A target extraction method based on the cat's eye effect

Abstract

This invention belongs to the field of laser detection technology and proposes a target extraction method based on the cat's eye effect. By controlling the laser switch, one frame of the image is in the laser-on state and the other is in the laser-off state. Using three adjacent frames, the frame difference between two adjacent frames is calculated and superimposed. The superposition of the two frame difference results makes the pixel value of the detected target position much larger than that of other positions, thereby extracting the corresponding threshold. In addition, the superposition of the two frame difference results can eliminate most interference and moving targets, improving the accuracy of target detection. The current input frame and the two frames before it are used as the threshold extraction scheme for the current frame. After the current frame is buffered, the corresponding threshold is obtained as the discrimination condition for its output. The overall implementation is delayed by one frame of image time, with small system latency. Moreover, this method is simple to design, easy to implement, and has wide applicability.

Description

Technical Field

[0001] This invention belongs to the field of laser detection technology, and in particular relates to a target extraction method based on the cat's eye effect. Background Technology

[0002] Currently, commonly used military optical observation and aiming devices, such as telescopes and sniper scopes, all share a characteristic: their optical windows exhibit a strong back-reflection characteristic along the original light path when reflecting parallel incident light beams. This back-reflection is typically stronger than the echo reflected from the surrounding diffuse background. This is similar to the strong reflectivity of a cat's eye at night, hence the term "cat's eye effect."

[0003] Under laser illumination, targets are detected and extracted through the "cat's eye effect." The echo intensity of this effect is much greater than the diffuse reflection intensity of objects in the environment. The target's position can be obtained by using the frame difference between laser-illuminated and unilluminated states. This method is simple to implement, has low latency, and is a commonly used technique. However, the simple two-frame difference can cause significant interference, especially during movement, generating many false targets. Furthermore, after the frame difference, a threshold comparison is needed to extract the target. Threshold settings are generally either automatic or manual. Automatic settings are usually based on experience, but this method is ineffective under different environments and lighting conditions. Manual settings require more human intervention under different environments and lighting conditions, making them inconvenient and cumbersome to operate. Summary of the Invention

[0004] To address the above problems, this invention provides a target extraction method based on the cat's eye effect. After detection begins, the laser is turned on in one frame and off in another, alternating in this manner. Using the middle frame of three adjacent frames as a reference, frame difference processing is performed on both sides, and the results of the two frame difference operations are superimposed to obtain the maximum value of the current pixel, which serves as the baseline threshold for target extraction. The superposition of the two frame difference operations can eliminate interfering targets. The "cat's eye effect" generated by the detected target (a target exhibiting the "cat's eye effect") is superimposed, resulting in a larger pixel value, thus ensuring the accuracy of the baseline threshold. Furthermore, this method has strong resistance to environmental factors and light intensity.

[0005] The technical solution adopted in this invention is: a target extraction method based on the cat's eye effect, with the following specific operation steps:

[0006] S1. By controlling the laser switch, the laser is turned on in one frame and turned off in another frame. The current input frame and the two frames before it are processed. The three frames are denoted as n1, n2 and n3 respectively. The frame difference is calculated based on n2 to obtain two matrices a = (n2-n1) and b = (n3-n2). The absolute values ​​of the two frame difference results are superimposed to obtain matrix c = (|a|+|b|).

[0007] S2. In step S1, the superposition of the two frame differences can not only eliminate most of the interference, but also make the value of the detected target position larger after superposition. Traversing the matrix c can obtain the maximum value q. The threshold is set as ψ*q, where ψ takes the value [0.5, 0.95].

[0008] S3. Even without a detected target, the threshold generated in step S2 can still yield the target. Therefore, a second threshold is set based on the average grayscale value p of the three frames. in Values ​​[1, 2], threshold It is used to determine whether there is a target in the current detection after the threshold in step S2 is met;

[0009] S4. At the end of the current frame buffer, the discrimination threshold ψ*q can be obtained from steps S2 and S3. As a criterion for the output of matrix c generated in step S1, and when the pixel value in matrix c is greater than or equal to ψ*q and If a target is detected at a certain location, it can be marked on the image or information can be reported. At the same time, a new threshold is generated and updated.

[0010] Furthermore, two thresholds ψ*q and Then, combining the image output, the frame difference between the n3rd frame and the n2nd and n4th frames is calculated and superimposed. The result is then compared with the threshold ψ*q and... Compare, and include those greater than or equal to ψ*q and The value is up to 1, less than ψ*q and The value is reduced to 0 for binarization. At the same time, a new threshold is obtained and updated according to steps s1-s3. The binarized image is subjected to morphological filtering, and the target is extracted by the connected component algorithm. The target extraction is based on the input time sequence of the current input frame, i.e., the n4th frame. The target extraction ends when the buffer of the n4th frame ends. When it is output, the target information can be marked at the corresponding position in the image. At the same time, the target can also report information through the communication interface.

[0011] The beneficial effects of this invention are as follows: By controlling the laser switch, one frame of the image is in the laser-on state and the other is in the laser-off state. Utilizing three adjacent frames, the method calculates the frame difference between adjacent frames and superimposes them. The superposition of the two frame differences results in the pixel value at the target location being much larger than at other locations, thereby extracting the corresponding threshold. Furthermore, the superposition of the two frame difference results can eliminate most interference and moving targets, improving the accuracy of target detection. Using the current input frame and its two preceding frames as the threshold extraction scheme for the current frame, the corresponding threshold is obtained after the current frame buffering is completed, serving as the output discrimination condition. The overall system achieves a time lag of one frame, resulting in low system latency. Moreover, this method is simple to design, easy to implement, and has wide applicability. Attached Figure Description

[0012] Figure 1 This is a flowchart of the automatic threshold extraction method in this invention.

[0013] Figure 2 This is a flowchart of an embodiment of the present invention that incorporates video output. Specific implementation methods

[0014] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0015] See Figure 1 , 2 This invention creates a target extraction method based on the cat's eye effect, with the following specific steps:

[0016] S1. By controlling the laser switch, the laser is turned on in one frame and turned off in another frame. The current input frame and the two frames before it are processed. The three frames are denoted as n1, n2 and n3 respectively. The frame difference is calculated based on n2 to obtain two matrices a = (n2-n1) and b = (n3-n2). The absolute values ​​of the two frame difference results are superimposed to obtain matrix c = (|a|+|b|).

[0017] S2. In step S1, the superposition of the two frame differences can not only eliminate most of the interference, but also make the value of the detected target position larger after superposition. Traversing the matrix c can obtain the maximum value q. The threshold is set as ψ*q, where ψ takes the value [0.5, 0.95].

[0018] S3. Even without a target to be detected (a target exhibiting the "cat's eye effect"), the threshold generated in step S2 can still yield the target. Therefore, a second threshold is set based on the average grayscale value p of the three frames. in Values ​​[1, 2], threshold It is used to determine whether there is a target in the current detection after the threshold in step S2 is met;

[0019] S4. At the end of the current frame buffer, the discrimination threshold ψ*q can be obtained from steps S2 and S3. As a criterion for the output of matrix c generated in step S1, and when the pixel value in matrix c is greater than or equal to ψ*q and If a target is detected at a certain location, it can be marked on the image or information can be reported. At the same time, a new threshold is generated and updated.

[0020] See Figure 1 , Figure 2 Furthermore, two thresholds ψ*q and Then, combining the image output, the frame difference between the n3rd frame and the n2nd and n4th frames is calculated and superimposed. The result is then compared with the threshold ψ*q and... Compare, and include those greater than or equal to ψ*q and The value is up to 1, less than ψ*q and The value is reduced to 0 for binarization. At the same time, a new threshold is obtained and updated according to steps s1-s3. The binarized image is subjected to morphological filtering, and the target is extracted by the connected component algorithm. The target extraction is based on the input time sequence of the current input frame, i.e., the n4th frame. The target extraction ends when the n4th frame buffer ends. When it is output, the target information can be marked at the corresponding position in the image. At the same time, the target can also report information through the communication interface. In the whole process, the video is equivalent to the delay of one frame buffer. The latency is low, and the threshold is updated in real time with the video frame. No other operation is required. The overall process design is simple and easy to implement.

Claims

1. A cat-eye effect based target extraction method, characterized in that, Operating steps: S1. By controlling the laser switch, the laser is turned on in one frame and turned off in another frame. The current input frame and the two frames before it are processed. The three frames are denoted as n1, n2 and n3 respectively. The frame difference is calculated based on n2 to obtain two matrices a=(n2-n1) and b=(n3-n2). The absolute values ​​of the two frame difference results are superimposed to obtain matrix c=(|a|+|b|). S2. In step S1, the superposition of the two frame differences can not only eliminate most of the interference, but also make the value of the detected target position larger after superposition. Traversing the matrix c can obtain the maximum value q. The threshold is set as ψ×q, where ψ takes the value [0.5, 0.95]. S3. Under the condition that there is no target to be detected, the corresponding target can still be obtained by the threshold generated in step S2. In this regard, according to the average gray value p of the three frames of images, a second threshold φ×p is set, where φ takes the value [1,2]. The threshold φ×p is used to determine whether there is a target to be detected in the current detection after the threshold in step S2 is met. S4. When the current frame buffer ends, the discrimination thresholds ψ×q and φ×p can be obtained from steps S2 and S3, which are used as the discrimination conditions of the matrix c generated by step S1 when it is output. When the pixel value in matrix c is greater than or equal to ψ×q and φ×p, it is considered that there is a target at that position, and the image can be marked or the information can be reported. At the same time, a new threshold is generated and updated. After obtaining two thresholds ψ×q and φ×p, the frame difference of the n3rd frame is calculated and superimposed with the n2th and n4th frames, and the result is compared with the thresholds ψ×q and φ×p. Values ​​greater than or equal to ψ×q and φ×p are set to 1, and values ​​less than ψ×q and φ×p are set to 0 for binarization. At the same time, new thresholds are obtained and updated according to steps s1-s3. The binarized image is subjected to morphological filtering, and the target is extracted by the connected component algorithm. The target extraction is based on the input time sequence of the current input frame, i.e., the n4th frame. The target extraction ends when the buffer of the n4th frame ends. When it is output, the target information can be marked at the corresponding position in the image. At the same time, the target can also report information through the communication interface.

Images