An Automatic Forensic Injury Annotation Method and System Based on Image Segmentation Technology

Abstract

This invention belongs to the field of forensic identification technology. It provides a method and system for automatic annotation of forensic injuries based on image segmentation technology. The method includes: an automatic annotation device capturing video images of the wound area targeted in the forensic examination; using image segmentation technology to extract the external shape data of the wound from the video images; a posture prediction model analyzing and processing the external shape data to predict the target's display posture; determining the target shooting angle based on the target's display posture; while the forensic examiner displays the wound area of ​​the target object, the automatic annotation device capturing high-definition images of the displayed wound area according to the target shooting angle; extracting wound feature data from the high-definition images; assessing the injury severity level of the wound area based on the wound feature data; automatically annotating the injury severity level in the high-definition images; and saving the data. This method enables automatic assessment of wound severity levels and automatic annotation in high-definition images, improving the efficiency of forensic injury examination.

Description

Technical Field

[0001] This invention relates to the field of forensic identification technology, and more specifically, to an automatic annotation method and system for forensic injury identification based on image segmentation technology. Background Technology

[0002] Injury examination is a professional process, usually conducted by forensic pathologists or personnel with forensic identification qualifications. However, current injury examination methods still require forensic pathologists to manually fill out injury reports, resulting in low efficiency in forensic injury assessment and failing to meet practical needs. Summary of the Invention

[0003] To address this issue, the present invention provides a method, system, electronic device, computer storage medium, and computer program product for automatic annotation of forensic injuries based on image segmentation technology, in order to solve the aforementioned technical problems.

[0004] This invention discloses an automatic annotation method for forensic injury identification based on image segmentation technology, the method comprising the following steps:

[0005] An automatic annotation device captures video images of the wound area targeted in the injury assessment operation, and uses image segmentation technology to extract the external shape data of the wound from the video images;

[0006] The posture prediction model analyzes and processes the external shape data to predict the target's display posture, and determines the target's shooting angle based on the target's display posture.

[0007] When a forensic examiner displays the wound area of ​​the target object, an automatic annotation device captures a high-definition image of the wound area in the display according to the target shooting angle, and extracts wound feature data from the high-definition image;

[0008] The severity level of the wound area is assessed based on the wound feature data, and the severity level is automatically marked in the high-definition image and saved.

[0009] In some embodiments, the automatic labeling device is deployed on the forensic examiner's body.

[0010] In some embodiments, image segmentation techniques are used to extract the external shape data of the wound from the video image, including:

[0011] The video image is analyzed based on color and texture features to identify the initial wound area and assess the linear complexity of the initial wound area.

[0012] The boundary distance is determined based on the linear complexity, and a first wound boundary line is determined in the wound region based on the boundary distance. The target wound region is extracted using image segmentation technology based on the first wound boundary line. The boundary distance refers to the distance between the first wound boundary line and the outer boundary of the initial wound region.

[0013] The target wound area is binarized, and the second wound boundary line is extracted from the binarized image. The second wound boundary line is used as the external shape data.

[0014] In some embodiments, the boundary distance and the linear complexity are negatively correlated, that is, the higher the linear complexity, the smaller the boundary distance, and the lower the linear complexity, the larger the boundary distance.

[0015] In some embodiments, the posture prediction model analyzes and processes the external shape data to predict the target's display posture, including:

[0016] The pose prediction model includes a semantic segmentation layer, several feature extraction layers, and a corresponding number of activation layers, fusion layers, and output layers.

[0017] The semantic segmentation layer is used to perform semantic processing and segmentation on the second wound boundary line to obtain several wound boundary line segments.

[0018] Each of the feature extraction layers extracts features from each of the wound boundary segments to obtain shape features;

[0019] The activation layer processes the shape features extracted by the corresponding feature extraction layer to obtain the predicted display posture;

[0020] The fusion layer performs fusion processing on each of the predicted display poses to obtain the target display pose;

[0021] The output layer displays the pose of the target.

[0022] In some embodiments, the automatic annotation device captures a high-resolution image of the wound area displayed at the target shooting angle, including:

[0023] The automatic annotation device extracts real-time external shape data of the wound area based on the video image and calculates the approximation between the real-time external shape data and the external shape data.

[0024] If the approximation is higher than the approximation threshold, then a high-resolution image of the wound area displayed is captured according to the target shooting angle.

[0025] This invention also discloses an automatic annotation system for forensic injury identification based on image segmentation technology. The system includes an automatic annotation device, which is used to implement the following method steps:

[0026] An automatic annotation device captures video images of the wound area targeted in the injury assessment operation, and uses image segmentation technology to extract the external shape data of the wound from the video images;

[0027] The attitude prediction model in the automatic annotation device analyzes and processes the external shape data to predict the target display attitude, and determines the target shooting angle based on the target display attitude.

[0028] When a forensic examiner displays the wound area of ​​the target object, an automatic annotation device captures a high-definition image of the wound area in the display according to the target shooting angle, and extracts wound feature data from the high-definition image;

[0029] The severity level of the wound area is assessed based on the wound feature data, and the severity level is automatically marked in the high-definition image and saved.

[0030] The present invention also discloses an electronic device comprising: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, the processor executing the computer program to implement the method as described in any of the preceding methods.

[0031] The present invention also discloses a computer storage medium storing a computer program that is executed by a processor to implement the methods described in any of the preceding methods.

[0032] The present invention also discloses a computer program product that, when run on a terminal, implements the method described in any of the preceding methods.

[0033] This invention enables automated wound assessment, quickly determining the severity of the injury, and automatically annotating it in high-definition images. This significantly reduces the workload of forensic pathologists and improves assessment efficiency. Furthermore, because the automatic annotation device of this invention captures wound images from a determined optimal shooting angle, the accuracy of subsequently extracted wound features is also high, ensuring the accuracy of the assessed injury severity. Attached Figure Description

[0034] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a flowchart illustrating an automatic annotation method for forensic injury identification based on image segmentation technology disclosed in an embodiment of the present invention.

[0036] Figure 2 This is a schematic diagram of the process for extracting the external shape data of a wound from video images, as disclosed in an embodiment of the present invention.

[0037] Figure 3 This is a schematic diagram of the preliminary wound area and the target wound area disclosed in the embodiments of the present invention;

[0038] Figure 4 This is a flowchart illustrating the process of capturing and displaying high-definition images of wound areas using an automatic annotation device disclosed in an embodiment of the present invention. Detailed Implementation

[0039] The following specific embodiments illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0040] Furthermore, the technical features involved in the different embodiments of this application described below can be combined with each other as long as they do not conflict with each other.

[0041] like Figure 1 As shown in the figure, this invention discloses an automatic annotation method for forensic injury identification based on image segmentation technology. The method includes the following steps:

[0042] An automatic annotation device captures video images of the wound area targeted in the injury assessment operation, and uses image segmentation technology to extract the external shape data of the wound from the video images;

[0043] The posture prediction model analyzes and processes the external shape data to predict the target's display posture, and determines the target's shooting angle based on the target's display posture.

[0044] When a forensic examiner displays the wound area of ​​the target object, an automatic annotation device captures a high-definition image of the wound area in the display according to the target shooting angle, and extracts wound feature data from the high-definition image;

[0045] The severity level of the wound area is assessed based on the wound feature data, and the severity level is automatically marked in the high-definition image and saved.

[0046] This invention deploys an automatic annotation device at the forensic injury scene. The device is equipped with a high-definition camera and an angle adjustment mechanism, enabling multi-angle imaging of the wound area. The automatic annotation device first captures video images of the wound area targeted for the forensic examination, from which the external shape data of the wound can be extracted using image segmentation technology.

[0047] When conducting injury assessments, forensic pathologists need to use specialized tools to properly display the wound, such as opening it to a certain degree. This allows automatic annotation equipment to capture high-definition images of the wound area with more detail, thereby accurately assessing the severity of the injury. The forensic pathologist's display posture is predictable. The posture prediction model constructed in this invention can analyze the external shape data of the wound to predict the forensic pathologist's display posture, i.e., the target display posture. After determining the target display posture, the target shooting angle of the automatic annotation equipment can be determined accordingly. That is, the automatic annotation equipment adjusts the shooting angle of the high-definition camera to the target shooting angle to obtain the best wound imaging effect.

[0048] This invention enables automated wound assessment, quickly determining the severity of the injury, and automatically annotating it in high-definition images. This significantly reduces the workload of forensic pathologists and improves assessment efficiency. Furthermore, because the automatic annotation device of this invention captures wound images from a determined optimal shooting angle, the accuracy of subsequently extracted wound features is also high, ensuring the accuracy of the assessed injury severity.

[0049] In some embodiments, the automatic labeling device is deployed on the forensic examiner's body.

[0050] Specifically, automatic annotation devices can be installed at the injury assessment site or worn, such as on the forensic examiner's head or shoulders, similar to the way action cameras are worn. Wearable automatic annotation devices offer greater flexibility and are less likely to obstruct the forensic examiner's view.

[0051] In some embodiments, such as Figure 2 , 3 As shown, image segmentation technology is used to extract the external shape data of the wound from the video image, including:

[0052] The video image is analyzed based on color and texture features to identify the initial wound area and assess the linear complexity of the initial wound area.

[0053] The boundary distance is determined based on the linear complexity, and a first wound boundary line is determined in the wound region based on the boundary distance. The target wound region is extracted using image segmentation technology based on the first wound boundary line. The boundary distance refers to the distance between the first wound boundary line and the outer boundary of the initial wound region.

[0054] The target wound area is binarized, and the second wound boundary line is extracted from the binarized image. The second wound boundary line is used as the external shape data.

[0055] Specifically, because wounds differ significantly from normal skin in color and texture, after capturing video images of the wound, the approximate area of ​​the wound is first identified based on color and texture feature analysis. A more accurate wound boundary line, i.e., the outer contour of the wound, also needs to be determined. Specifically, the linear complexity of the initial wound area is first assessed. Linear complexity refers to the number of intersecting lines contained in the image corresponding to the initial wound area. A higher number of intersecting lines indicates a more complex wound (for example, a wound from a triangular stab has more lines and more intersecting lines than a wound from a regular knife), and vice versa. Based on the linear complexity, the boundary distance can be determined, i.e., the distance between the first wound boundary line and the outer boundary of the initial wound area. Using the first wound boundary line as a reference, image segmentation techniques can segment the target wound area with as few non-wound areas as possible from the larger initial wound area that includes non-wound areas. This helps improve the accuracy of the subsequent second wound boundary line.

[0056] Binarizing the target wound area makes the contour boundary more prominent, allowing the extraction of the second wound boundary line (which largely coincides with the first wound boundary line but is more easily identified in the binarized image). The second wound boundary line contains shape features related to the trend of each point along the boundary line; for example, there is an inward fold at the lower left corner, an outward fold at the upper right corner, and other areas are smooth arcs. The second wound boundary line represents the external shape data.

[0057] In some embodiments, the boundary distance and the linear complexity are negatively correlated, that is, the higher the linear complexity, the smaller the boundary distance, and the lower the linear complexity, the larger the boundary distance.

[0058] Specifically, the aforementioned boundary distance and linear complexity are negatively correlated. A smaller boundary distance results in a larger target wound area cut from the initial wound area, reducing the probability of the cutting line crossing the true wound boundary and preserving more shape features. Conversely, a larger boundary distance results in a smaller target wound area cut from the initial wound area, ensuring the cutting line is as close as possible to the true wound boundary, reducing the image of non-wound areas, lowering the probability of misidentifying the wound boundary from non-wound areas, and thus ensuring the accuracy of the identified wound boundary.

[0059] In some embodiments, the posture prediction model analyzes and processes the external shape data to predict the target's display posture, including:

[0060] The pose prediction model includes a semantic segmentation layer, several feature extraction layers, and a corresponding number of activation layers, fusion layers, and output layers.

[0061] The semantic segmentation layer is used to perform semantic processing and segmentation on the second wound boundary line to obtain several wound boundary line segments.

[0062] Each of the feature extraction layers extracts features from each of the wound boundary segments to obtain shape features;

[0063] The activation layer processes the shape features extracted by the corresponding feature extraction layer to obtain the predicted display posture;

[0064] The fusion layer performs fusion processing on each of the predicted display poses to obtain the target display pose;

[0065] The output layer displays the pose of the target.

[0066] Specifically, the pose prediction model in this invention includes a semantic segmentation layer, several feature extraction layers, and a corresponding number of activation layers, fusion layers, and an output layer. The semantic segmentation layer (with an attention mechanism) first performs semantic analysis on the external shape data, i.e., the boundary line of the second wound, to determine several significant features of the boundary line (such as the inward folding direction of the lower left corner). Based on these significant features, the boundary line of the second wound is then segmented into multiple wound boundary line segments. The feature extraction layer and the corresponding activation layer then perform feature extraction and processing to obtain the corresponding prediction results, i.e., the predicted display pose. The fusion layer fuses all the prediction results, i.e., the predicted display pose, to obtain the final target display pose.

[0067] It should be noted that the target display posture mainly refers to the degree and angle (or orientation) of the tissue in the wound area being opened. The target shooting angle ensures that the shooting angle of the high-definition camera corresponds to the wound area when it is displayed according to the aforementioned degree and angle (or orientation). That is, at the target shooting angle, the high-definition camera can capture a "panoramic" image of the opened wound area, minimizing the situation where other tissues in the wound area obscure the details of the wound area (mainly the details inside the wound).

[0068] In some embodiments, such as Figure 4 As shown, the automatic annotation device captures high-resolution images of the wound area displayed at the target shooting angle, including:

[0069] The automatic annotation device extracts real-time external shape data of the wound area based on the video image and calculates the approximation between the real-time external shape data and the external shape data.

[0070] If the approximation is higher than the approximation threshold, then a high-resolution image of the wound area displayed is captured according to the target shooting angle.

[0071] Specifically, forensic pathologists use specialized tools to appropriately open up the wound area to reveal more wound details (especially internal wound details) to the automatic annotation device, which can then determine a more accurate injury severity level. When the automatic annotation device captures high-definition images of the wound area, it's necessary to determine if the forensic pathologist has adequately prepared for the proper presentation of the wound, i.e., whether the forensic pathologist has properly opened up the wound using specialized tools. Only if so can a high-definition image of the wound area be captured from the target shooting angle. To this end, the automatic annotation device in this invention extracts real-time external shape data of the wound area from the video image and calculates the approximation between this real-time external shape data and the external shape data (corresponding to the aforementioned target presentation posture). If the approximation is higher than an approximation threshold, it indicates that the forensic pathologist has adequately prepared for the proper presentation of the wound, meaning the internal wound details have been displayed to the expected degree. At this point, a high-definition image of the wound area can be captured from the target shooting angle.

[0072] This invention also discloses an automatic annotation system for forensic injury identification based on image segmentation technology. The system includes an automatic annotation device, which is used to implement the following method steps:

[0073] An automatic annotation device captures video images of the wound area targeted in the injury assessment operation, and uses image segmentation technology to extract the external shape data of the wound from the video images;

[0074] The attitude prediction model in the automatic annotation device analyzes and processes the external shape data to predict the target display attitude, and determines the target shooting angle based on the target display attitude.

[0075] When a forensic examiner displays the wound area of ​​the target object, an automatic annotation device captures a high-definition image of the wound area in the display according to the target shooting angle, and extracts wound feature data from the high-definition image;

[0076] The injury severity level of the wound area is assessed based on the wound feature data, and the injury severity level is automatically labeled in the high-definition image and saved. This invention also discloses an electronic device, including: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor. The processor executes the computer program to implement the method described in the foregoing embodiments.

[0077] This invention also discloses a computer storage medium storing a computer program, which is executed by a processor to implement the method described in the foregoing embodiments.

[0078] This invention also discloses a computer program product that, when run on a terminal, implements the methods described in the foregoing embodiments.

[0079] The various component embodiments of the present invention can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some or all of the components in the device according to embodiments of the present invention. The present invention can also be implemented as a device or apparatus program (e.g., a computer program and computer program product) for performing part or all of the methods described herein. Such programs implementing the present invention can be stored on a computer-readable medium or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0080] The terms "an embodiment," "embodiment," or "one or more embodiments" as used herein mean that a particular feature, structure, or characteristic described in connection with an embodiment is included in at least one embodiment of the invention. Furthermore, please note that the examples of the phrase "in one embodiment" do not necessarily all refer to the same embodiment.

[0081] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0082] It should be noted that the above embodiments are illustrative of the invention and not restrictive, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

[0083] Furthermore, it should be noted that the language used in this specification has been chosen primarily for readability and instructional purposes, rather than for interpreting or limiting the subject matter of the invention. Therefore, many modifications and alterations will be apparent to those skilled in the art without departing from the scope and spirit of the appended claims. The disclosure of this invention is illustrative, not restrictive, and the scope of the invention is defined by the appended claims.

Claims

1. A method for automatic annotation of forensic injuries based on image segmentation technology, characterized in that, The method includes the following steps: An automatic annotation device captures video images of the wound area targeted in the injury assessment operation, and uses image segmentation technology to extract the external shape data of the wound from the video images; The posture prediction model analyzes and processes the external shape data to predict the target's display posture, and determines the target's shooting angle based on the target's display posture. When a forensic examiner displays the wound area of ​​the target object, an automatic annotation device captures a high-definition image of the wound area in the display according to the target shooting angle, and extracts wound feature data from the high-definition image; The severity level of the wound area is assessed based on the wound feature data, and the severity level is automatically marked in the high-definition image and saved. The external shape data of the wound is extracted from the video image using image segmentation technology, including: The video image is analyzed based on color and texture features to identify the initial wound area and assess the linear complexity of the initial wound area. The boundary distance is determined based on the linear complexity, and a first wound boundary line is determined in the wound region based on the boundary distance. The target wound region is extracted using image segmentation technology based on the first wound boundary line. The boundary distance refers to the distance between the first wound boundary line and the outer boundary of the initial wound region. The target wound area is binarized, and the second wound boundary line is extracted from the binarized image. The second wound boundary line is used as the external shape data.

2. The automatic annotation method for forensic injury identification based on image segmentation technology according to claim 1, characterized in that: Automatic labeling equipment is installed on the forensic doctor's body.

3. The automatic annotation method for forensic injury identification based on image segmentation technology according to claim 1, characterized in that: The boundary distance and the linear complexity are negatively correlated; that is, the higher the linear complexity, the smaller the boundary distance, and the lower the linear complexity, the larger the boundary distance.

4. The automatic annotation method for forensic injury identification based on image segmentation technology according to claim 3, characterized in that: The posture prediction model analyzes and processes the external shape data to predict the target's display posture, including: The pose prediction model includes a semantic segmentation layer, several feature extraction layers, and a corresponding number of activation layers, fusion layers, and output layers. The semantic segmentation layer is used to perform semantic processing and segmentation on the second wound boundary line to obtain several wound boundary line segments. Each of the feature extraction layers extracts features from each of the wound boundary segments to obtain shape features; The activation layer processes the shape features extracted by the corresponding feature extraction layer to obtain the predicted display posture; The fusion layer performs fusion processing on each of the predicted display poses to obtain the target display pose; The output layer displays the pose of the target.

5. The automatic annotation method for forensic injury identification based on image segmentation technology according to claim 1, characterized in that: The automatic annotation device captures high-resolution images of the wound area displayed according to the target shooting angle, including: The automatic annotation device extracts real-time external shape data of the wound area based on the video image and calculates the approximation between the real-time external shape data and the external shape data. If the approximation is higher than the approximation threshold, then a high-resolution image of the wound area displayed is captured according to the target shooting angle.

6. An automatic annotation system for forensic injury identification based on image segmentation technology, the system comprising an automatic annotation device, characterized in that, The automatic labeling device is used to implement the following method steps: An automatic annotation device captures video images of the wound area targeted in the injury assessment operation, and uses image segmentation technology to extract the external shape data of the wound from the video images; The attitude prediction model in the automatic annotation device analyzes and processes the external shape data to predict the target display attitude, and determines the target shooting angle based on the target display attitude. When a forensic examiner displays the wound area of ​​the target object, an automatic annotation device captures a high-definition image of the wound area in the display according to the target shooting angle, and extracts wound feature data from the high-definition image; The severity level of the wound area is assessed based on the wound feature data, and the severity level is automatically marked in the high-definition image and saved. The external shape data of the wound is extracted from the video image using image segmentation technology, including: The video image is analyzed based on color and texture features to identify the initial wound area and assess the linear complexity of the initial wound area. The boundary distance is determined based on the linear complexity, and a first wound boundary line is determined in the wound region based on the boundary distance. The target wound region is extracted using image segmentation technology based on the first wound boundary line. The boundary distance refers to the distance between the first wound boundary line and the outer boundary of the initial wound region. The target wound area is binarized, and the second wound boundary line is extracted from the binarized image. The second wound boundary line is used as the external shape data.

7. An electronic device, comprising: At least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, characterized in that: the processor executes the computer program to implement the method as claimed in any one of claims 1-5.

8. A computer storage medium storing a computer program, characterized in that: The computer program is executed by a processor to implement the method as described in any one of claims 1-5.

9. A computer program product, characterized in that: When the computer program product is run on a terminal, it implements the method as described in any one of claims 1-5.

Images