[0015] Preferably, the specimen includes a transparent acrylic plate.

[0016] Preferably, the source dirty image for the image data acquisition system to simulate and generate a dirty image includes:

[0017] Turn on the LED lights in the light box to ensure a constant illumination environment inside the light box;

[0018] Set the background image and calibrate the camera focal length;

[0019] Ensure that the position and focal length of the camera remain unchanged, and set the background image to a pure white background;

[0020] Place the prepared specimen in front of the optical window of the camera;

[0021] The camera captures an image to obtain the source dirty image.

[0022] Preferably, calibrating the camera focal length specifically means: setting the background image as a checkerboard, performing a focusing operation on the camera so that the camera can clearly capture the contour boundaries in the checkerboard, to ensure that the distance between the camera and the background image is the camera focal length L.

[0023] Preferably, in this method, the position where the dirt appears in the image data is simulated by moving the relative position of the specimen with respect to the camera.

[0024] Preferably, the process of pixel point fusion of the source dirty image and the target scene image includes:

[0025] Determine the weight values of each point of the source dirty image and the target scene image according to the pixel values of each pixel point in the source dirty image and the target scene image;

[0026] Perform a weighted sum of the pixel values of each pixel point in the source dirty image and the target scene image to obtain the pixel value of the corresponding pixel point of the produced dirty image.

[0027] Preferably, the determination method of the weight value is specifically:

[0028] For the pixel point (x, y) in the source dirty image, its pixel value is The weight is determined as

[0029]

[0030] For the pixel point (x, y) in the target scene image, its pixel value is The weight is determined as

[0031]

[0032] Among them, p is the pixel threshold.

[0033] Preferably, the pixel value of each pixel point of the simulated dirty image is determined by the following formula:

[0034]

[0035] Among them, is the pixel value of the pixel point (x, y) in the simulated dirty image.

[0036] Preferably, the size of the pixel threshold p is selected according to different types of dirt during the determination of the weight value.

[0037] Compared with the prior art, the present invention has the following advantages:

[0038] (1) The present invention can quickly and effectively simulate dirty images in various target scenarios;

[0039] (2) When the present invention performs image fusion based on image pixels, the weight information is calculated and fused for each pixel point to achieve the purpose of simulating reality. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a flowchart of a method for simulating and generating a dirty image based on pixel point fusion according to the present invention;

[0041] Figure 2 is a schematic structural diagram of an image data acquisition system according to the present invention;

[0042] Figure 3 is a schematic diagram of a checkerboard for calibrating the focal length of a camera according to the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0043] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. Note that the following description of the embodiments is only illustrative in nature, and the present invention is not intended to limit the objects or uses to which it is applicable, and the present invention is not limited to the following embodiments.

[0044] Embodiment

[0045] As Figure 1 shown, this embodiment provides a method for simulating and generating a dirty image based on pixel point fusion, and the method includes:

[0046] Build an image data acquisition system;

[0047] The image data acquisition system acquires the source dirty image for simulating and generating the dirty image;

[0048] Obtain a target scene image for simulating the generation of dirt. The target scene image is the scene image data collected when the camera optical window is clean, and is used to simulate the image data collected when the camera optical window is dirty;

[0049] Perform pixel fusion on the source dirt image and the target scene image to simulate the dirt image data collected when the surface of the camera optical window is dirty.

[0050] The present invention mainly includes two parts: First, an image data acquisition system, which is used to collect the source image data for simulating the generation of dirt images. Its characteristic is that the illumination intensity inside the light box is constant, excluding the influence of external light. At the same time, when collecting images, the background is pure white to ensure that the influence of the background image is excluded during pixel fusion of the images. Second, image fusion based on image pixel points. This fusion method sets weight information for pixel points according to the value of each pixel point. The smaller the value of the pixel point in the source image, the greater its weight information, so as to achieve the purpose of well simulating the image data collected when the camera optical window is dirty during image fusion.

[0051] The following specifically describes the two parts of the present invention:

[0052] I. Image data acquisition system

[0053] As Figure 2 shown, the image data acquisition system includes:

[0054] Light box: Provide a collection environment with a constant illumination intensity;

[0055] Camera: Used to collect images;

[0056] Specimen: Used to be set in front of the camera optical window when collecting the source dirt image. There is dirt on the specimen. The specimen with dirt is used to simulate the dirt on the surface of the camera optical window. The specimen is made of transparent acrylic board, and dirt is set on it. The shape and size of the dirt can be set according to needs, and the relative position of the specimen to the camera can be moved to simulate the position where the dirt appears in the image data;

[0057] Background image: Serve as the background when the camera performs image collection.

[0058] The specific process of collecting the source dirt image for simulating the generation of dirt images based on the above image data acquisition includes:

[0059] Turn on the LED lights in the light box to ensure a constant illumination environment inside the light box;

[0060] Set the background image and calibrate the camera focal length: Set the background image as a checkerboard. The checkerboard image is as Figure 3As shown in the figure, perform a focusing operation on the camera so that the camera can clearly capture the contour boundaries in the checkerboard, ensuring that the distance between the camera and the background image is the camera focal length L;

[0061] Ensure that the position and focal length of the camera remain unchanged, and set the background image to a pure white background;

[0062] Place the prepared specimen in front of the optical window of the camera;

[0063] The camera captures an image to obtain the source dirty image.

[0064] II. Image Fusion Based on Image Pixel Points

[0065] The present invention performs fusion on image data with a pixel value range of 0 to 255. When performing image fusion based on image pixels, the weight information is calculated and fused for each pixel point to achieve the purpose of simulating reality.

[0066] The process of pixel point fusion between the source dirty image and the target scene image includes:

[0067] Determine the weight values of each point in the source dirty image and the target scene image according to the pixel values of each pixel point in the source dirty image and the target scene image;

[0068] Perform weighted summation on the pixel values of each pixel point in the source dirty image and the target scene image to obtain the pixel value of the corresponding pixel point of the generated dirty image.

[0069] The specific method for determining the weight value is as follows:

[0070] For the pixel point (x, y) in the source dirty image, its pixel value is The weight is determined as

[0071]

[0072] For the pixel point (x, y) in the target scene image, its pixel value is The weight is determined as

[0073]

[0074] Where p is the pixel threshold. In this embodiment, p is taken as 180, and other thresholds can be selected according to different types of dirt.

[0075] The pixel value of each pixel point of the simulated generated dirty image is determined by the following formula:

[0076]

[0077] Where, is the pixel value of the pixel point (x, y) in the simulated dirty image.

[0078] The above embodiments are only examples and do not represent a limitation on the scope of the present invention. These embodiments can also be implemented in various other ways, and various omissions, substitutions, and changes can be made without departing from the technical idea of the present invention.

Claims

1. A method for simulating and generating a dirty image based on pixel point fusion, characterized in that, the method includes: Construct an image data acquisition system; The image data acquisition system acquires a source dirty image for simulating and generating a dirty image; Obtain a target scene image for simulating and generating dirt; Perform pixel point fusion on the source dirty image and the target scene image to simulate and generate dirty image data collected when the surface of the camera optical window is dirty; The image data acquisition system includes: Light box: Provide a collection environment with a constant light intensity; Camera: Used to collect images; Specimen: Used to be set in front of the camera optical window when collecting the source dirty image. There is dirt on the specimen, and the specimen with dirt is used to simulate the dirt on the surface of the camera optical window; Background image: Serve as the background when the camera performs image acquisition; The process of pixel point fusion of the source dirty image and the target scene image includes: Determine the weight values of each point of the source dirty image and the target scene image according to the pixel values of each pixel point in the source dirty image and the target scene image; Perform weighted summation on the pixel values of each pixel point in the source dirty image and the target scene image to obtain the pixel value of the corresponding pixel point of the generated dirty image; The specific way to determine the weight value is: For the pixel point (x, y) in the source dirty image, its pixel value is The weight is determined as For the pixel point (x, y) in the target scene image, its pixel value is The weight is determined as where p is the pixel threshold; The pixel value of each pixel point of the simulated and generated dirty image is determined by the following formula: Among them, is the pixel value of the pixel point (x, y) in the simulated dirty image; When determining the weight value, the size of the pixel threshold p is selected according to different dirt types.

2. A method for simulating and generating a dirty image based on pixel point fusion according to claim 1, characterized in that, the specimen includes a transparent acrylic plate.

3. A method for simulating and generating a dirty image based on pixel point fusion according to claim 1, characterized in that, The image data acquisition system acquires a source dirty image for simulating and generating a dirty image, including: Turn on the LED lights in the light box to ensure a constant light environment inside the light box; Set the background image and calibrate the camera focal length; Ensure that the position and focal length of the camera remain unchanged, and set the background image to a pure white background; Place the fabricated specimen in front of the camera optical window; The camera takes an image to obtain the source dirty image.

4. A method for simulating and generating a dirty image based on pixel point fusion according to claim 3, characterized in that, Calibrating the camera focal length specifically means: setting the background image as a checkerboard, and performing a focusing operation on the camera so that the camera can clearly capture the contour boundaries in the checkerboard to ensure that the distance between the camera and the background image is the camera focal length L.

5. A method for simulating and generating a dirty image based on pixel point fusion according to claim 1, characterized in that, In this method, the position of the dirt appearing in the image data is simulated by moving the relative position of the specimen relative to the camera.

Images