CN110415208B - Self-adaptive target detection method and device, equipment and storage medium thereof - Google Patents
Self-adaptive target detection method and device, equipment and storage medium thereof Download PDFInfo
- Publication number
- CN110415208B CN110415208B CN201910498094.2A CN201910498094A CN110415208B CN 110415208 B CN110415208 B CN 110415208B CN 201910498094 A CN201910498094 A CN 201910498094A CN 110415208 B CN110415208 B CN 110415208B
- Authority
- CN
- China
- Prior art keywords
- image
- interest
- target
- original image
- region
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10048—Infrared image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20004—Adaptive image processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20092—Interactive image processing based on input by user
- G06T2207/20101—Interactive definition of point of interest, landmark or seed
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20092—Interactive image processing based on input by user
- G06T2207/20104—Interactive definition of region of interest [ROI]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Quality & Reliability (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a self-adaptive target detection method, a device, equipment and a storage medium thereof, wherein the method comprises the steps of acquiring an original image; performing differential filtering processing on the original image to obtain a filtering processing image; clustering and merging the filtering image to obtain a plurality of image interested areas; respectively calculating the contrast ratio of each image region of interest and the original image; and obtaining a target image according to the contrast. According to the invention, the differential filter is adopted to filter the original image, so that most background areas are eliminated, then a plurality of image regions of interest (ROIS) are obtained based on a clustering and merging method, effective segmentation of the original image is realized, and finally, the contrast of the image regions of interest (ROIS) and the original image is combined to further eliminate false targets, so that the false alarm rate is reduced in target detection of simple background or complex background, and the detection effect is improved.
Description
Technical neighborhood
The invention belongs to the technical field of target detection, and particularly relates to a self-adaptive target detection method, a device, equipment and a storage medium thereof.
Background
With the development of target detection technology, detection, tracking and identification of small infrared targets mainly come from infrared search and tracking systems, and how to detect and track targets from acquired infrared images becomes important. Therefore, the detection of the infrared small target is always a hot spot subject of the infrared detection neighborhood, and the research of the detection method of the infrared small target has far-reaching significance for the reverse fight.
The existing infrared small target detection method mainly comprises two steps: the image is subjected to background suppression processing, and the image after the background suppression is subjected to segmentation processing. The method comprises the steps of performing background inhibition processing on an image to strengthen a target, wherein common background inhibition methods comprise a spatial domain filtering method, a morphological filtering method and a background prediction method, and the background inhibition methods realize the separation of high-frequency signals and low-frequency signals in the image so as to achieve the effect of highlighting the target; the image after background suppression is segmented to realize effective segmentation of the image, the target image is obtained by effectively segmenting the image, common segmentation methods comprise thresholding segmentation and regional segmentation, a typical thresholding segmentation has a gray threshold segmentation method, the segmentation effect depends on the selection of a gray threshold, the regional segmentation has a regional growth method and a division merging method, and the segmentation effect depends on the selection of seed points and the selection of similarity criteria.
In the existing infrared small target detection method, under a simple scene, the detection effect is good because the effective information of the image is more and the background interference information is less, but under a complex background, the detection effect is poor because the detection effect is influenced by the background interference information.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a self-adaptive target detection method, a device, equipment and a storage medium thereof.
The embodiment of the invention provides a self-adaptive target detection method, which comprises the following steps:
acquiring an original image;
performing differential filtering processing on the original image to obtain a filtering processing image;
clustering and merging the filtering image to obtain a plurality of image interested areas;
respectively calculating the contrast ratio of each image region of interest and the original image; and
and obtaining a target image according to the contrast.
In one embodiment of the present invention, performing differential filtering processing on the original image to obtain a filtered image, including:
and carrying out differential filtering processing on the original image by using a Gaussian differential filter to obtain the filtering processing image.
In one embodiment of the present invention, the clustering and merging process is performed on the filtered images to obtain a plurality of image interested areas, including:
acquiring a plurality of target pixel points from the filtering processing image by using a first preset threshold value;
and clustering and merging the target pixel points to obtain the image interested areas.
In one embodiment of the present invention, clustering and merging the plurality of target pixel points to obtain the plurality of image interested areas includes:
and clustering and merging the target pixel points by adopting a density clustering method according to the distance threshold, the gray threshold and the neighborhood sample number threshold to obtain the image interested areas.
In one embodiment of the present invention, calculating the contrast of each of the image regions of interest with the original image includes:
respectively acquiring a plurality of adjacent original image blocks of each image region of interest;
respectively calculating a first gray value and a second gray value of each image region of interest;
respectively calculating a third gray value of each original image block adjacent to each image region of interest;
and respectively calculating the contrast ratio between each image region of interest and the original image blocks adjacent to the image region of interest according to the first gray level value, the second gray level value and the third gray level value.
In one embodiment of the present invention, the size of the image region of interest is the same as the size of each adjacent original image block of the image region of interest, wherein the size of the image region of interest is the smallest bounding rectangle containing all target pixels within the image region of interest.
In one embodiment of the present invention, obtaining a target image according to the contrast includes:
and obtaining the target image according to the contrast and a second preset threshold value.
Another embodiment of the present invention provides an adaptive target detection apparatus, the apparatus comprising:
the data acquisition module is used for acquiring the original image;
the first data processing module is used for carrying out differential filtering processing on the original image to obtain a filtering processing image;
the second data processing module is used for carrying out clustering and merging processing on the filtering processing images to obtain a plurality of image interested areas;
the third data processing module is used for respectively calculating the contrast ratio of each image region of interest and the original image;
and the data determining module is used for obtaining the target image according to the contrast.
Still another embodiment of the present invention provides an adaptive object detection electronic device, including a processor, a communication interface, a memory, and a communication bus, where the processor, the communication interface, and the memory complete communication with each other through the communication bus;
the memory is used for storing a computer program;
the processor is configured to implement any of the methods described above when executing the computer program stored on the memory.
Yet another embodiment of the present invention provides a computer-readable storage medium having a computer program stored therein, which when executed by a processor, implements any of the methods described above.
Compared with the prior art, the invention has the beneficial effects that:
according to the invention, the differential filter is adopted to filter the original image, so that most background areas are eliminated, then a plurality of image regions of interest (ROIS) are obtained based on a clustering combination method, effective segmentation of the original image is realized, and finally, the contrast of the image regions of interest (ROIS) and the original image is combined to further eliminate false targets, so that the false alarm rate can be reduced and the detection effect can be improved in target detection of simple background or complex background.
The present invention will be described in further detail with reference to the accompanying drawings and examples.
Drawings
Fig. 1 is a schematic flow chart of an adaptive target detection method according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of an eight-neighborhood block in an adaptive target detection method according to an embodiment of the present invention;
fig. 3 is a schematic structural diagram of an adaptive target detection apparatus according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of an adaptive target detection electronic device according to an embodiment of the present invention;
fig. 5 is a schematic structural diagram of a computer-readable storage medium according to an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to specific examples, but embodiments of the present invention are not limited thereto.
Example 1
The existing infrared small target detection method mainly comprises two steps: background suppression and image segmentation. The conventional background suppression method comprises a conventional high-pass filtering or low-pass filtering method, a gray morphology-based method, a frequency domain filtering-based method and a background prediction-based method, wherein the conventional background suppression method is to realize background estimation after performing linear transformation or nonlinear transformation on pixels near a given pixel point, and can be converted into convolution operation based on a filtering template in specific implementation; the conventional image segmentation is to perform effective image segmentation on an image after background suppression so as to obtain a target image, the common image segmentation method comprises a segmentation method based on a threshold value and a segmentation method based on a region, wherein the segmentation method based on the threshold value is to select a proper criterion function to solve an optimal gray threshold value, the common threshold value segmentation method comprises a fixed threshold value method, a fuzzy threshold value method and a maximum inter-class variance method, the region-based segmentation method is to divide the image into different regions according to similarity criteria, and the common region segmentation method comprises a seed growth method, a region splitting and merging method and a watershed method.
The traditional infrared small target detection method has good detection effect in a simple scene and high instantaneity, and in a complex background, the background suppression technology extracts the background information while extracting the target image, the effect of background suppression is poor, and more false alarm points are generated after the image segmentation processing. In addition, the conventional infrared small target detection method can only detect targets under fixed scales, the effectiveness of the target detection method is lost for multi-scale targets, and a scale space is generally required to be established for multi-scale target detection, but the target detection time is increased in multiple times, so that the real-time detection of the targets is not facilitated.
Based on the above-mentioned problems, please refer to fig. 1, fig. 1 is a flow chart of an adaptive target detection method according to an embodiment of the present invention. The embodiment of the invention provides a self-adaptive target detection method, which comprises the following steps:
and step 1, acquiring an original image.
In this embodiment, the original image includes infrared small target image data, background area data, and noise data.
And step 2, performing differential filtering processing on the original image to obtain a filtering processing image.
Specifically, the conventional background suppression method is essentially background prediction, and the gray value of the central pixel point is replaced by the conversion value of the pixels around the given pixel point, so that more or less deviation exists, and even if weak noise and background information exist, more false alarm points exist in the detection result, so that larger interference is brought to target detection. Therefore, the present embodiment performs differential filtering processing on the original image by using the differential filter, so that most of the background area or noise information can be eliminated. The differential filter convolves with the original image, and the central area of the filtering template captures the target pixel point of the infrared small target in the original image, which is more beneficial to extracting the target pixel point at the edge of the original image, so that the target detection is performed under more effective target pixel points, and the target detection effect is improved.
Preferably, the differential filter is a gaussian differential filter, a differential low-pass filter, a differential band-pass filter.
And step 3, clustering and merging the filtered images to obtain a plurality of image interested areas.
Specifically, most of the traditional image segmentation methods are based on pixel-level operation, the image is processed pixel by pixel, and due to the limitation of the size of a filtering template, target detection can only be performed by a target detection frame with a fixed scale, multi-scale target detection can not be effectively performed in real time, false targets and false alarm problems can exist in detection results, the traditional multi-scale detection methods generally need to perform multiple filtering or establish a scale pyramid, the multi-scale detection methods are serious in time consumption, poor in image segmentation robustness, and a large number of false alarm points exist in detection results. Therefore, the embodiment provides an image segmentation method based on a clustering method, which adopts a clustering combination method to perform clustering combination processing on a plurality of target pixel points in a filtering processing image obtained by filtering through a Gaussian differential filter, and the clustering combination result corresponds to a plurality of image interested areas (Region of Interest, for short, ROIS) of the embodiment, and the image interested areas ROIS reserve most effective target pixel points in an original image, so that background areas and noise information are further eliminated, false targets and false alarms are reduced, and the effect of target detection is improved. Meanwhile, in this embodiment, the shape and size of each image region of interest ROIS may be different, and the size of the corresponding target detection frame may be continuously adjusted according to the size of each image region of interest ROIS, so as to implement scale adaptive detection of the target. The embodiment is a block-level-based target scale self-adaptive detection method, and compared with a pixel-level-based detection method, the target detection time is obviously reduced.
Preferably, the cluster merging method comprises a mixed Gaussian cluster method, a density cluster method, a cluster method based on a neural network model and a hierarchical aggregation cluster method.
And 4, respectively calculating the contrast ratio of the interested region of each image and the original image.
Specifically, in this embodiment, the discontinuity between the target pixel point and the surrounding background pixel point or the noise pixel point is considered, so a contrast function including the image region of interest and the original image information is designed, the contrast between each image region of interest ROIS and the surrounding original image is calculated through the contrast function, the smaller the value of the contrast is, the lower the contrast of the corresponding image region of interest ROIS indicates that the image region of interest ROIS has a certain continuity with the surrounding region, i.e. the surrounding has a similar region, and the image region of interest ROIS is likely to be the background region or the noise region.
And 5, obtaining a target image according to the contrast.
Specifically, according to the contrast function in step 4, the embodiment obtains the contrast between each image region of interest ROIS and the original image, retains the image region of interest ROIS with large contrast, and removes the image region of interest ROIS with small contrast, thereby obtaining the target image. In the embodiment, the discontinuity between the target pixel point and the surrounding background pixel points or the noise pixel points is utilized to eliminate false edge areas, reduce false alarm rate and improve detection effect. The contrast is related to the empirical value in the actual scene, and the specific value of the empirical value is finely adjusted according to the actual scene.
In summary, in this embodiment, the differential filter is used to filter the original image, so that most of background area or noise information is removed, then a plurality of image interested areas ROIS are obtained based on the clustering and merging method, so as to effectively divide the original image, and finally, the contrast ratio of the image interested areas ROIS and the original image is combined to further remove false targets, so that the false alarm rate is reduced and the detection effect is improved in simple background or complex background target detection.
Example two
Based on the first embodiment, the difference filtering process of the step 2 in the first embodiment is a gaussian difference filter, the clustering combining process of the step 3 in the first embodiment is an improved density clustering method, then a contrast function is designed, and the contrast of the region of interest ROIS of the image and the original image is combined to realize the self-adaptive detection of the target, and the specific implementation process includes:
and step 1, acquiring an original image.
And step 2, performing differential filtering processing on the original image by adopting a Gaussian differential filter to obtain a filtering processing image.
Specifically, in this embodiment, considering that the gray level of the infrared small target area is generally larger than the gray level of the surrounding background or noise, the gray level distribution of the infrared small target area in the whole original image has an obvious peak value, and the infrared small target area has a larger contrast with the surrounding areas in all directions, therefore, this embodiment proposes to effectively extract the target pixel point of the infrared small target in the original image by using the gaussian differential filter, and meanwhile, excludes some background areas or noise information, thereby effectively enhancing the target pixel point of the infrared small target.
In this embodiment, by deducting the gaussian function, the difference between two different gaussian functions is obtained, and a gaussian differential filter with positive center and negative periphery is obtained as the differential filter processor of this embodiment, where the gaussian differential filter is specifically designed as follows:
wherein x and y are target pixel points in the original image, and sigma 1 and sigma 2 are Gaussian functions G respectively 1 (x, y) and Gaussian function G 2 Scale parameters of (x, y) and such that σ1<σ2。
Preferably, σ1 is a 3*3 matrix and σ2 is a 5*5 matrix.
According to the embodiment, the Gaussian differential filter is used as a filtering template, convolution processing is carried out on the filtering template and an original image, an infrared small target in the original image is captured by the central area of the filtering template, the filtering processing image obtained by filtering is an area contrast map in a difference mode, and the pixel points with larger area contrast are more likely to be target pixel points, so that suppression of most of background or noise is achieved, and the target pixel points are enhanced. In this embodiment, because the three-dimensional characteristic of the gaussian differential filter better conforms to the three-dimensional characteristic of the infrared small target image, most background pixel points and noise pixel points can be eliminated by performing differential filtering processing on the original image through the gaussian differential filter, namely, false alarm points are reduced, so that the target pixel points of the infrared small target are captured more effectively.
And step 3, adopting an improved density clustering method to perform clustering merging processing on the filtering processing images to obtain a plurality of image interested areas.
In the embodiment, step 3 of clustering and merging the filtered images by adopting an improved density clustering method to obtain a plurality of image interested areas, the specific implementation can comprise the following steps 3.1 and 3.2, wherein,
and 3.1, acquiring a plurality of target pixel points from the filtering processing image by using a first preset threshold value.
Specifically, before the filter processing images are clustered and combined, a plurality of target pixel points are selected from the filter processing images through a fixed threshold, such as a first preset threshold, and then the selected target pixel points are subjected to subsequent processing, so that the subsequent calculation amount can be reduced, the image processing speed is improved, and meanwhile, a plurality of target pixel points are selected through judging the pixel points of the filter processing images and the first preset threshold, so that a part of background pixel points and noise pixel points can be rapidly eliminated, and the false alarm rate is reduced. The target pixels are suspected target pixels, which may be target pixels, background pixels or noise pixels; the specific value of the first preset threshold needs to be fine-tuned in different scenes.
Preferably, the value of the first preset threshold is 100.
And 3.2, carrying out clustering and merging processing on the plurality of target pixel points to obtain a plurality of image interested areas.
Specifically, in order to improve the effect of target detection, the embodiment improves the traditional density clustering method, and adopts the improved density clustering method to perform clustering merging processing on a plurality of target pixel points according to a distance threshold, a gray threshold and a neighborhood sample number threshold, so as to obtain a plurality of image interesting areas ROIS. Specifically, in this embodiment, the distance threshold and the gray threshold are simultaneously used as constraint conditions of the density clustering method, and specific inputs for the improved density clustering method are: the target pixel point set D= { D filtered by the first preset threshold value 1 ，d 2 ，…，d m -and parameters related to an improved density clustering method, the parameters comprising a distance threshold α, a gray threshold β, a neighborhood sample threshold MinP, wherein d 1 ，d 2 ，…，d m All are target pixel points, the Euclidean distance and gray level are adopted in the distance measurement mode of the distance threshold value alphaThe gray level measurement mode of the threshold value beta adopts gray level difference, and the embodiment adopts Euclidean distance and gray level difference to carry out comprehensive measurement; the output is: image region of interest ROIS set b= { B 1 ，B 2 ，…，B k The number of the image regions of interest ROIS output by the embodiment is k, k is uncertain data, specifically determined by target pixel points input by an improved density clustering method and parameter design of the method, and a plurality of detected image regions of interest ROIS are correspondingly obtained through a clustering result of the improved density clustering method.
Preferably, the distance threshold α has a value of 5, the gray threshold β has a value of 30, and the neighborhood sample number threshold MinP has a value of 3.
Compared with the density clustering method before improvement, the improved density clustering method provided by the embodiment combines the gray threshold value to perform self-adaptive clustering, can better adapt to target pixel points, has more reasonable and accurate merging results, and ensures the clustering speed; compared with the traditional seed growth method, the improved density clustering method provided by the embodiment does not need to determine the selection of seed pixel points and the number of seeds, and the clustering combination is performed in a self-adaptive manner completely according to the distance threshold alpha, the gray threshold beta and the neighborhood sample number threshold MinP, so that the image region of interest (ROIS) obtained by the improved density clustering method is in any shape and is not limited by the shape of a target image, the number of target pixel points does not need to be known in advance, the accurate scale of the target is obtained while the self-adaptive target detection frame is effectively segmented, the scale self-adaptation of the target is realized, the real-time performance of image detection is ensured, and a part of background pixel points and noise pixel points can be eliminated. For each adaptive target detection frame, the embodiment preferably takes a minimum bounding rectangle including all target pixels in the region of interest ROIS of the image.
It should be noted that the adaptive target detection frame of the present embodiment is not limited to the above-mentioned access method.
And 4, respectively calculating the contrast ratio of the interested region of each image and the original image.
The specific implementation of calculating the contrast ratio between the region of interest of each image and the original image in step 4 of this embodiment includes the following steps 4.1, 4.2, 4.3 and 4.4, wherein,
and 4.1, respectively acquiring a plurality of adjacent original image blocks of the interested region of each image.
Specifically, in this embodiment, firstly, an image region of interest ROIS is located in an original image, where the image region of interest ROIS is the image region of interest ROIS obtained in the step 3, a minimum circumscribed rectangle is obtained according to all target pixels in the image region of interest ROIS as a target image block, the target image block corresponds to a specific position and scale of the image region of interest ROIS in the original image, and then a plurality of original image blocks adjacent to the target image block are obtained from the periphery of the target image block, and the size of each original image block is equal to the size of the target image block. For example, referring to fig. 2, fig. 2 is a schematic structural diagram of an eight-neighborhood block in an adaptive target detection method according to an embodiment of the present invention, as shown in fig. 2, a target image block corresponding to a region of interest ROIS of the image of the embodiment is shown in a position 0 in fig. 2, and 8 original image blocks adjacent to the target image block are taken, which are shown in positions 1,2, … and 8 in fig. 2, respectively.
And 4.2, respectively calculating a first gray value and a second gray value of the region of interest of each image.
Specifically, a gray level average value of a target image block corresponding to each image region of interest (ROIS) is calculated by a gray level calculation method, wherein the gray level average value is a first gray level value, and the first gray level value is recorded as m 0 Simultaneously calculating the maximum gray value in the target image block corresponding to each image region of interest (ROIS), wherein the maximum gray value is a second gray value, and the second gray value is recorded as L max 。
And 4.3, respectively calculating a third gray value of each original image block adjacent to the region of interest of each image.
Specifically, the gray average value of 8 original image blocks corresponding to each target image block is calculated by a gray calculation method, wherein the gray average value is a third gray valueRespectively marked as m 1 ，…，m 8 . Wherein the target image block corresponds to the image region of interest ROIS.
And 4.4, respectively calculating the contrast ratio between each image region of interest and a plurality of original image blocks adjacent to the image region of interest according to the first gray level value, the second gray level value and the third gray level value.
Specifically, based on the first gray value calculated in step 4.2, the second gray value, and the third gray value calculated in step 4.3, the contrast function C of the target image block corresponding to the region of interest ROIS of the image and 8 original image blocks adjacent to the target image block is constructed in this embodiment, and the specific contrast function C is designed as follows:
in this embodiment, the contrast ratio between the region of interest ROIS of each image and the original image can be calculated by the formula (2), and the smaller the contrast ratio value of the contrast function C, the lower the contrast ratio corresponding to the region of interest ROIS of the image, which indicates that the region of interest ROIS of the image has a certain continuity with the surrounding region, i.e. the surrounding region has a similar region, the region of interest ROIS of the image is likely to be a background region or a noise region.
And 5, obtaining a target image according to the contrast.
In step 5 of this embodiment, it is specifically determined whether the region of interest ROIS of the image is the target image region according to the contrast obtained by the contrast function C in step 4 and the second preset threshold.
Specifically, in this embodiment, a second preset threshold is empirically set, the contrast between the region of interest ROIS of the image and the original image is obtained through the above-mentioned contrast function C formula (2), the contrast is compared with the second preset threshold, and the contrast is less than or equal to the second preset threshold, which indicates that the region of interest ROIS of the image is a target image region, and the contrast is greater than the second preset threshold, which indicates that the region of interest ROIS of the image is a background region or a noise region, so as to exclude false targets such as the background region or the noise region, further exclude false edge regions, reduce false alarm rate, improve detection effect, and obtain a final real target image region. The second preset threshold specific value needs to be finely adjusted according to an actual scene.
Preferably, the value of the second preset threshold is 1.1.
And (3) carrying out contrast calculation on each image region of interest (ROSS) through a contrast function C formula (2), comparing the calculated contrast with a second preset threshold value, confirming whether each image region of interest (ROSS) is a target image region, excluding non-target image regions, reserving target image regions, and obtaining a final target image from the target image regions.
In summary, the adaptive target detection method provided in this embodiment improves the detection effect of the conventional infrared small target detection method, reduces the false alarm rate, and simultaneously can accurately implement the adaptation of the target detection scale in real time, thereby improving the robustness of target detection and enabling the adaptive target detection method to be better applied to actual scenes.
Example III
On the basis of the second embodiment, please refer to fig. 3, fig. 3 is a schematic structural diagram of an adaptive target detection apparatus according to an embodiment of the present invention. The present embodiment provides an adaptive target detection apparatus, including:
and the data acquisition module is used for acquiring the original image.
The first data processing module is used for carrying out differential filtering processing on the original image to obtain a filtering processing image.
Specifically, in this embodiment, the original image is subjected to differential filtering processing by using a gaussian differential filter, and a filtered image is obtained.
The second data processing module is used for carrying out clustering combination processing on the filtering processing images to obtain a plurality of image interested areas;
specifically, in this embodiment, a plurality of target pixel points are first obtained from a filtering image by using a first preset threshold, and then, the plurality of target pixel points are clustered and combined by using an improved density clustering method according to a distance threshold, a gray threshold and a neighborhood sample number threshold, so as to obtain a plurality of image interested areas.
The third data processing module is used for respectively calculating the contrast ratio of the interested area of each image and the original image;
specifically, in this embodiment, first, a plurality of original image blocks adjacent to each image region of interest are obtained respectively, then, a first gray value and a second gray value of each image region of interest are calculated respectively, and a third gray value of each original image block adjacent to each image region of interest is calculated respectively, and according to the first gray value, the second gray value and the third gray value, the contrast between each image region of interest and the plurality of original image blocks adjacent to the image region of interest is calculated respectively.
And the data determining module is used for obtaining the target image according to the contrast.
Specifically, when the target image is obtained according to the contrast, the second preset threshold is combined at the same time, and the target image is obtained according to the contrast and the second preset threshold.
The adaptive target detection device provided in this embodiment may perform the above method embodiment, and its implementation principle and technical effects are similar, and will not be described herein.
Example IV
On the basis of the third embodiment, please refer to fig. 4, fig. 4 is a schematic structural diagram of an adaptive target detection electronic device according to an embodiment of the present invention. The embodiment provides self-adaptive target detection electronic equipment, which comprises a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory are communicated with each other through the communication bus;
a memory for storing a computer program;
a processor for executing a computer program stored on a memory, the computer program when executed by the processor performing the steps of:
and step 1, acquiring an original image.
And step 2, performing differential filtering processing on the original image to obtain a filtering processing image.
Specifically, in this embodiment, the original image is subjected to differential filtering processing by using a gaussian differential filter, and a filtered image is obtained.
And step 3, clustering and merging the filtered images to obtain a plurality of image interested areas.
Specifically, in this embodiment, a plurality of target pixel points are first obtained from a filtering image by using a first preset threshold, and then, the plurality of target pixel points are clustered and combined by using an improved density clustering method according to a distance threshold, a gray threshold and a neighborhood sample number threshold, so as to obtain a plurality of image interested areas.
And 4, respectively calculating the contrast ratio of the interested region of each image and the original image.
Specifically, in this embodiment, first, a plurality of original image blocks adjacent to each image region of interest are obtained respectively, then, a first gray value and a second gray value of each image region of interest are calculated respectively, and a third gray value of each original image block adjacent to each image region of interest is calculated respectively, and according to the first gray value, the second gray value and the third gray value, the contrast between each image region of interest and the plurality of original image blocks adjacent to the image region of interest is calculated respectively.
And 5, obtaining a target image according to the contrast.
Specifically, when the target image is obtained according to the contrast, the second preset threshold is combined at the same time, and the target image is obtained according to the contrast and the second preset threshold.
The adaptive target detection electronic device provided in this embodiment may perform the above method embodiment and the above apparatus embodiment, and its implementation principle and technical effects are similar, and are not described herein again.
Example five
On the basis of the fourth embodiment, please refer to fig. 5, fig. 5 is a schematic structural diagram of a computer readable storage medium according to an embodiment of the present invention. The present embodiment provides a computer readable storage medium having a computer program stored thereon, which when executed by a processor, implements the steps of:
and step 1, acquiring an original image.
And step 2, performing differential filtering processing on the original image to obtain a filtering processing image.
Specifically, in this embodiment, the original image is subjected to differential filtering processing by using a gaussian differential filter, and a filtered image is obtained.
And step 3, clustering and merging the filtered images to obtain a plurality of image interested areas.
Specifically, in this embodiment, a plurality of target pixel points are first obtained from a filtering image by using a first preset threshold, and then, the plurality of target pixel points are clustered and combined by using an improved density clustering method according to a distance threshold, a gray threshold and a neighborhood sample number threshold, so as to obtain a plurality of image interested areas.
And 4, respectively calculating the contrast ratio of the interested region of each image and the original image.
Specifically, in this embodiment, first, a plurality of original image blocks adjacent to each image region of interest are obtained respectively, then, a first gray value and a second gray value of each image region of interest are calculated respectively, and a third gray value of each original image block adjacent to each image region of interest is calculated respectively, and according to the first gray value, the second gray value and the third gray value, the contrast between each image region of interest and the plurality of original image blocks adjacent to the image region of interest is calculated respectively.
And 5, obtaining a target image according to the contrast.
Specifically, when the target image is obtained according to the contrast, the second preset threshold is combined at the same time, and the target image is obtained according to the contrast and the second preset threshold.
The computer readable storage medium provided in this embodiment may perform the above method embodiment, the above apparatus embodiment, and the above electronic device embodiment, and achieve similar principles and technical effects, which are not described herein again.
The foregoing is a further detailed description of the invention in connection with the preferred embodiments, and it is not intended that the invention be limited to the specific embodiments described. It should be understood by those skilled in the art that several simple deductions or substitutions may be made without departing from the spirit of the present invention, and the present invention is not limited to the above-described embodiments.
Claims (5)
1. An adaptive target detection method, comprising:
acquiring an original image;
performing differential filtering processing on the original image to obtain a filtering processing image;
acquiring a plurality of target pixel points from the filtering processing image by using a first preset threshold value to exclude a part of background pixel points and noise pixel points, and carrying out self-adaptive clustering and merging processing on the plurality of target pixel points by using a density clustering method according to a distance threshold value, a gray level threshold value and a neighborhood sample number threshold value to obtain a plurality of image interested areas, wherein the plurality of target pixel points are suspected target pixel points;
calculating the contrast ratio of each image region of interest to the original image, respectively, including:
a plurality of original image blocks adjacent to each image region of interest are respectively obtained, and the method comprises the following steps: firstly positioning the original image to the image interested region, obtaining a minimum circumscribed rectangle as a target image block according to all target pixel points in the image interested region, wherein the target image block corresponds to the specific position and scale of the image interested region in the original image, then acquiring a plurality of adjacent original image blocks from the periphery of the target image block, and the size of each original image block is equal to that of the target image block;
respectively calculating a first gray value and a second gray value of each image region of interest;
respectively calculating a third gray value of each original image block adjacent to each image region of interest;
respectively calculating the contrast ratio between each image region of interest and the original image blocks adjacent to the image region of interest according to the first gray value, the second gray value and the third gray value;
the size of the image region of interest is the same as the size of each adjacent original image block of the image region of interest, wherein the size of the image region of interest is a minimum circumscribed rectangle containing all target pixel points in the image region of interest;
obtaining a target image according to the contrast, including: and obtaining the target image according to the contrast and a second preset threshold value.
2. The method of claim 1, wherein performing differential filtering on the original image to obtain a filtered image comprises:
and carrying out differential filtering processing on the original image by using a Gaussian differential filter to obtain the filtering processing image.
3. An adaptive target detection apparatus, the apparatus comprising:
the data acquisition module is used for acquiring an original image;
the first data processing module is used for carrying out differential filtering processing on the original image to obtain a filtering processing image;
the second data processing module is used for acquiring a plurality of target pixel points from the filtering processing image by utilizing a first preset threshold value to exclude a part of background pixel points and noise pixel points, and carrying out clustering and merging processing on the plurality of target pixel points in a self-adaptive manner by adopting a density clustering method according to a distance threshold value, a gray level threshold value and a neighborhood sample number threshold value to obtain a plurality of image interested areas, wherein the plurality of target pixel points are suspected target pixel points;
a third data processing module, configured to calculate a contrast ratio between each image region of interest and the original image, where the third data processing module includes:
a plurality of original image blocks adjacent to each image region of interest are respectively obtained, and the method comprises the following steps: firstly positioning the original image to the image interested region, obtaining a minimum circumscribed rectangle as a target image block according to all target pixel points in the image interested region, wherein the target image block corresponds to the specific position and scale of the image interested region in the original image, then acquiring a plurality of adjacent original image blocks from the periphery of the target image block, and the size of each original image block is equal to that of the target image block;
respectively calculating a first gray value and a second gray value of each image region of interest;
respectively calculating a third gray value of each original image block adjacent to each image region of interest;
respectively calculating the contrast ratio between each image region of interest and the original image blocks adjacent to the image region of interest according to the first gray value, the second gray value and the third gray value;
the size of the image region of interest is the same as the size of each adjacent original image block of the image region of interest, wherein the size of the image region of interest is a minimum circumscribed rectangle containing all target pixel points in the image region of interest;
the data determining module is used for obtaining a target image according to the contrast, and comprises the following steps: and obtaining the target image according to the contrast and a second preset threshold value.
4. An electronic device for adaptive target detection, wherein the electronic device comprises a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory complete communication with each other through the communication bus;
the memory is used for storing a computer program;
the processor, when executing the computer program stored on the memory, is configured to implement the method of any one of claims 1-2.
5. A computer readable storage medium, characterized in that the computer readable storage medium has stored therein a computer program which, when executed by a processor, implements the method of any of claims 1-2.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910498094.2A CN110415208B (en) | 2019-06-10 | 2019-06-10 | Self-adaptive target detection method and device, equipment and storage medium thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910498094.2A CN110415208B (en) | 2019-06-10 | 2019-06-10 | Self-adaptive target detection method and device, equipment and storage medium thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110415208A CN110415208A (en) | 2019-11-05 |
| CN110415208B true CN110415208B (en) | 2023-10-17 |
Family
ID=68358917
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910498094.2A Active CN110415208B (en) | 2019-06-10 | 2019-06-10 | Self-adaptive target detection method and device, equipment and storage medium thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110415208B (en) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110838123B (en) * | 2019-11-06 | 2022-02-11 | 南京止善智能科技研究院有限公司 | Segmentation method for illumination highlight area of indoor design effect image |
| CN110910421B (en) * | 2019-11-11 | 2022-11-11 | 西北工业大学 | Weak and small moving object detection method based on block characterization and variable neighborhood clustering |
| CN111047624A (en) * | 2019-12-27 | 2020-04-21 | 成都英飞睿技术有限公司 | Image dim target detection method, device, equipment and storage medium |
| WO2021146952A1 (en) * | 2020-01-21 | 2021-07-29 | 深圳市大疆创新科技有限公司 | Following method and device, movable platform, and storage medium |
| CN111476971A (en) * | 2020-04-09 | 2020-07-31 | 英大智能电气有限公司 | Transmission line closely mountain fire monitoring devices |
| CN111985555B (en) * | 2020-08-19 | 2021-04-13 | 中国科学院上海微系统与信息技术研究所 | Millimeter wave three-dimensional holographic image denoising method |
| CN112837335B (en) * | 2021-01-27 | 2023-05-09 | 上海航天控制技术研究所 | Medium-long wave infrared composite anti-interference method |
| CN113810720A (en) * | 2021-08-09 | 2021-12-17 | 北京博雅慧视智能技术研究院有限公司 | Image processing method, device, equipment and medium |
| CN117557784B (en) * | 2024-01-09 | 2024-04-26 | 腾讯科技（深圳）有限公司 | Target detection method, target detection device, electronic equipment and storage medium |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103198480A (en) * | 2013-04-02 | 2013-07-10 | 西安电子科技大学 | Remote sensing image change detection method based on area and Kmeans clustering |
| CN103996198A (en) * | 2014-06-04 | 2014-08-20 | 天津工业大学 | Method for detecting region of interest in complicated natural environment |
| CN104123561A (en) * | 2014-07-10 | 2014-10-29 | 中国矿业大学 | Spatial gravity model based fuzzy c-means remote sensing image automatic classification method |
| CN105261004A (en) * | 2015-09-10 | 2016-01-20 | 西安电子科技大学 | Mean shift and neighborhood information based fuzzy C-mean image segmentation method |
| CN105374029A (en) * | 2015-10-12 | 2016-03-02 | 国家电网公司 | Segmenting method and system of transformer substation equipment infrared image interest areas |
| CN105513080A (en) * | 2015-12-21 | 2016-04-20 | 南京邮电大学 | Infrared image target salience evaluating method |
| CN105654453A (en) * | 2014-11-10 | 2016-06-08 | 华东师范大学 | Robust FCM image segmentation method |
| CN107256560A (en) * | 2017-05-16 | 2017-10-17 | 北京环境特性研究所 | A kind of method for detecting infrared puniness target and its system |
| CN107392885A (en) * | 2017-06-08 | 2017-11-24 | 江苏科技大学 | A kind of method for detecting infrared puniness target of view-based access control model contrast mechanism |
| CN107424146A (en) * | 2017-06-28 | 2017-12-01 | 北京理工大学 | A kind of infrared polarization method for objectively evaluating image quality and system |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP6388507B2 (en) * | 2014-07-30 | 2018-09-12 | オリンパス株式会社 | Image processing device |
-
2019
- 2019-06-10 CN CN201910498094.2A patent/CN110415208B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103198480A (en) * | 2013-04-02 | 2013-07-10 | 西安电子科技大学 | Remote sensing image change detection method based on area and Kmeans clustering |
| CN103996198A (en) * | 2014-06-04 | 2014-08-20 | 天津工业大学 | Method for detecting region of interest in complicated natural environment |
| CN104123561A (en) * | 2014-07-10 | 2014-10-29 | 中国矿业大学 | Spatial gravity model based fuzzy c-means remote sensing image automatic classification method |
| CN105654453A (en) * | 2014-11-10 | 2016-06-08 | 华东师范大学 | Robust FCM image segmentation method |
| CN105261004A (en) * | 2015-09-10 | 2016-01-20 | 西安电子科技大学 | Mean shift and neighborhood information based fuzzy C-mean image segmentation method |
| CN105374029A (en) * | 2015-10-12 | 2016-03-02 | 国家电网公司 | Segmenting method and system of transformer substation equipment infrared image interest areas |
| CN105513080A (en) * | 2015-12-21 | 2016-04-20 | 南京邮电大学 | Infrared image target salience evaluating method |
| CN107256560A (en) * | 2017-05-16 | 2017-10-17 | 北京环境特性研究所 | A kind of method for detecting infrared puniness target and its system |
| CN107392885A (en) * | 2017-06-08 | 2017-11-24 | 江苏科技大学 | A kind of method for detecting infrared puniness target of view-based access control model contrast mechanism |
| CN107424146A (en) * | 2017-06-28 | 2017-12-01 | 北京理工大学 | A kind of infrared polarization method for objectively evaluating image quality and system |
Non-Patent Citations (1)
| Title |
|---|
| A Robust Infrared Small Target Detection Algorithm Based on Human Visual System;J.H. Han等;《IEEE GEOSCIENCE AND REMOTE SENSING LETTERS》;20141231;第2168-2172页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110415208A (en) | 2019-11-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110415208B (en) | Self-adaptive target detection method and device, equipment and storage medium thereof | |
| CN108010021B (en) | Medical image processing system and method | |
| CN107358258B (en) | SAR image target classification based on NSCT double CNN channels and selective attention mechanism | |
| CN111738342B (en) | Pantograph foreign matter detection method, storage medium and computer equipment | |
| CN110517288A (en) | Real-time target detecting and tracking method based on panorama multichannel 4k video image | |
| CN109086724B (en) | Accelerated human face detection method and storage medium | |
| CN111145209A (en) | Medical image segmentation method, device, equipment and storage medium | |
| Er-Sen et al. | An adaptive edge-detection method based on the canny operator | |
| CN110660065B (en) | Infrared fault detection and identification algorithm | |
| CN109685045A (en) | A kind of Moving Targets Based on Video Streams tracking and system | |
| CN113780110A (en) | Method and device for detecting weak and small targets in image sequence in real time | |
| CN111476804A (en) | Method, device and equipment for efficiently segmenting carrier roller image and storage medium | |
| CN111369570A (en) | Multi-target detection tracking method for video image | |
| CN112288780B (en) | Multi-feature dynamically weighted target tracking algorithm | |
| CN114581658A (en) | Target detection method and device based on computer vision | |
| CN113723314A (en) | Sugarcane stem node identification method based on YOLOv3 algorithm | |
| Lin et al. | Image segmentation based on edge detection and region growing for thinprep-cervical smear | |
| Zhu et al. | Corn leaf diseases diagnostic techniques based on image recognition | |
| CN110288616B (en) | Method for segmenting hard exudation in fundus image based on fractal and RPCA | |
| CN112435272A (en) | High-voltage transmission line connected domain removing method based on image contour analysis | |
| CN110111355B (en) | Moving vehicle tracking method capable of resisting strong shadow interference | |
| CN112084957A (en) | Mobile target retention detection method and system | |
| CN116524269A (en) | Visual recognition detection system | |
| CN115147613A (en) | Infrared small target detection method based on multidirectional fusion | |
| Liu et al. | A Generative Adversarial Network for infrared and visible image fusion using adaptive dense generator and Markovian discriminator |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |