A target detection method and device based on information fusion

Abstract

The present application discloses a target detection method and device based on information fusion. The target detection method provided by the present application first performs four-channel threshold segmentation, then fuses the segmentation results of the four channels using the D-S evidence theory and binarizes the fused image, then compensates for the information loss in the fusion process by the region expansion method, combines the line segment connectivity judgment condition, determines the connectivity relationship between each line segment in the binary image, obtains a number of connected domains, and finally calculates the average probability distribution value of the connected domain to determine the target object in the target image based on the average probability distribution value of each connected domain. The detection process does not need to rely on the training of a large amount of data, nor does it need to wait for a long time for the algorithm training to be completed, thereby solving the technical problem that the existing target detection method is time-consuming.

Description

Technical Field

[0001] The present application relates to the field of image recognition technology, and in particular to a target detection method and device based on information fusion. Background Art

[0002] Object detection technology aims to identify and locate objects in images, and to determine their size and category. It is widely used in many scenarios, such as autonomous driving, augmented reality, and video surveillance.

[0003] The current mainstream approach to object detection is still based on deep learning. These methods are mainly divided into two-stage and single-stage approaches. Two-stage methods, such as the R-CNN series, first generate an area where an object may be located and then perform fine-grained object detection. Single-stage methods, such as YOLO and SSD, simultaneously calculate the object location and type in the image. However, these methods rely on large training sets and require a long training time. Summary of the Invention

[0004] The embodiments of the present application provide a target detection method and device based on information fusion, which are used to solve the technical problems that existing target detection methods rely on a large number of training sets, require a lot of time for training, and are time-consuming.

[0005] To solve the above technical problems, the first aspect of the present application provides a target detection method based on information fusion, comprising:

[0006] Obtain the target image to be processed;

[0007] Traversing the pixel points in the target image, determining the parameter values of the four channels of each pixel point, wherein the four channels are specifically: R channel, G channel, B channel and V channel;

[0008] Calculate the image segmentation thresholds of the four channels respectively by a preset image threshold segmentation method, and perform adaptive segmentation according to each of the image segmentation thresholds to obtain a segmented image corresponding to each channel;

[0009] By using DS evidence theory, each of the segmented images is fused to obtain a fused image;

[0010] Binarizing the fused image to obtain a binarized image;

[0011] Compensating the binary image by region expansion processing, and determining connectivity relationships between line segments in the binary image based on the compensated binary image and a line segment connectivity determination condition, connecting the line segments based on the connectivity relationships between the line segments, and merging connected domains to which the connected line segments belong;

[0012] An average probability distribution value of the connected component is calculated to identify the target object in the target image based on the average probability distribution value.

[0013] Preferably, the image segmentation thresholds of the four channels are calculated respectively by a preset image threshold segmentation method, and adaptive segmentation is performed according to each of the image segmentation thresholds to obtain the segmented image corresponding to each channel, specifically comprising:

[0014] The image segmentation thresholds of the four channels are calculated respectively by the maximum inter-class variance method, and adaptive segmentation is performed according to each of the image segmentation thresholds to obtain the segmented images corresponding to each channel.

[0015] Preferably, fusing the segmented images by DS evidence theory to obtain a fused image specifically includes:

[0016] Acquiring target position information of the target image, and performing ellipse fitting processing on the detected target object image based on the target position information to obtain a circumscribed ellipse area of the target object image, wherein the detected target position information is position information of the target object in the target image obtained by pre-detecting the target image;

[0017] Calculating a probability distribution value for each pixel based on a relative positional relationship between the pixel coordinates in each segmented image and the circumscribed elliptical area;

[0018] Based on the probability distribution value of each pixel point, the segmented images are fused in combination with the Dempster rule to obtain a fused image.

[0019] Preferably, calculating the average probability distribution value of the connected component to identify the target object in the target image based on the average probability distribution value specifically includes:

[0020] Based on the pixels in the connected domain and the probability distribution values of the pixels in the connected domain, an average probability distribution value of the connected domain is calculated, so as to identify the target object in the target image based on the average probability distribution value.

[0021] Preferably, the line segment connectivity determination condition is specifically:

[0022]

[0023] Where seg1 and seg2 are two different line segments in the same row, start and end represent the coordinates of the left and right endpoints of the line segments, respectively. l is the distance factor, which is used to control the threshold of the distance between line segments. k is the scale factor, which is used to fine-tune the connection criteria of line segments at different scales.

[0024] A second aspect of the present application provides an object detection device based on information fusion, comprising:

[0025] An image acquisition unit, configured to acquire a target image to be processed;

[0026] A channel parameter calculation unit is used to traverse the pixel points in the target image and determine the parameter values of the four channels of each pixel point, wherein the four channels are specifically: R channel, G channel, B channel and V channel;

[0027] An image segmentation unit is configured to calculate the image segmentation thresholds of the four channels respectively through a preset image threshold segmentation device, and perform adaptive segmentation according to each of the image segmentation thresholds to obtain a segmented image corresponding to each channel;

[0028] An image fusion unit, configured to fuse the segmented images using DS evidence theory to obtain a fused image;

[0029] An image binarization processing unit, configured to binarize the fused image to obtain a binarized image;

[0030] an image compensation processing unit, configured to compensate the binary image by using a region dilation processing method, and determine, based on the compensated binary image and in combination with a line segment connectivity determination condition, connectivity relationships between line segments in the binary image, and then connect the line segments based on the connectivity relationships between the line segments, and merge connected domains to which the connected line segments belong;

[0031] The target object recognition unit is configured to calculate an average probability distribution value of the connected component, so as to recognize the target object in the target image based on the average probability distribution value.

[0032] Preferably, the image segmentation unit is specifically used for:

[0033] The image segmentation thresholds of the four channels are calculated respectively by the maximum inter-class variance method, and adaptive segmentation is performed according to each of the image segmentation thresholds to obtain the segmented images corresponding to each channel.

[0034] Preferably, the image fusion unit is specifically used for:

[0035] Obtain target position information of the target image, perform ellipse fitting processing on the detected target object image based on the target position information to obtain a circumscribed elliptical area of the target object image, wherein the detected target position information is position information of the target object in the target image obtained by pre-detecting the target image; calculate a probability distribution value of each pixel point based on the relative positional relationship between the coordinates of each pixel point in the segmented image and the circumscribed elliptical area; and fuse each of the segmented images based on the probability distribution values of each pixel point in combination with the Dempster rule to obtain a fused image.

[0036] Preferably, the target object recognition unit is specifically used to:

[0037] Based on the pixels in the connected domain and the probability distribution values of the pixels in the connected domain, an average probability distribution value of the connected domain is calculated, so as to identify the target object in the target image based on the average probability distribution value.

[0038] Preferably, the line segment connectivity determination condition is specifically:

[0039]

[0040] Where seg1 and seg2 are two different line segments in the same row, start and end represent the coordinates of the left and right endpoints of the line segments, respectively. l is the distance factor, which is used to control the threshold of the distance between line segments. k is the scale factor, which is used to fine-tune the connection criteria of line segments at different scales.

[0041] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0042] The target detection method provided in the present application first performs four-channel threshold segmentation, then uses the DS evidence theory to fuse the segmentation results of the four channels and binarizes the fused image, then compensates for the information loss in the fusion process through the region expansion method, combines the line segment connectivity judgment conditions, determines the connectivity relationship between each line segment in the binary image, obtains several connected domains, and finally calculates the average probability distribution value of the connected domain to determine the target object in the target image based on the average probability distribution value of each connected domain. The detection process does not need to rely on the training of a large amount of data, nor does it require a long time to wait for the algorithm training to be completed, which solves the technical problem of the time-consuming existing target detection methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0044] Figure 1 This is a flow chart of an embodiment of a target detection method based on information fusion provided in this application.

[0045] Figure 2 A flowchart of another embodiment of a target detection method based on information fusion provided in this application.

[0046] Figure 3 Schematic diagram of the connected domain merging process.

[0047] Figure 4 This is a diagram showing the relationship between basic probability distribution and reliability value and position.

[0048] Figure 5 This is a structural diagram of an embodiment of a target detection device based on information fusion provided by the present application. DETAILED DESCRIPTION

[0049] The embodiments of the present application provide a target detection method and device based on information fusion, which are used to solve the technical problems that existing target detection methods rely on a large number of training sets, require a lot of time for training, and are time-consuming.

[0050] In order to make the purpose, features, and advantages of the invention of this application more obvious and easy to understand, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the embodiments described below are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0051] See also Figure 1 The first embodiment of the present application provides a target detection method based on information fusion, including:

[0052] Step 101: Acquire a target image to be processed.

[0053] It should be noted that the target image to be processed is obtained through step 101. This target image should contain the target object to be detected. Then, it is processed through the three-stage processing method mentioned in the subsequent steps to detect the target object in the image.

[0054] Step 102: traverse the pixels in the target image and determine the parameter values of the four channels of each pixel.

[0055] The four channels are: R channel, G channel, B channel and V channel.

[0056] Step 103: Calculate the image segmentation thresholds of the four channels respectively using a preset image threshold segmentation method, and perform adaptive segmentation according to each image segmentation threshold to obtain a segmented image corresponding to each channel.

[0057] It should be noted that the threshold segmentation phase corresponding to steps 102 and 103 of this embodiment is first executed. The target image is traversed to determine the parameter values of the four channels for each pixel. Generally, the original image is an RGB three-channel image. In this case, the image can be converted into an HSV image to obtain the V channel. Based on the relationship between RGB and HSV channels, the entire image is traversed, and for each pixel, V(i,j) = max(R,G,B) is set to obtain the value of the V channel.

[0058] Next, calculate the R, G, B, V channels respectively to find the corresponding image segmentation threshold t R ,t G ,t B ,t V Then, these thresholds are applied to the four channels for segmentation to obtain a preliminary segmented image. The threshold calculation in the threshold segmentation stage can use the traditional grayscale histogram method, or the deep learning training method and the maximum inter-class variance method.

[0059] Step 104: fuse the segmented images using DS evidence theory to obtain a fused image.

[0060] Step 105: Binarize the fused image to obtain a binary image.

[0061] It should be noted that DS evidence theory uses hypothesis space, basic probability distribution function, belief function, and likelihood function to describe uncertainty. If Θ = a1, a2, ..., a n , where a i Represents the possible event i, then the power set of Θ is Θ = {Φ, {a1}, {a2}, {a1, a2}, ···, Θ}, forming a hypothesis space.

[0062] Steps 104 and 105 of this embodiment correspond to the information fusion stage. In the information fusion stage, the method of this embodiment uses the DS evidence theory to assign a probability function based on the target position information in the history, and then fuses the various segmented images according to the Dempster rule to obtain a fused image, and binarizes each pixel point.

[0063] Step 106: Compensate the binary image through regional expansion processing, and based on the compensated binary image and combined with the line segment connectivity determination conditions, determine the connectivity relationship between the various line segments in the binary image. Then, based on the connectivity relationship between the various line segments, connect the line segments and merge the connected domains to which the connected line segments belong.

[0064] Step 107 : Calculate the average probability distribution value of the connected component to identify the target object in the target image based on the average probability distribution value.

[0065] It should be noted that step 106 and step 107 of this embodiment correspond to the information compensation and detection stage. This stage compensates for the lost information and identifies the target object based on the region expansion method. First, the binary image obtained in the previous stage is traversed, the line segments in the image are counted, and expansion is performed based on the line segments. If two line segments meet the preset line segment connectivity judgment conditions, they are connected. After completing the line segment detection, when looking for connected domains, we no longer need to process at the pixel level, but can directly process the line segment data structure. In this section, we use the union-find algorithm to detect connected domains. This method regards each line segment as a node, each connected domain as a tree, and the merging of connected domains as the merging of trees. The algorithm steps are as follows:

[0066] 1) Create a tag array father with the length of the number of line segments, and initialize the entire array with father[i]=i, indicating that the parent node of the i-th line segment is itself. The initialization process is as follows Figure 2 As shown, each run now constitutes a tree, indicating that each run constitutes a separate connected domain during initialization.

[0067] 2) Traverse the entire structure array storing line segments in order from left to right and from top to bottom. i and L j Connected, first look for L i The root node and L j The root node of . Let L i The root point is L x ,L j The root node is L y , then let father[x] = y to merge the connected domain represented by x into y. The process is as follows Figure 3 As shown, Lx and Ly on the left have no parent nodes, indicating that each represents a connected domain. After traversing to nodes Li and Lj and determining that Li and Lj are connected, the algorithm then searches upward to Lx and Ly, respectively, assigning the father of one root node to the other. At this point, Lx is no longer the parent of itself, leaving only Ly as the parent. This completes the merging of the two connected domains. After traversing all paths, each remaining node with father[i] = i represents a connected domain. By counting these nodes, the number of connected domains can be determined.

[0068] All remaining connected domains are sorted by area, first screening them based on prior information about the object's area size. Ellipse fitting is then performed on the connected domains, calculating the ellipse's major axis, minor axis, eccentricity, and other parameters. Simultaneously, the average confidence level of the connected domains is calculated using the trust function for each pixel obtained based on DS evidence theory in step 104. This allows the target object in the target image to be identified based on the average probability distribution.

[0069] Furthermore, the conditions for determining whether a line segment is connected are specifically as follows:

[0070]

[0071] Where seg1 and seg2 are two different line segments in the same row, start and end represent the coordinates of the left and right endpoints of the line segments, respectively. l is the distance factor, which is used to control the threshold of the distance between line segments. k is the scale factor, which is used to fine-tune the connection criteria of line segments at different scales.

[0072] The above is a detailed description of the first embodiment of a target detection method based on information fusion provided by the present application. The target detection method provided in this embodiment first performs four-channel threshold segmentation, and then uses the DS evidence theory to fuse the segmentation results of the four channels and binarizes the fused image. Then, the regional expansion method is used to compensate for the information loss in the fusion process. Combined with the line segment connectivity judgment conditions, the connectivity relationship between each line segment in the binary image is determined to obtain several connected domains. Finally, the average probability distribution value of the connected domain is calculated to determine the target object in the target image based on the average probability distribution value of each connected domain. The detection process does not need to rely on the training of a large amount of data, nor does it require a long time to wait for the algorithm training to be completed, which solves the technical problem of the existing target detection method being time-consuming.

[0073] Based on the above embodiment, the present application further provides another embodiment of a target detection method based on information fusion, which specifically includes:

[0074] Furthermore, step 103 in the previous embodiment specifically includes the following steps:

[0075] The image segmentation thresholds of the four channels are calculated respectively by the maximum inter-class variance method, and adaptive segmentation is performed according to each image segmentation threshold to obtain the segmented image corresponding to each channel.

[0076] It should be noted that the segmentation threshold calculation in the threshold segmentation stage mentioned in the previous embodiment can use the traditional grayscale histogram method, the deep learning training method, and the maximum inter-class variance method. However, the actual application effect of the grayscale histogram method is general, and the deep learning training is very dependent on a large number of data sets, which cannot be provided in many actual engineering projects. Therefore, this embodiment preferably uses the maximum inter-class variance method to calculate the segmentation threshold. The process of calculating the segmentation threshold by the maximum inter-class variance method can refer to the following example:

[0077] For any threshold t, the image is divided into two categories C0 and C1. The pixel values in C0 are lower than the threshold, and the pixel values in C1 are higher than the threshold. For the entire image, the average grayscale value is:

[0078]

[0079] Calculate the grayscale mean of C0 and C1 respectively:

[0080]

[0081]

[0082]

[0083]

[0084] In summary:

[0085] u t =w0u0+w1u1

[0086] The between-class variance is defined as:

[0087]

[0088] When the above formula takes the maximum value, the inter-class variance of C0 and C1 is the largest, so the threshold t at this time can best separate the target object and the background.

[0089] Then find the corresponding segmentation threshold t for the four channels R, G, B, and V respectively. R ,t G ,t B ,t V Afterwards, these thresholds are applied to the four channels for segmentation, and the segmented image is preliminarily obtained.

[0090] Furthermore, the process of step 104 in the previous embodiment specifically includes:

[0091] Step 1041 : Acquire target position information of the target image, perform ellipse fitting processing on the detected target object image based on the target position information, and obtain a circumscribed ellipse area of the target object image.

[0092] The detected target position information is the position information of the target object in the target image obtained by pre-detecting the target image.

[0093] Step 1042: Calculate the probability distribution value of each pixel based on the relative positional relationship between the pixel coordinates in each segmented image and the circumscribed elliptical area.

[0094] Step 1043: Based on the probability distribution value of each pixel point, the segmented images are fused in combination with the Dempster rule to obtain a fused image.

[0095] It should be noted that the first step is to utilize the historical location information. Figure 4 As shown in the figure, an ellipse is fitted to the target object detected last time to obtain the circumscribed ellipse of the object. Assume that the ellipse equation is Then for each pixel point, if it falls inside the ellipse, it is assigned a probability of 1, otherwise the point will also fall on the concentric ellipse, and the basic probability distribution is based on the position of the concentric ellipse. Let the value of any channel C at point (i, j) after threshold segmentation be C(i, j). Define event<i,j> In channel C, pixel (i, j) belongs to the drone. The assignment probability is as follows:

[0096]

[0097]

[0098] in, In this way, after introducing historical location information, there will be a reasonable probability distribution.

[0099] Perform four-channel fusion according to Dempster's rule and binarize each pixel

[0100]

[0101] in

[0102]

[0103] The final fused pixel values are as follows:

[0104]

[0105] Furthermore, the process of step 107 in the previous embodiment specifically includes:

[0106] Based on the pixel points in the connected domain and the probability distribution values of each pixel point in the connected domain, an average probability distribution value of the connected domain is calculated to identify the target object in the target image based on the average probability distribution value.

[0107] The above content is a detailed description of the second embodiment of a target detection method based on information fusion provided by the present application. The following is a detailed description of an embodiment of a target detection device based on information fusion provided by the present application.

[0108] See also Figure 5 The third embodiment of the present application provides a target detection device based on information fusion, comprising:

[0109] The image acquisition unit 201 is used to acquire a target image to be processed.

[0110] The channel parameter calculation unit 202 is used to traverse the pixel points in the target image and determine the parameter values of the four channels of each pixel point, wherein the four channels are specifically: R channel, G channel, B channel and V channel.

[0111] The image segmentation unit 203 is used to calculate the image segmentation thresholds of the four channels respectively through a preset image threshold segmentation device, and perform adaptive segmentation according to each image segmentation threshold to obtain a segmented image corresponding to each channel.

[0112] The image fusion unit 204 is configured to fuse the segmented images using DS evidence theory to obtain a fused image.

[0113] The image binarization processing unit 205 is used to binarize the fused image to obtain a binarized image.

[0114] The image compensation processing unit 206 is used to compensate the binary image through regional expansion processing, and based on the compensated binary image and combined with the line segment connectivity determination conditions, determine the connectivity relationship between each line segment in the binary image, and then connect the line segments based on the connectivity relationship between the line segments and merge the connected domains to which the connected line segments belong.

[0115] The target object recognition unit 207 is configured to calculate an average probability distribution value of the connected component, so as to recognize the target object in the target image based on the average probability distribution value.

[0116] Furthermore, the image segmentation unit 203 is specifically configured to:

[0117] The image segmentation thresholds of the four channels are calculated respectively by the maximum inter-class variance method, and adaptive segmentation is performed according to each image segmentation threshold to obtain the segmented image corresponding to each channel.

[0118] Furthermore, the image fusion unit 204 is specifically configured to:

[0119] The target position information of the target image is obtained. Based on the target position information, an ellipse fitting process is performed on the detected target object image to obtain a circumscribed ellipse region of the target object image. The detected target position information is the position information of the target object in the target image obtained by pre-detection of the target image. The probability distribution value of each pixel in each segmented image is calculated based on the relative positional relationship between the coordinates of the pixel points and the circumscribed ellipse region. Based on the probability distribution values of each pixel point, the segmented images are fused in combination with the Dempster rule to obtain a fused image.

[0120] Furthermore, the target object recognition unit 207 is specifically configured to:

[0121] Based on the pixel points in the connected domain and the probability distribution values of each pixel point in the connected domain, an average probability distribution value of the connected domain is calculated to identify the target object in the target image based on the average probability distribution value.

[0122] Furthermore, the conditions for determining whether a line segment is connected are specifically as follows:

[0123]

[0124] Where seg1 and seg2 are two different line segments in the same row, start and end represent the coordinates of the left and right endpoints of the line segments, respectively. l is the distance factor, which is used to control the threshold of the distance between line segments. k is the scale factor, which is used to fine-tune the connection criteria of line segments at different scales.

[0125] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the terminals, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0126] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0127] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present application described herein can, for example, be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0128] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0129] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0130] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0131] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A target detection method based on information fusion, characterized in that: include: Obtain the target image to be processed; Traversing the pixel points in the target image, determining the parameter values of the four channels of each pixel point, wherein the four channels are specifically: R channel, G channel, B channel and V channel; Calculate the image segmentation thresholds of the four channels respectively by a preset image threshold segmentation method, and perform adaptive segmentation according to each of the image segmentation thresholds to obtain a segmented image corresponding to each channel; By using DS evidence theory, each of the segmented images is fused to obtain a fused image; Binarizing the fused image to obtain a binarized image; Compensating the binary image by region expansion processing, and determining connectivity relationships between line segments in the binary image based on the compensated binary image and a line segment connectivity determination condition, connecting the line segments based on the connectivity relationships between the line segments, and merging connected domains to which the connected line segments belong; An average probability distribution value of the connected component is calculated to identify the target object in the target image based on the average probability distribution value.

2. The target detection method based on information fusion according to claim 1, characterized in that: The method of calculating the image segmentation thresholds of the four channels respectively by using a preset image threshold segmentation method, and performing adaptive segmentation according to each of the image segmentation thresholds to obtain the segmented images corresponding to each channel specifically includes: The image segmentation thresholds of the four channels are calculated respectively by the maximum inter-class variance method, and adaptive segmentation is performed according to each of the image segmentation thresholds to obtain the segmented images corresponding to each channel.

3. The target detection method based on information fusion according to claim 1, characterized in that: The step of fusing the segmented images to obtain the fused image by using the DS evidence theory specifically includes: Acquiring target position information of the target image, and performing ellipse fitting processing on the detected target object image based on the target position information to obtain a circumscribed ellipse area of the target object image, wherein the detected target position information is position information of the target object in the target image obtained by pre-detecting the target image; Calculating a probability distribution value for each pixel based on a relative positional relationship between the pixel coordinates in each segmented image and the circumscribed elliptical area; Based on the probability distribution value of each pixel point, the segmented images are fused in combination with the Dempster rule to obtain a fused image.

4. The target detection method based on information fusion according to claim 3, characterized in that: Calculating the average probability distribution value of the connected component to identify the target object in the target image based on the average probability distribution value specifically includes: Based on the pixels in the connected domain and the probability distribution values of the pixels in the connected domain, an average probability distribution value of the connected domain is calculated, so as to identify the target object in the target image based on the average probability distribution value.

5. The target detection method based on information fusion according to claim 1, characterized in that: The line segment connectivity determination condition is specifically: Where seg1 and seg2 are two different line segments in the same row, start and end represent the coordinates of the left and right endpoints of the line segments, respectively. l is the distance factor, which is used to control the threshold of the distance between line segments. k is the scale factor, which is used to fine-tune the connection criteria of line segments at different scales.

6. A target detection device based on information fusion, characterized in that: include: An image acquisition unit, configured to acquire a target image to be processed; A channel parameter calculation unit is used to traverse the pixel points in the target image and determine the parameter values of the four channels of each pixel point, wherein the four channels are specifically: R channel, G channel, B channel and V channel; An image segmentation unit is configured to calculate the image segmentation thresholds of the four channels respectively through a preset image threshold segmentation device, and perform adaptive segmentation according to each of the image segmentation thresholds to obtain a segmented image corresponding to each channel; An image fusion unit, configured to fuse the segmented images using DS evidence theory to obtain a fused image; An image binarization processing unit, configured to binarize the fused image to obtain a binarized image; an image compensation processing unit, configured to compensate the binary image by using a region dilation processing method, and determine, based on the compensated binary image and in combination with a line segment connectivity determination condition, connectivity relationships between line segments in the binary image, and then connect the line segments based on the connectivity relationships between the line segments, and merge connected domains to which the connected line segments belong; The target object recognition unit is configured to calculate an average probability distribution value of the connected component, so as to recognize the target object in the target image based on the average probability distribution value.

7. The target detection device based on information fusion according to claim 6, characterized in that: The image segmentation unit is specifically used for: The image segmentation thresholds of the four channels are calculated respectively by the maximum inter-class variance method, and adaptive segmentation is performed according to each of the image segmentation thresholds to obtain the segmented images corresponding to each channel.

8. The target detection device based on information fusion according to claim 6, characterized in that: The image fusion unit is specifically used for: Obtain target position information of the target image, perform ellipse fitting processing on the detected target object image based on the target position information to obtain a circumscribed elliptical area of the target object image, wherein the detected target position information is position information of the target object in the target image obtained by pre-detecting the target image; calculate a probability distribution value of each pixel point based on the relative positional relationship between the coordinates of each pixel point in the segmented image and the circumscribed elliptical area; and fuse each of the segmented images based on the probability distribution values of each pixel point in combination with the Dempster rule to obtain a fused image.

9. The target detection device based on information fusion according to claim 8, characterized in that: The target object recognition unit is specifically used for: Based on the pixels in the connected domain and the probability distribution values of the pixels in the connected domain, an average probability distribution value of the connected domain is calculated, so as to identify the target object in the target image based on the average probability distribution value.

10. The target detection device based on information fusion according to claim 6, characterized in that: The line segment connectivity determination condition is specifically: Where seg1 and seg2 are two different line segments in the same row, start and end represent the coordinates of the left and right endpoints of the line segments, respectively. l is the distance factor, which is used to control the threshold of the distance between line segments. k is the scale factor, which is used to fine-tune the connection criteria of line segments at different scales.

Images