Shadow removal method for moving object detection with color invariant feature in emergency safety scene

Abstract

The invention provides a shadow removal method for detecting a moving object with a color invariant feature in an emergency safety scene. The shadow removal method comprises the following steps: firstly, connecting a national standard video platform to access a related video stream; carrying out decoding analysis on the video stream to obtain decoded image frame data; foreground and background pixels of the moving target in the monitoring scene are classified through a background modeling algorithm; continuously classifying illumination shadow pixels and normal motion pixels in the foreground pixels through color invariant features in the motion foreground pixels; and finally extracting a completed moving target pixel point. When the method is used for classifying and removing the shadow moving target, the accuracy is high, the practicability is high, and the method is better than the scene application of more shadow removal-free methods.

Description

Technical Field

[0001] The present invention relates to the technical field of computer vision and video surveillance, and particularly relates to a method for removing shadows of moving objects with color-invariant features in an emergency security scenario, which is applicable to scenarios such as video surveillance and intelligent target tracking in emergency security scenarios. Background Art

[0002] In video surveillance, shadows cast by moving objects are often misclassified as foreground targets, resulting in distorted target shapes, multi-target adhesion, and even target loss. Although there are many existing shadow removal methods, they have the following problems:

[0003] 1. The shadow segmentation threshold depends on empirical settings and lacks robustness;

[0004] 2. The color space conversion is complex and the real-time performance is poor;

[0005] 3. It is difficult to accurately separate the mixed area of shadows and foreground targets. Summary of the Invention

[0006] A shadow removal method combining ViBe background modeling, C1C2C3 color-invariant features, and physical shadow theory is proposed to solve the problems of high shadow misdetection rate and insufficient real-time performance in the prior art. To achieve the above object, a method for removing shadows of moving objects with color-invariant features in an emergency security scenario of the present application includes the following steps:

[0007] S1. Access the video stream through the national standard GB / T28181 and decode it into image frame data;

[0008] S2. Use a background modeling algorithm to segment the foreground and background in the image;

[0009] S3. Calculate the color-invariant features and convert the RGB color space to the C1C2C3 color space;

[0010] S4. Set a threshold to distinguish shadow and non-shadow pixel points in the C1C2C3 color space;

[0011] S5. Repair the foreground pixels misjudged as shadows to obtain the complete moving detection target of each frame of image;

[0012] S6. The shadow removal of the moving object detection with color-invariant features is completed.

[0013] Beneficial effects: The shadow removal method for detecting moving objects with color - invariant features in an emergency security scenario of the present invention effectively solves the problem of accurately extracting moving objects in the emergency security scenario. Through the accurate extraction of moving objects, it can be applied and analyzed in important algorithm directions such as intelligent security and moving object behavior tracking. Through practical application, the moving target detection after removing shadows can improve the fault - tolerance rate of the algorithm tracking and can also better analyze the complete recognition target effectively. Brief Description of the Drawings

[0014] Figure 1 It is the flowchart of the method steps of the present invention; Detailed Embodiment

[0015] In order to more clearly describe the technical solutions in the embodiments of the present invention, the following will describe the present invention in detail with reference to the accompanying drawings in the embodiments. The specific embodiments described in the present invention, as well as other embodiments based on the present invention, all fall within the protection scope of the present invention.

[0016] The object of the present invention is a shadow removal method for detecting moving objects with color - invariant features in an emergency security scenario. This method mainly includes three modules: background classification before motion detection, transformation from RGB to C1C2C3 color space, and intersection of moving foreground and non - shadow areas.

[0017] As Figure 1 shown is the flowchart of the steps of a shadow removal method for detecting moving objects with color - invariant features in an emergency security scenario according to an embodiment of the present invention.

[0018] A shadow removal method for detecting moving objects with color - invariant features provided by an embodiment of the present invention includes the following steps:

[0019] Step S1. Access the video stream through the national standard GB / T28181 and decode it into image frame data.

[0020] Step S2. Classify the foreground and background pixels of moving objects in the image.

[0021] S21. Model the background to randomly store 20 historical samples for each pixel, and judge whether the current pixel is the foreground through the Euclidean distance;

[0022] S22. If the distance between the current pixel value v(x) and at least two background samples is less than the threshold R, it is determined as the background, otherwise it is the foreground;

[0023] S23. Randomly replace the background samples in the model and spread the update to the neighboring pixels to improve the anti - camera - jitter ability.

[0024] Step S3. Detect the shadow pixels with color - invariant features.

[0025] S31. Convert the RGB space to the lighting-invariant C1C2C3 space:

[0026]

[0027] S32. Construct reference image RGB ref is the background sample mean:

[0028]

[0029] Among them, Bgs is the RGB channel background sample, n is the sample data size, RGB ref is the mean of background samples of three channels;

[0030] S33. Calculate the difference between the current pixel and the reference pixel in the C1C2C3 space:

[0031]

[0032] Among them, Th=(Th1, Th2, Th3) is an adaptive threshold used to distinguish between shadows and foregrounds.

[0033] Step S4: In the moving foreground image, the foreground pixels that are misjudged as shadows are repaired. If a shadow pixel is detected in the moving foreground pixel, the pixel is removed.

[0034] Step S5: At this point, the shadow removal of the moving object detection with color-invariant features is completed.

Claims

1. A method for removing shadows in the detection of moving objects with color-invariant features in an emergency safety scenario, comprising the following steps: (1) Access the video stream through the national standard GB28181 and decode it into image frame data; (2) Use a background modeling algorithm to segment the foreground and background in the image for classification; (3) Calculate color-invariant features and transform the RGB color space into the C1C2C3 color space; (4) Set a threshold to distinguish shadow and non-shadow pixel points in the C1C2C3 color space; (5) Repair the foreground pixels misjudged as shadows to obtain the complete moving detection target for each frame of image; (6) The shadow removal in the detection of moving objects with color-invariant features is completed.

Images