High-reflectivity target detection and identification method based on difference image multi-factor collaborative discrimination

Abstract

The invention relates to a high-reflectivity target detection and recognition method based on differential image multi-factor collaborative discrimination, and belongs to the technical field of image processing and mode recognition. The method comprises the steps of performing filtering preprocessing on continuous frames of active and passive images, performing differential processing on the two images after registration, and preliminarily extracting a possible target area; binarizing the difference image according to the gray value distribution of the difference image, and extracting a target candidate region; performing feature analysis on the extracted target candidate region; calculating the average intensity of each target candidate area in the difference image; and screening the extracted target candidate areas according to set conditions, screening an area with the maximum average intensity and variance from the screened areas, marking a finally determined target area on the original image, and outputting a detection result. According to the method, the problem that the detection precision is reduced due to the fact that a single-feature target detection method is easily interfered by background noise is solved, and accurate detection of the target is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of image processing and pattern recognition, and in particular relates to a high-reflectivity target detection and recognition method based on differential image multi-factor collaborative discrimination. Background Art

[0002] Although a variety of methods have been proposed and applied in existing image target detection technologies, there are still many challenges in target detection in complex backgrounds. Traditional target detection methods mainly rely on single features, such as shape, color, or texture. These methods perform well when the background is simple and the target features are obvious. However, in complex backgrounds, single-feature methods are easily interfered by background noise, resulting in a decrease in detection accuracy, and traditional methods often have difficulty in accurate recognition. In recent years, target detection methods based on deep learning have made significant progress, but these methods usually require a large amount of training data and computing resources, and perform poorly in scenarios with high real-time requirements. Therefore, it is of great significance to develop a high-definition target detection and recognition method based on multi-factor collaborative discrimination of differential images. Summary of the Invention

[0003] The purpose of the present invention is to overcome the defects of the prior art and provide a high-reflection target detection and recognition method based on multi-factor collaborative discrimination of differential images.

[0004] The principle of the present invention is as follows: The present invention effectively suppresses background noise and highlights the target area through image filtering, image registration, and image difference. Comprehensively analyze the various features of the target candidate area in the differential image (such as area, eccentricity, roundness, etc.), and combine the average intensity and variance for screening. First, extract the regional features and calculate the area for preliminary screening, then further screen the candidate area based on the area, eccentricity and roundness, and finally determine the target area and mark it through the dual screening mechanism (average intensity and variance) of the candidate area in the differential image. This method maintains high detection accuracy under complex backgrounds while reducing dependence on computing resources, and is suitable for application scenarios with high real-time requirements.

[0005] The hardware equipment conditions that this invention relies on are as follows:

[0006] Active and passive image acquisition system, including a laser and an ordinary RGB industrial camera.

[0007] To achieve the above-mentioned purpose, the technical solution of the present invention is: a high-reflectivity target detection and recognition method based on multi-factor collaborative discrimination of differential images, which comprehensively analyzes multiple features of the target candidate area, calculates the average intensity of each target candidate area in the difference image, evaluates the activity level of the target candidate area, and realizes accurate detection of high-reflectivity targets in complex scenes through conditional screening.

[0008] Furthermore, the various features include area, eccentricity, and circularity.

[0009] Furthermore, the target candidate area is obtained by filtering and preprocessing the active and passive images of consecutive frames, and then performing difference processing after aligning the two images to obtain a difference image, binarizing the difference image according to the gray value distribution of the difference image, and extracting the target candidate area.

[0010] Furthermore, the differential image binarization is performed based on an adaptively determined threshold.

[0011] Furthermore, continuous frame active and passive images are acquired based on a laser detection system.

[0012] Furthermore, the specific implementation steps of this method are as follows:

[0013] Step S1: using a laser detection system to obtain continuous active and passive images of the target scene;

[0014] Step S2: pre-process the active and passive images of consecutive frames using Gaussian filtering, perform image difference after registration, and obtain a differential image;

[0015] Step S3: Based on the differential image, adaptively determine the threshold value, perform image binarization processing to obtain a binary image;

[0016] Step S4: extracting target candidate regions and their features in the binary image, where the features include area, eccentricity, and roundness;

[0017] Step S5: Based on the binary image, traverse all target candidate regions, perform preliminary screening based on area, eccentricity and roundness, and exclude target candidate regions that do not meet the preset conditions;

[0018] Step S6: Based on the initially screened target candidate regions, the radius is calculated according to the region area in the binary image, the coordinate range of each screened region is obtained, and the image grayscale data within each coordinate range is correspondingly extracted in the differential image;

[0019] Step S7: Based on the differential image data obtained in step S6, calculate the average intensity of each region, select the regions with average intensity greater than a preset value, further calculate the grayscale variance of these regions, and select the region with the largest variance;

[0020] Step S8: Based on the screened area, mark the final target area on the active image.

[0021] Furthermore, the step S2 is specifically as follows:

[0022] Step S21: Perform Gaussian filtering on the active and passive images to smooth the details in the images and reduce the background noise of the images;

[0023] Step S22, using SURF algorithm to detect the active and passive images, extracting feature points and feature descriptors in the images, matching the feature descriptors of the two images, finding corresponding feature point pairs, estimating the geometric transformation relationship between the images according to the matched feature point pairs, and applying the estimated geometric transformation to register the images;

[0024] Step S23, performing difference on the registered active and passive images, and the difference formula is:

[0025] I 差 =I 主 -I 被

[0026] In the formula, I 主 represents the image obtained under laser active illumination, i.e., the active image, and I 被 represents the image obtained without laser active illumination, i.e., the passive image.

[0027] Further, the step S5 specifically includes:

[0028] Step S51, screening the regions with a circularity greater than 0.8 from the features of the target candidate regions extracted in step S4, to obtain regions with a shape close to a circle;

[0029] Step S52, excluding regions with an area too small or too large from the regions screened in step S51 by setting upper and lower limits of the area of the regions;

[0030] Step S53, screening regions with a regular shape from the regions screened in step S52 by setting an eccentricity range of 0.8 to 1.25.

[0031] The application further provides a high-reflective target detection and recognition system based on differential image multi-factor collaborative discrimination, which includes a memory, a processor, and computer program instructions stored in the memory and capable of being executed by the processor, and when the processor executes the computer program instructions, the steps of any of the above methods can be realized.

[0032] The application further provides a computer readable storage medium, which stores computer program instructions capable of being executed by a processor, and when the processor executes the computer program instructions, the steps of any of the above methods can be realized.

[0033] Compared with the prior art, the application has the following beneficial effects:

[0034] The application provides a high-reflection target detection and recognition method based on differential image multi-factor collaborative discrimination.

[0035] The application provides a high-reflection target detection and recognition method based on differential image multi-factor collaborative discrimination. The method can effectively distinguish target regions and background noise by comprehensively analyzing multiple features of the target regions, thereby significantly improving detection accuracy. Specifically, multiple factors such as the shape, size, eccentricity, average intensity and variance of the target are combined for comprehensive judgment, which can more comprehensively describe the characteristics of the target, thereby effectively reducing the false positive rate and the false negative rate. An adaptive threshold adjustment mechanism is introduced, and the detection threshold is dynamically adjusted according to the difference between the target region and the background noise, thereby maintaining stable detection performance under different light conditions and complex backgrounds. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 FIG. 1 is a schematic diagram of an image differential cat-eye effect target detection algorithm based on multi-factor joint judgment provided by an embodiment of the application.

[0037] Figure 2 FIG. 2 is an active image provided by an embodiment of the application.

[0038] Figure 3 FIG. 3 is a passive image provided by an embodiment of the application.

[0039] Figure 4 FIG. 4 is a differential image provided by an embodiment of the application.

[0040] Figure 5 FIG. 5 is a binary image provided by an embodiment of the application.

[0041] Figure 6 FIG. 6 is a marked image provided by an embodiment of the application. DETAILED DESCRIPTION

[0042] The technical solutions of the application will be specifically described below with reference to the drawings.

[0043] The application provides a high-reflection target detection and recognition method based on differential image multi-factor collaborative discrimination, which comprehensively analyzes multiple features of target candidate regions, calculates the average intensity of each target candidate region in the difference image, evaluates the activity level of the target candidate region, and realizes accurate detection of high-reflection targets in complex scenes through conditional screening.

[0044] The following is a specific implementation process of the present application.

[0045] The present application is a high-reflective target detection and recognition method based on differential image multi-factor collaborative discrimination, which uses image filtering, image registration, image difference, and multiple features (such as area, eccentricity, roundness, etc.) of target candidate regions in comprehensive analysis based on differential images, and combines average intensity and variance screening to finally determine the target region and mark it. The specific implementation steps of the method are as follows:

[0046] Step S1, using a laser detection system, acquiring continuous frames of active and passive images of a target scene;

[0047] Step S2, Gaussian filtering preprocessing of the active and passive images, image difference after registration;

[0048] Step S3, based on the differential image, adaptively selecting a threshold, performing image binarization processing to obtain a binary image;

[0049] Step S4, extracting target candidate regions and their features in the binary image, including area, eccentricity and roundness;

[0050] Step S5, based on the binary image, traversing all target candidate regions, and preliminarily screening according to the conditions of roundness, area and eccentricity, etc., to exclude regions that do not meet the conditions;

[0051] Step S6, based on the target regions preliminarily screened, calculating the radius according to the area of the region in the binary image, obtaining the coordinate range of each screened region, and extracting the image gray data in each coordinate range in the differential image;

[0052] Step S7, based on the obtained differential image data, calculating the average intensity of each region, selecting regions with higher average intensity, further calculating the gray variance of these regions, and screening the region with the largest variance;

[0053] Step S8, based on the screened region, marking the finally determined target region on the active image.

[0054] The present application will be further described in detail below in combination with the drawings and examples. The present embodiment is implemented on the premise of the technical solution of the present application, and detailed implementation modes and specific operation processes are given, but the protection scope of the present application includes but is not limited to the following examples.

[0055] Embodiment

[0056] In the method of the present application, the active and passive image acquisition system is composed of a laser and a common RGB industrial camera.

[0057] Reference Figure 1 The method of the present embodiment includes the following steps:

[0058] Step S1, using a laser detection system, acquiring continuous frame active and passive images of a target scene;

[0059] Step S2, Gaussian filtering preprocessing of the active and passive images, image difference after registration;

[0060] Step S3, based on the difference image, adaptive selection of threshold, image binarization processing to obtain a binary image;

[0061] Step S4, extracting a target candidate region and its features in the binary image, including area, eccentricity and circularity;

[0062] Step S5, based on the binary image, traversing all target candidate regions, and according to the circularity, area and eccentricity conditions for preliminary screening, excluding regions that do not meet the conditions;

[0063] Step S6, based on the target region after preliminary screening, calculating the radius in the binary image according to the area of the region, obtaining the coordinate range of each screened region, and extracting the image gray data in each coordinate range in the difference image;

[0064] Step S7, based on the obtained difference image data, calculating the average intensity of each region, selecting the region with higher average intensity, further calculating the gray variance of these regions, and screening the region with the largest variance;

[0065] Step S8, based on the screened region, marking the finally determined target region on the active image.

[0066] In this embodiment, step S2 is specifically:

[0067] Step S21, Gaussian filtering is performed on the active and passive images to smooth the details in the image and reduce the image background noise;

[0068] Step S22, using SURF algorithm to detect the active and passive images, extracting feature points and feature descriptors in the images, matching the feature descriptors of the two images, finding corresponding feature point pairs, estimating the geometric transformation relationship between the images according to the matched feature point pairs, and applying the estimated geometric transformation to register the images;

[0069] Step S23, difference of the registered active and passive images, the difference formula is:

[0070] I 差 = I 主 -I 被

[0071] Wherein I 主 represents the image obtained under laser active illumination, I 被 represents the image obtained without laser active illumination.

[0072] The step S5 is specifically:

[0073] The step S51, the features of the region extracted in the step S4 are screened to obtain a region with a circular shape close to a circle, wherein the region with a circularity greater than 0.8 is screened.

[0074] The step S52, the region screened in the step S51 is excluded by setting an upper limit and a lower limit of the area of the region, wherein the region with an area too small or too large is excluded.

[0075] The step S53, the region screened in the step S52 is further screened by setting an eccentricity range of 0.8 to 1.25, wherein a region with a regular shape is screened.

[0076] In the embodiment, the steps S4-S6 extract the features of the target candidate region in the binary image, and the target region is screened based on the multiple features of the target candidate region and in combination with the average intensity and the variance, so that the target marking is finally realized. Figure 4 It can be obviously observed that the background noise in the difference image is effectively suppressed, and the target region in the figure is more prominent, Figure 5 It can be seen that the target candidate region of the binary image has various shapes, Figure 6 It can be seen that the final target region can be screened based on the multiple features of the target candidate region, so that the target marking is realized.

[0077] The above embodiment shows that the method proposed in the application can be based on the joint judgment of multiple factors, and the method of image difference can effectively suppress the background noise and highlight the target region. The multiple features of the target candidate region in the difference image are comprehensively analyzed, and the screening is combined with the average intensity and the variance, so that the high detection accuracy can be maintained in the complex background, the dependence on the computing resources is reduced, and the method is suitable for the application scenarios with high real-time requirements.

[0078] The application further provides a high-reflection target detection and recognition system based on the difference image and the multiple-factor cooperative discrimination, which comprises a memory, a processor and computer program instructions stored in the memory and capable of being executed by the processor, and when the processor executes the computer program instructions, the steps of any one of the above-mentioned methods can be realized.

[0079] The application further provides a computer readable storage medium, which stores computer program instructions capable of being executed by a processor, and when the processor executes the computer program instructions, the steps of any one of the above-mentioned methods can be realized.

[0080] The above is the preferred embodiment of the application, and any change made according to the technical solution of the application, as long as the function effect does not exceed the range of the technical solution of the application, belongs to the protection scope of the application.

Claims

1. A high-reflection target detection and recognition method based on multi-factor collaborative discrimination of differential images, characterized by: Comprehensively analyze the various features of the target candidate area, calculate the average intensity of each target candidate area in the difference image, evaluate the activity level of the target candidate area, and through conditional screening, achieve accurate detection of high-reflectivity targets in complex scenes.

2. The method for detecting and identifying high-reflective targets based on multi-factor collaborative discrimination of differential images according to claim 1, characterized in that: Various features include area, eccentricity, and circularity.

3. The method for detecting and identifying high-reflective targets based on multi-factor collaborative discrimination of differential images according to claim 1, characterized in that: The target candidate area is obtained by filtering and preprocessing the active and passive images of consecutive frames, and then performing difference processing after aligning the two images to obtain a difference image. According to the gray value distribution of the difference image, the difference image is binarized to extract the target candidate area.

4. The method for detecting and identifying high-reflective targets based on multi-factor collaborative discrimination of differential images according to claim 3 is characterized in that: The differential image binarization is performed based on an adaptively determined threshold value.

5. The method for detecting and identifying high-reflective targets based on multi-factor collaborative discrimination of differential images according to claim 1, characterized in that: Continuous frame active and passive images are acquired based on the laser detection system.

6. The method for detecting and identifying high-reflective targets based on multi-factor collaborative discrimination of differential images according to claim 1, characterized in that: The specific implementation steps of this method are as follows: Step S1: using a laser detection system to obtain continuous active and passive images of the target scene; Step S2: pre-process the active and passive images of consecutive frames using Gaussian filtering, perform image difference after registration, and obtain a differential image; Step S3: Based on the differential image, adaptively determine the threshold value, perform image binarization processing to obtain a binary image; Step S4: extracting target candidate regions and their features in the binary image, where the features include area, eccentricity, and roundness; Step S5: Based on the binary image, traverse all target candidate regions, perform preliminary screening based on area, eccentricity and roundness, and exclude target candidate regions that do not meet the preset conditions; Step S6: Based on the initially screened target candidate regions, the radius is calculated according to the region area in the binary image, the coordinate range of each screened region is obtained, and the image grayscale data within each coordinate range is correspondingly extracted in the differential image; Step S7: Based on the differential image data obtained in step S6, calculate the average intensity of each region, select the regions with average intensity greater than a preset value, further calculate the grayscale variance of these regions, and select the region with the largest variance; Step S8: Based on the screened area, mark the final target area on the active image.

7. The method for detecting and identifying high-reflective targets based on multi-factor collaborative discrimination of differential images according to claim 6, characterized in that: The step S2 is specifically as follows: Step S21: Perform Gaussian filtering on the active and passive images to smooth the details in the images and reduce the background noise of the images; Step S22: Use the SURF algorithm to detect the active and passive images, extract feature points and feature descriptors in the images, match the feature descriptors of the two images, find corresponding feature point pairs, estimate the geometric transformation relationship between the images based on the matched feature point pairs, and apply the estimated geometric transformation to align the images; Step S23: perform differential analysis on the registered active and passive images. The differential formula is: I 差 =I 主 -I 被 Where I 主 The image obtained under laser active illumination is the active image, I 被 The image obtained without active laser illumination is called a passive image.

8. The method for detecting and identifying high-reflective targets based on multi-factor collaborative discrimination of differential images according to claim 6, characterized in that: The step S5 is specifically as follows: Step S51: for the features of the target candidate region extracted in step S4, filter the region with a roundness greater than 0.8 to obtain a region with a shape close to a circle; Step S52: For the area selected in step S51, set upper and lower limits for the area to exclude areas that are too small or too large; Step S53: For the area screened in step S52, the eccentricity range is set to 0.8 to 1.25 to screen out areas with regular shapes.

9. A high-reflection target detection and recognition system based on differential image multi-factor collaborative discrimination, characterized by: The method comprises a memory, a processor, and computer program instructions stored in the memory and capable of being executed by the processor. When the processor executes the computer program instructions, the steps of the method according to any one of claims 1 to 8 can be implemented.

10. A computer-readable storage medium having stored thereon computer program instructions that can be executed by a processor, wherein when the processor executes the computer program instructions, the steps of the method according to any one of claims 1 to 8 can be implemented.

Images