WO2021147435A1 - 伤情图片自动化审核方法、装置、电子设备及存储介质 - Google Patents

伤情图片自动化审核方法、装置、电子设备及存储介质 Download PDF

Info

Publication number
WO2021147435A1
WO2021147435A1 PCT/CN2020/125071 CN2020125071W WO2021147435A1 WO 2021147435 A1 WO2021147435 A1 WO 2021147435A1 CN 2020125071 W CN2020125071 W CN 2020125071W WO 2021147435 A1 WO2021147435 A1 WO 2021147435A1
Authority
WO
WIPO (PCT)
Prior art keywords
area
face
injury
picture
meets
Prior art date
Application number
PCT/CN2020/125071
Other languages
English (en)
French (fr)
Inventor
宁培阳
Original Assignee
平安科技（深圳）有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 平安科技（深圳）有限公司 filed Critical 平安科技（深圳）有限公司
Publication of WO2021147435A1 publication Critical patent/WO2021147435A1/zh

Links

Images

Classifications

    • 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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/08—Insurance
    • 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
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/90—Determination of colour characteristics
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/161—Detection; Localisation; Normalisation

Definitions

  • This application relates to smart medical care, and in particular to an automated review method, device, electronic equipment, and computer-readable storage medium for injury pictures for auto insurance claims.
  • the traditional method of obtaining injury pictures is to send a car insurance claim adjuster to the scene or hospital to take pictures of the injury. Although this solution can reliably obtain injury pictures that meet the claims requirements, it requires a high labor cost and transportation cost.
  • the current popular deep learning methods can better build an automated review process for injury pictures and achieve better practical results. But there are other outstanding problems.
  • One is that training deep learning models requires a large number of injury pictures. Injury pictures are obviously private information of customers. Therefore, using customer injury pictures for model training has a high risk of information violations.
  • the inventor found that although the effect of the deep learning-based solution is ideal, it may not be realized due to the lack of compliance data; the second is that the deep learning model is complicated in calculation and requires high hardware resources. At present, it is mainly through the function of It is difficult to deploy on remote high-performance servers, but it is difficult to deploy directly on the client's mobile terminal. This results in high operating costs for deep learning solutions, and the fundamental purpose of reducing claims costs is still not well achieved.
  • This application provides an automatic review, device, electronic equipment, and computer-readable storage medium for injury pictures for auto insurance claims.
  • the main purpose of the application is to increase the probability of success in shooting an injury picture that meets the requirements and improve customer experience.
  • the present application provides an automated review method for injury pictures, the method includes:
  • Reviewing the skin area of the injury picture includes: detecting the skin area in the injury picture based on the color space threshold segmentation; judging whether the area of the detected skin area meets the area size requirement; if the area of the detected skin area is not Meet the area size requirements, send the first unqualified instruction to the client to re-obtain the injury picture to be reviewed.
  • the first unqualified instruction includes the abnormal light instruction; if the detected area of the skin area meets the area size requirements, The detected skin area is separated from the wound area;
  • the unqualified instruction includes the long shooting distance; if the separated wound area meets the wound area requirements, locate the injury picture based on the template matching method 'S face
  • Reviewing the face of the injury picture includes: judging whether the positioned face meets the face requirements; if the positioned face does not meet the face requirements, send the third unqualified instruction to the client, and obtain the desired review again
  • the third unqualified instruction includes adjusting the shooting angle or adjusting the shooting distance; if the positioned face meets the face requirements, obtain the injured person’s certificate, perform face detection on the injured person’s certificate, and obtain the information in the injured person’s certificate human face;
  • Review the similarity between the face in the injury picture and the face in the injured person’s certificate including: comparing the similarity between the face in the certificate and the face in the injury picture; judging the comparison result of the similarity comparison Whether it meets the similarity requirements; if the comparison result does not meet the similarity requirements, send the fourth unqualified instruction to the client to re-obtain the injury picture to be reviewed.
  • the fourth unqualified instruction includes the wrong document; if the comparison result Meet the similarity requirements, upload the injury picture.
  • this application also provides an automatic review device for injury pictures, the device includes:
  • the first reviewing department which reviews the skin area of the injury picture, includes: a skin area detection module, which detects the skin area in the injury picture based on color space threshold segmentation; the first area judgment module, which judges the detected skin area Whether the area meets the area size requirements, if the detected area of the skin area does not meet the area size requirements, send the first unqualified instruction to the client, and the acquisition department will re-obtain the injury picture to be reviewed.
  • the first unqualified instruction includes Light abnormality instruction; if the detected area of the skin area meets the area size requirement, send a signal to the wound area acquisition module; the wound area acquisition module separates the wound area from the detected skin area and sends it to the second review department;
  • the second reviewing part reviewing the wound area of the injury picture, includes: a wound area detection module, which detects the wound area in the skin area obtained by the first reviewing part based on color space threshold segmentation; a second area judgment module, judging Whether the separated wound area meets the wound area requirements, if the separated wound area does not meet the wound area requirements, send a second unqualified instruction to the client, and the acquisition department re-obtains the injury picture to be reviewed, and the second unqualified
  • the instructions include a long shooting distance; if the separated wound area meets the wound area requirements, send a signal to the first face acquisition module; the first face acquisition module locates the face that appears in the injury picture based on the template matching method and sends it To the third review department and the similarity acquisition department;
  • the third review department reviews the face of the injury picture, including: a face judging module, to determine whether the positioned face meets the face requirements, if the positioned face does not meet the face requirements, send the third unqualified Instruction is given to the client, and the acquisition department reacquires the injury picture to be reviewed.
  • the third unqualified instruction includes adjusting the shooting angle or adjusting the shooting distance; if the positioned face meets the requirements of the face, send a signal to the second face to obtain Module; the second face acquisition module, which obtains the injured person’s certificate, performs face detection on the injured person’s certificate, obtains the face in the injured person’s certificate and sends it to the similarity acquisition department;
  • the similarity obtaining section compares the face in the certificate with the face in the injury picture, and judges whether the comparison result of the similarity comparison meets the similarity requirements, and if the comparison result does not meet the similarity requirements, send the first Four unqualified instructions are sent to the client, and the acquisition department re-obtains the injury picture to be reviewed.
  • the fourth unqualified instruction includes the wrong document; if the comparison result meets the similarity requirements, a signal is sent to the uploading department;
  • Upload department upload injury pictures that meet the requirements of the first review department, the second review department, the third review department and the similarity acquisition department.
  • an electronic device which includes:
  • At least one processor and,
  • a memory communicatively connected with the at least one processor; wherein,
  • the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor, so that the at least one processor can execute the following steps:
  • Reviewing the skin area of the injury picture includes: detecting the skin area in the injury picture based on the color space threshold segmentation; judging whether the area of the detected skin area meets the area size requirement; if the area of the detected skin area is not Meet the area size requirements, send the first unqualified instruction to the client to re-obtain the injury picture to be reviewed.
  • the first unqualified instruction includes the abnormal light instruction; if the detected area of the skin area meets the area size requirements, The detected skin area is separated from the wound area;
  • the unqualified instruction includes the long shooting distance; if the separated wound area meets the wound area requirements, locate the injury picture based on the template matching method 'S face
  • Reviewing the face of the injury picture includes: judging whether the positioned face meets the face requirements; if the positioned face does not meet the face requirements, send the third unqualified instruction to the client, and obtain the desired review again
  • the third unqualified instruction includes adjusting the shooting angle or adjusting the shooting distance; if the positioned face meets the face requirements, obtain the injured person’s certificate, perform face detection on the injured person’s certificate, and obtain the information in the injured person’s certificate human face;
  • Review the similarity between the face in the injury picture and the face in the injured person’s certificate including: comparing the similarity between the face in the certificate and the face in the injury picture; judging the comparison result of the similarity comparison Whether it meets the similarity requirements; if the comparison result does not meet the similarity requirements, send the fourth unqualified instruction to the client to re-obtain the injury picture to be reviewed.
  • the fourth unqualified instruction includes the wrong document; if the comparison result Meet the similarity requirements, upload the injury picture.
  • the present application also provides a computer-readable storage medium storing a computer program, wherein the computer program is executed by a processor to implement the following steps:
  • Reviewing the skin area of the injury picture includes: detecting the skin area in the injury picture based on the color space threshold segmentation; judging whether the area of the detected skin area meets the area size requirement; if the area of the detected skin area is not Meet the area size requirements, send the first unqualified instruction to the client to re-obtain the injury picture to be reviewed.
  • the first unqualified instruction includes the abnormal light instruction; if the detected area of the skin area meets the area size requirements, The detected skin area is separated from the wound area;
  • the unqualified instruction includes the long shooting distance; if the separated wound area meets the wound area requirements, locate the injury picture based on the template matching method 'S face
  • Reviewing the face of the injury picture includes: judging whether the positioned face meets the face requirements; if the positioned face does not meet the face requirements, send the third unqualified instruction to the client, and obtain the desired review again
  • the third unqualified instruction includes adjusting the shooting angle or adjusting the shooting distance; if the positioned face meets the face requirements, obtain the injured person’s certificate, perform face detection on the injured person’s certificate, and obtain the information in the injured person’s certificate human face;
  • Review the similarity between the face in the injury picture and the face in the injured person’s certificate including: comparing the similarity between the face in the certificate and the face in the injury picture; judging the comparison result of the similarity comparison Whether it meets the similarity requirements; if the comparison result does not meet the similarity requirements, send the fourth unqualified instruction to the client to re-obtain the injury picture to be reviewed.
  • the fourth unqualified instruction includes the wrong document; if the comparison result Meet the similarity requirements, upload the injury picture.
  • the automated review method, device, electronic equipment, and computer-readable storage medium for injury pictures described in this application guide customers in real time to adjust the ambient light, distance, and shooting angle when shooting injury pictures, so as to increase the probability of success in one shot of injury pictures that meet the requirements
  • it adopts a variety of lightweight digital image processing methods based on color space threshold segmentation and template matching. The calculation is small, and it can be deployed on the customer’s mobile phone, reducing the cost of deploying a remote server.
  • users were reminded in real time to make corresponding adjustments, so as to ensure that effective injury pictures were taken while improving the customer’s experience. More importantly, it can be automatically reviewed offline.
  • Figure 1 is a flow chart of the automatic review method for injury pictures described in this application.
  • FIG. 2 is a schematic diagram of modules of an automatic review device for injury pictures provided by an embodiment of the present application
  • FIG. 3 is a schematic diagram of the internal structure of an electronic device for realizing an automatic review method for injury pictures according to an embodiment of the present application.
  • Fig. 1 is a flowchart of the method for automatically reviewing injury pictures described in this application. As shown in Fig. 1, the method for automatically reviewing injury pictures includes:
  • Step S1 Obtain the injury picture to be reviewed
  • Step S2 reviewing the skin area of the injury picture, including: detecting the skin area in the injury picture based on color space threshold segmentation; judging whether the area of the detected skin area meets the area size requirement; if the detected skin area If the area does not meet the area size requirements, send the first unqualified instruction to the client, return to step S1, and re-obtain the injury picture to be reviewed.
  • the first unqualified instruction includes an abnormal light instruction; if the detected skin area is The area meets the area size requirement, the wound area is separated from the detected skin area, and step S3 is performed;
  • Step S3 reviewing the wound area of the injury picture, including: detecting the wound area based on color space threshold segmentation; judging whether the separated wound area meets the wound area requirement; if the separated wound area does not meet the wound area requirement, Send a second unqualified instruction to the client and return to step S1 to retrieve the injury picture to be reviewed.
  • the second unqualified instruction includes a long shooting distance; if the separated wound area meets the wound area requirements, based on template matching
  • the method locates the face appearing in the injury picture, and executes step S4;
  • Step S4 reviewing the face of the injury picture, including: determining whether the located face meets the face requirements; if the located face does not meet the face requirements, sending a third unqualified instruction to the client, and returning to step S1.
  • the third unqualified instruction includes adjusting the shooting angle or adjusting the shooting distance; if the positioned face meets the face requirements, obtain the injured person’s certificate, and perform face detection on the injured person’s certificate , Obtain the face in the injured person’s certificate, and execute step S5;
  • Step S5 reviewing the similarity between the face in the injury picture and the face in the injured person’s certificate, including: comparing the similarity between the face in the certificate and the face in the injury picture; judging the similarity comparison If the comparison result does not meet the similarity requirements; if the comparison result does not meet the similarity requirements, send a fourth unqualified instruction to the client, and return to step S1 to retrieve the injury picture to be reviewed.
  • the fourth unqualified instruction includes transmission Wrong document; if the comparison result meets the similarity requirements, go to step S6;
  • Step S6 upload an injury picture that meets the requirements of steps S2-S5.
  • step S2 the step of reviewing the skin area of the injury picture includes:
  • the illumination of the injury picture does not meet the first judgment condition, issue a first unqualified instruction, where the first unqualified instruction includes too dark, overexposed, or uneven lighting;
  • the illumination of the injury picture meets the first criterion, analyze the uniformity of the illumination of the injury picture based on the color space;
  • a first unqualified instruction is issued, and the first unqualified instruction includes partial over-darkness or partial over-exposure;
  • the illumination uniformity of the injury picture satisfies the second discrimination condition, segment the injury picture based on the color space threshold to obtain the skin area of the injury picture, and obtain the area of the skin area;
  • the wound area is separated from the detected skin area.
  • the step of segmenting the injury picture based on the color space threshold to obtain the skin area of the injury picture includes:
  • the internal voids of the area enclosed by the pixel points are filled by the morphological closing operation to obtain one or more closed areas, thereby obtaining the skin area.
  • step S2 includes:
  • Step S21 Convert the injury picture from the RGB color space to the HSV color space to obtain the average value of the R channel, G channel, B channel and V channel of the injury picture
  • the first criterion for normal light in the overall environment is:
  • ⁇ R , ⁇ G , and ⁇ B are respectively the upper limit of the average value of the R channel, the G channel, and the B channel. If the average value of a channel is higher than the upper limit, it indicates that there is a serious deviation in the light color temperature; ⁇ V1 , ⁇ V2 are the average values of the V channel the lower and upper limits, described below ⁇ V1 light is too dark, ⁇ V2 described above illumination is too bright; determining when a first condition is satisfied, performing step S22, the first determination condition is not satisfied when, according to the R channel, G channel , B channel and V channel average Different first unqualified instructions are issued, and the first unqualified instructions include too dark light, overexposed light, or uneven light.
  • ⁇ VL is the lower limit of the average relative brightness of the sub-block and the entire image. A value lower than this indicates that the injury picture has a local over-darkness problem; ⁇ VH is the upper limit of the average relative brightness of the sub-block and the entire image, which is higher than this value.
  • the injury picture has the problem of partial overexposure.
  • Step S23 skin area detection
  • the injury image is converted from RGB color space to YC r C b , HSV and other color spaces.
  • each pixel is represented by three color spaces
  • P ij (R, G, B, Y, C r , C b , H, S, V)
  • the pixels whose P ij meets the third criterion are filtered out, and the pixel points are filled in by the morphological closing operation
  • the fourth criterion is:
  • P ij belongs to a point in the skin area
  • the thresholds on both sides of each inequality are set for the skin
  • n is the number of pixels that meet the third criterion, if Indicates that the skin area is too small.
  • step S3 the step of reviewing the wound area of the injury picture includes:
  • a second unqualified instruction is issued, and the second unqualified instruction includes that the wound area is too small and the shooting distance is too far;
  • the face in the injury picture is located based on the template matching method.
  • the step of segmenting the skin area based on the color space threshold to obtain the wound area in the skin area includes:
  • the internal gaps of the area enclosed by the pixels are filled by the morphological closing operation to obtain one or more closed areas, thereby obtaining the wound area.
  • the step of locating the face appearing in the injury picture by the method based on template matching includes:
  • An average face is made from the currently public face data set and used as a template, and the gray-scale image is scanned and detected using a template matching algorithm to detect the face.
  • step S3 includes:
  • Step S31 Convert the image of the skin area from the RGB color space to YC r C b , HSV and other color spaces.
  • the sixth criterion is:
  • p′ ij belongs to a point in the wound area
  • the thresholds on both sides of each inequality are set for the wound
  • n′ is the number of pixels that meet the fifth criterion, if Indicates that the wound area is too small.
  • Gray ij R 'ij * 0.299 + G' ij * 0.587 + B 'ij * 0.114.
  • step S33 an "average face" is made from the currently public face data set and used as a template, and a template matching algorithm is used to scan and detect the grayscale image to detect a human face.
  • step S4 the step of reviewing the face of the injury picture includes:
  • Non-maximum suppression algorithm (Non Maximum Suppression) to eliminate overlapping and redundant faces detected at the same position
  • a third unqualified instruction is issued, and the third unqualified instruction includes adjusting the shooting distance of the injured person or adjusting the shooting angle of the face;
  • a fourth unqualified instruction is issued, and the fourth unqualified instruction includes that persons other than the injured temporarily leave the shooting screen;
  • the injured person’s certificate is obtained, and the face of the injured person’s certificate is detected to obtain the face of the injured person’s certificate.
  • the injured person’s certificate is an ID card, a driver’s license, etc., with a fixed format (with a fixed template).
  • the second face block is intercepted from the ID card, and the avatar is extracted due to the fixed format of the ID card. It is relatively simple and can be replaced with other simpler methods, for example, because the position of the avatar of the ID card is fixed, the avatar is intercepted according to the relative position.
  • the step of obtaining the face in the injured person’s certificate includes:
  • the head is taken based on the relative position.
  • step S5 the step of reviewing the similarity between the face in the injury picture and the face in the injured person’s certificate includes:
  • Face extraction according to the position of the detected face relative to the entire image, intercept one of the faces as the first face sub-block; intercept the second face sub-block from the ID card;
  • the first face sub-block and the second face sub-block are uniformly scaled, and LBP features (Local Binary Patterns) are extracted respectively;
  • the similarity does not meet the similarity requirements, send a fourth unqualified instruction to the client to re-obtain the injury picture to be reviewed.
  • the fourth unqualified instruction includes the wrong document (for example, the client uploads its own by mistake).
  • the degree of distinguishing features of different positions of the face is different.
  • the importance of the features of the positions of the eyes and the nose is greater than that of the edge of the face close to the background.
  • Different weights are assigned to the different positions of the face according to the importance of the features. The greater the importance of the feature, the greater the weight.
  • a fixed weight adjustment vector W is designed to distinguish the weights of different positions of the human face.
  • F source is the face LBP feature vector extracted from the injury image
  • F target is the face LBP feature vector extracted from the ID card image
  • ⁇ is the multiplication operation of the vector elements at the same position
  • ⁇ is the vector modulus Long operation
  • S is the similarity between the first face sub-block and the second face sub-block, 0 ⁇ S ⁇ 1, the larger S is, the more similar the two LBP features are, which can be used to characterize the similarity of the faces.
  • step S6 If S ⁇ S min , that is, S is higher than the threshold S min , it is considered that the face matches, and step S6 is executed; otherwise, the fourth unqualified instruction is issued.
  • skin area detection skin area separated from injury pictures.
  • the purpose of this step is to reduce the probability of false detection of the wound area (the wound only exists on the skin).
  • the skin area is separated from the wound area.
  • the purpose of this step is to verify that there is an injured part in the picture.
  • the skin of the injured person fails to pass because the distance between the lens and the injured person is too far, resulting in the detection If the wound area is too small, prompt the customer to adjust the shooting distance accordingly; face detection: locate the face that appears in the image, and if the previous steps are passed, the injured person’s face detection fails because the lens did not capture the face Or the face is not facing the lens, accordingly, the customer is prompted to adjust the shooting distance or adjust the shooting angle.
  • Possible technical solutions include a face detection method based on template matching.
  • the template matching algorithm realizes fast face detection; face similarity comparison: compare the similarity of the face in the ID (usually ID card) picture with the face in the injury picture.
  • face similarity comparison compare the similarity of the face in the ID (usually ID card) picture with the face in the injury picture.
  • the purpose of this step is to verify the injury. The identity of the person who protects against fraudulent situations such as insurance fraud.
  • the steps S2-S5 of the automated review method for injury pictures described in this application have a small overall calculation amount, so the equipment performance requirements are not high.
  • the mobile CPU can complete the calculations required for processing within a few seconds, meeting the "offline" design requirements , While realizing the fully automated review of injury pictures, it reduces the deployment cost of high-performance servers and further achieves the fundamental purpose of reducing injury picture reviews.
  • the automatic review method for injury pictures described in this application can be implemented to remind the user to make corresponding adjustments in real time before the user presses the shutter, so as to ensure that the effective shooting is achieved. Improve the customer experience while hurting the picture.
  • the automatic review method for injury pictures described in this application may not directly use injury pictures.
  • use the parameters recommended by published papers to complete the design of the skin detection model and use the published face detection data set to complete the design of the face detection model. And so on, so as to avoid the use of private data.
  • FIG. 2 is a block diagram of the structure of the device for automatic review of injury pictures according to the present application.
  • the device for automatic review of injury pictures 100 can be installed in an electronic device.
  • the data auditing device may include an obtaining unit 110, a first auditing unit 120, a second auditing unit 130, a third auditing unit 140, a similarity obtaining unit 150, and an uploading unit 160.
  • the first auditing unit 120 includes The skin area detection module 121, the first area judgment module 122, and the wound area obtaining module 123.
  • the second review unit 130 includes the wound area detection module 131, the second area judgment module 132, and the first face acquisition module 133.
  • the third review unit 140 includes a face judging module 141 and a second face obtaining module 142.
  • the part/module of the present invention refers to a series of computer program segments that can be executed by the processor of an electronic device and can complete fixed functions, which are stored in the memory of the electronic device.
  • each part/module is as follows:
  • the first review unit 120 reviews the skin area of the injury picture, including: a skin area detection module 121, which detects the skin area in the injury picture based on color space threshold segmentation; the first area determination module 122, determines the detected skin area Whether the area of the skin area meets the area size requirement, if the detected area of the skin area does not meet the area size requirement, the first unqualified instruction is sent to the client, and the acquisition department 110 reacquires the injury picture to be reviewed.
  • the unqualified instruction includes an abnormal light instruction; if the area of the detected skin area meets the area size requirement, a signal is sent to the wound area obtaining module 123; the wound area obtaining module 123 separates the wound area from the detected skin area and sends it to the first 2. Audit Department 130;
  • the second reviewing unit 130 reviews the wound area of the injury picture, including: a wound area detection module 131, which detects the wound area in the skin area obtained by the first reviewing unit 120 based on color space threshold segmentation; second area judgment Module 132, judge whether the separated wound area meets the wound area requirement, if the separated wound area does not meet the wound area requirement, send a second unqualified instruction to the client, the acquisition department 110 re-obtains the injury picture to be reviewed, so The second unqualified instruction includes a long shooting distance; if the separated wound area meets the wound area requirement, a signal is sent to the first face obtaining module 133; the first face obtaining module 133 locates the injury picture based on the template matching method And send the face appearing in the third review unit 140 and the similarity obtaining unit 150;
  • the third review unit 140 reviews the face of the injury picture, including: a face judging module 141, judging whether the positioned face meets the face requirements, and if the positioned face does not meet the face requirements, send the third The unqualified instruction is sent to the client, and the acquisition unit 110 reacquires the injury picture to be reviewed.
  • the third unqualified instruction includes adjusting the shooting angle or adjusting the shooting distance; if the positioned face meets the requirements of the face, it sends a signal to the second Face obtaining module 142;
  • the second face obtaining module 142 obtains the injured person’s certificate, performs face detection on the injured person’s certificate, obtains the face in the injured person’s certificate and sends it to the similarity obtaining unit 150;
  • the similarity obtaining unit 150 compares the human face in the certificate with the human face in the injury picture to determine whether the comparison result of the similarity comparison meets the similarity requirement, and if the comparison result does not meet the similarity requirement, send The fourth unqualified instruction is sent to the client, and the obtaining unit 110 reacquires the injury picture to be reviewed.
  • the fourth unqualified instruction includes the wrong document; if the comparison result meets the similarity requirements, a signal is sent to the uploading unit 160;
  • the uploading unit 160 uploads injury pictures that meet the requirements of the first reviewing unit 120, the second reviewing unit 130, the third reviewing unit 140, and the similarity obtaining unit 150.
  • the aforementioned skin area detection module 121 includes:
  • the illumination acquisition unit analyzes the illumination of the injury picture based on the color space
  • the first discrimination condition setting unit sets the illumination threshold of each parameter in the color space to obtain the first discrimination condition
  • the illumination judgment unit judges whether the illumination of the injury picture meets the first judgment condition; if the illumination of the injury picture does not meet the first judgment condition, a first unqualified instruction is issued.
  • the first unqualified instruction includes excessively dark illumination and illumination Overexposure or uneven illumination; if the illumination of the injury picture meets the first judgment condition, send a signal to the uniformity obtaining unit;
  • the uniformity obtaining unit analyzes the uniformity of illumination of the injury picture based on the color space
  • the second discriminant condition setting unit sets the threshold of the illumination uniformity to obtain the second discriminant condition
  • the uniformity judgment unit judges whether the uniformity of the illumination of the injury picture meets the second judgment condition; if the uniformity of the illumination of the injury picture does not meet the second judgment condition, a first disqualification instruction is issued, and the first disqualification
  • the instructions include partial over-darkness or partial over-exposure; if the uniformity of the illumination of the injury picture meets the second judgment condition, a signal is sent to the skin area obtaining unit;
  • the skin area obtaining unit divides the injury picture based on the color space threshold to obtain the skin area and the area of the skin area of the injury picture.
  • the skin area obtaining unit includes:
  • Color space conversion sub-unit to obtain data in multiple color spaces of the injury picture
  • the third discriminating condition obtaining subunit sets the threshold of each parameter in the multiple color spaces corresponding to the skin area to obtain the third discriminating condition
  • the screening subunit selects multiple pixels that meet the third criterion from the data in multiple color spaces of the injury picture
  • the filling unit fills the internal voids of the area enclosed by the pixel points through the morphological closing operation to obtain one or more closed areas, thereby obtaining the skin area.
  • the implementation of the wound area detection module 131 is similar to the implementation of the aforementioned skin area detection module 121. Since the skin area detection module 121 has already detected the illumination and uniformity of the injury picture, the wound area detection module 131 only needs to detect the wound. The area of the area.
  • the aforementioned first face obtaining module 133 includes:
  • the gray scale conversion unit converts the RGB image of the injury picture into a gray scale image
  • the template matching unit uses the currently public face data set to produce an average face and uses it as a template, and scans and detects the gray image using a template matching algorithm to detect the face.
  • the aforementioned face judgment module 141 includes:
  • Elimination unit using non-maximum suppression algorithm to eliminate overlapping and redundant faces detected at the same position
  • the face counting unit counts the faces in the injury picture processed by the elimination unit, and if no face is detected, a third unqualified instruction is issued.
  • the third unqualified instruction includes the injured person adjusting the shooting distance or adjusting the person The camera angle of the face; if two or more faces are detected, a fourth unqualified instruction is issued.
  • the fourth unqualified instruction includes persons other than the injured temporarily leaving the shooting screen; if one face is detected, a signal is sent to The second face obtaining module 142.
  • the similarity obtaining unit 150 includes:
  • the face extraction unit according to the position of the detected face relative to the entire image, intercepts one face as the first face sub-block; intercepts the second face sub-block from the ID card;
  • the feature extraction unit scales the first face sub-block and the second face sub-block uniformly, and extracts LBP features respectively;
  • the similarity matching unit uses the similarity method to perform face feature matching according to the extracted LBP features of the first face sub-block and the second face sub-block to obtain the similarity between the first face sub-block and the second face sub-block;
  • the similarity judging unit judges whether the similarity meets the similarity requirements; if the similarity does not meet the similarity requirements, it sends a fourth unqualified instruction to the client to re-obtain the injury picture to be reviewed, and the fourth The unqualified instruction includes the wrong document; if the similarity meets the similarity requirement, a signal is sent to the uploading unit 160.
  • the automated review device for injury pictures described in this application can effectively reduce the review cost of injury pictures, improve claims efficiency, and improve customer experience. Based on a variety of lightweight digital image processing methods and machine learning methods, offline injuries can be realized Picture review.
  • FIG. 3 it is a schematic diagram of the structure of an electronic device that implements the automatic review method for injury pictures in this application.
  • the electronic device 1 may include a processor 10, a memory 11, and a bus, and may also include a computer program stored in the memory 11 and running on the processor 10, such as an injury picture automated review program 12.
  • the memory 11 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (such as SD or DX memory, etc.), magnetic memory, magnetic disk, CD etc.
  • the memory 11 may be an internal storage unit of the electronic device 1 in some embodiments, for example, a mobile hard disk of the electronic device 1.
  • the memory 11 may also be an external storage device of the electronic device 1, such as a plug-in mobile hard disk, a smart media card (SMC), and a secure digital (Secure Digital) equipped on the electronic device 1. , SD) card, flash card (Flash Card), etc.
  • the memory 11 may also include both an internal storage unit of the electronic device 1 and an external storage device.
  • the memory 11 can be used not only to store application software and various data installed in the electronic device 1, such as the code of an automated review program for injury pictures, but also to temporarily store data that has been output or will be output.
  • the processor 10 may be composed of integrated circuits in some embodiments, for example, may be composed of a single packaged integrated circuit, or may be composed of multiple integrated circuits with the same function or different functions, including one or more Combinations of central processing unit (CPU), microprocessor, digital processing chip, graphics processor, and various control chips, etc.
  • the processor 10 is the control unit of the electronic device, which uses various interfaces and lines to connect various components of the entire electronic device, and runs or executes programs or modules (such as injury The automatic review program for love pictures, etc.), and call the data stored in the memory 11 to execute various functions of the electronic device 1 and process data.
  • the bus may be a peripheral component interconnect standard (PCI) bus or an extended industry standard architecture (EISA) bus, etc.
  • PCI peripheral component interconnect standard
  • EISA extended industry standard architecture
  • the bus can be divided into address bus, data bus, control bus and so on.
  • the bus is configured to implement connection and communication between the memory 11 and at least one processor 10 and the like.
  • FIG. 3 only shows an electronic device with components. Those skilled in the art can understand that the structure shown in FIG. 3 does not constitute a limitation on the electronic device 1, and may include fewer or more components than shown in the figure. Components, or a combination of certain components, or different component arrangements.
  • the electronic device 1 may also include a power source (such as a battery) for supplying power to various components.
  • the power source may be logically connected to the at least one processor 10 through a power management device, thereby controlling the power supply.
  • the device implements functions such as charge management, discharge management, and power consumption management.
  • the power supply may also include any components such as one or more DC or AC power supplies, recharging devices, power failure detection circuits, power converters or inverters, and power status indicators.
  • the electronic device 1 may also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.
  • the electronic device 1 may also include a network interface.
  • the network interface may include a wired interface and/or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is usually used in the electronic device 1 Establish a communication connection with other electronic devices.
  • the electronic device 1 may also include a user interface.
  • the user interface may be a display (Display) and an input unit (such as a keyboard (Keyboard)).
  • the user interface may also be a standard wired interface or a wireless interface.
  • the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode, organic light-emitting diode) touch device, etc.
  • the display can also be appropriately called a display screen or a display unit, which is used to display the information processed in the electronic device 1 and to display a visualized user interface.
  • the injury picture automatic review and interrogation program 12 stored in the memory 11 in the electronic device 1 is a combination of multiple instructions. When running in the processor 10, it can realize:
  • Reviewing the skin area of the injury picture includes: detecting the skin area in the injury picture based on the color space threshold segmentation; judging whether the area of the detected skin area meets the area size requirement; if the area of the detected skin area is not Meet the area size requirements, send the first unqualified instruction to the client to re-obtain the injury picture to be reviewed.
  • the first unqualified instruction includes the abnormal light instruction; if the detected area of the skin area meets the area size requirements, The detected skin area is separated from the wound area;
  • the unqualified instruction includes the long shooting distance; if the separated wound area meets the wound area requirements, locate the injury picture based on the template matching method 'S face
  • Reviewing the face of the injury picture includes: judging whether the positioned face meets the face requirements; if the positioned face does not meet the face requirements, send the third unqualified instruction to the client, and obtain the desired review again
  • the third unqualified instruction includes adjusting the shooting angle or adjusting the shooting distance; if the positioned face meets the face requirements, obtain the injured person’s certificate, perform face detection on the injured person’s certificate, and obtain the information in the injured person’s certificate human face;
  • Review the similarity between the face in the injury picture and the face in the injured person’s certificate including: comparing the similarity between the face in the certificate and the face in the injury picture; judging the comparison result of the similarity comparison Whether it meets the similarity requirements; if the comparison result does not meet the similarity requirements, send the fourth unqualified instruction to the client to re-obtain the injury picture to be reviewed.
  • the fourth unqualified instruction includes the wrong document; if the comparison result Meet the similarity requirements, upload the injury picture.
  • the integrated module/unit of the electronic device 1 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.
  • the computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory) .
  • the embodiments of the present application also propose a computer-readable storage medium.
  • the computer-readable storage medium may be non-volatile or volatile.
  • the computer-readable storage medium includes a computer program, and the computer program is The following operations are implemented when the processor is executed:
  • Reviewing the skin area of the injury picture includes: detecting the skin area in the injury picture based on the color space threshold segmentation; judging whether the area of the detected skin area meets the area size requirement; if the area of the detected skin area is not Meet the area size requirements, send the first unqualified instruction to the client to re-obtain the injury picture to be reviewed.
  • the first unqualified instruction includes the abnormal light instruction; if the detected area of the skin area meets the area size requirements, The detected skin area is separated from the wound area;
  • the unqualified instruction includes the long shooting distance; if the separated wound area meets the wound area requirements, locate the injury picture based on the template matching method 'S face
  • Reviewing the face of the injury picture includes: judging whether the positioned face meets the face requirements; if the positioned face does not meet the face requirements, send the third unqualified instruction to the client, and obtain the desired review again
  • the third unqualified instruction includes adjusting the shooting angle or adjusting the shooting distance; if the positioned face meets the face requirements, obtain the injured person’s certificate, perform face detection on the injured person’s certificate, and obtain the information in the injured person’s certificate human face;
  • Review the similarity between the face in the injury picture and the face in the injured person’s certificate including: comparing the similarity between the face in the certificate and the face in the injury picture; judging the comparison result of the similarity comparison Whether it meets the similarity requirements; if the comparison result does not meet the similarity requirements, send the fourth unqualified instruction to the client to re-obtain the injury picture to be reviewed.
  • the fourth unqualified instruction includes the wrong document; if the comparison result Meet the similarity requirements, upload the injury picture.
  • the automated review, device, electronic equipment, and computer readable storage medium of injury pictures for auto insurance claims described in this application are based on a variety of lightweight digital image processing methods and machine learning methods, which can construct an offline injury picture review program, Fully automated review of love pictures saves labor costs and transportation costs for injury claims; real-time guides customers to adjust the ambient light, distance, and shooting angle when taking pictures of injuries, so as to increase the probability of success in taking a picture of injuries that meet the requirements, thereby increasing Customer experience: The amount of calculation is small, and it can be deployed on the customer's mobile phone, reducing the cost of deploying remote servers; only a very small amount of customer injury pictures are used in the algorithm development, to protect user privacy as much as possible, and to meet the compliance requirements for data use.
  • modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the objectives of the solutions of the embodiments.
  • the functional modules in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit.
  • the above-mentioned integrated unit may be implemented in the form of hardware, or may be implemented in the form of hardware plus software functional modules.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Business, Economics & Management (AREA)
  • Data Mining & Analysis (AREA)
  • Finance (AREA)
  • Accounting & Taxation (AREA)
  • Evolutionary Biology (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Human Computer Interaction (AREA)
  • Multimedia (AREA)
  • Geometry (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Development Economics (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Strategic Management (AREA)
  • Technology Law (AREA)
  • General Business, Economics & Management (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

一种伤情图片自动化审核方法、装置、电子设备及计算机可读存储介质，涉及智慧医疗，所述方法包括：获得拟审核的伤情图片；审核所述伤情图片的皮肤区域；审核所述伤情图片的伤口区域；审核所述伤情图片的人脸；审核伤情图片中的人脸和伤者证件中的人脸的相似度；上传符合审核要求的伤情图片。所述方法可提高一次拍摄符合要求的伤情图片成功概率，提升客户体验。

Description

伤情图片自动化审核方法、装置、电子设备及存储介质
本申请要求于2020年8月27日提交中国专利局、申请号为202010879301.1，发明名称为“伤情图片自动化审核方法、装置、电子设备及存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及智慧医疗，尤其涉及一种面向车险理赔的伤情图片自动化审核方法、装置、电子设备及计算机可读存储介质。
背景技术
在智慧医疗中，人伤理赔是车辆保险中重要的理赔服务。进行该项理赔时，展示伤情图片(人身受伤情况的图片)是用于理赔的重要材料。
传统获取伤情图片的方法是派遣车险人伤理赔员赶往现场或医院拍摄伤情图片。这种方案虽然能够可靠获取符合理赔要求的伤情图片，但需要投入较高的人力成本和交通成本。
现有技术中，车险客户可以通过移动端软件提交伤情图片，交由远程的理赔员人工审核的方案。这种方案虽然节省了派遣理赔员的交通成本，但由于客户(特别是年级较大、受教育程度不高的客户)拍摄的伤情图片不一定满足理赔要求，可能需要多次与客户沟通并请求重新拍摄，一定程度上影响了理赔速度和服务质量。另外，投入人力成本并未显著减少。
当前流行的深度学习方法，可以较好地构建伤情图片自动化审核流程，并取得较好的实用化效果。但存在其他突出问题。一是训练深度学习模型需要大量伤情图片，伤情图片明显属于客户的隐私信息，因而使用客户伤情图片进行模型训练存在较高的信息违规风险。也就是说，发明人发现，基于深度学习的方案虽然效果理想，但很可能因为缺乏合规数据而不可实现；二是是深度学习模型计算复杂，对硬件资源要求较高，目前主要通过将功能部署在远程高性能服务器实现，而直接部署在客户移动端存在困难，这导致深度学习方案的运营成本较高，仍然不能很好地达到减少理赔成本的根本目的。
发明内容
本申请提供一种面向车险理赔的伤情图片自动化审核、装置、电子设备及计算机可读存储介质，其主要目的在于提高一次拍摄符合要求的伤情图片的成功概率，提升客户体验。
为实现上述目的，本申请提供的一种伤情图片自动化审核方法，所述方法包括：
获得拟审核的伤情图片；
审核所述伤情图片的皮肤区域，包括：基于色彩空间阈值分割对伤情图片中皮肤区域进行检测；判断检测到的皮肤区域的面积是否符合面积大小要求；如果检测到的皮肤区域的面积不符合面积大小要求，发送第一不合格指令给客户端，重新获得拟审核的伤情图片，所述第一不合格指令包括光照异常指令；如果检测到的皮肤区域的面积符合面积大小要求，从检测到的皮肤区域分离出伤口区域；
审核所述伤情图片的伤口区域，包括：基于色彩空间阈值分割对伤口区域进行检测；判断分离出的伤口面积是否符合伤口面积要求；如果分离出的伤口面积不符合伤口面积要求，发送第二不合格指令给客户端，重新获得拟审核的伤情图片，所述第二不合格指令包括拍摄距离远；如果分离出的伤口面积符合伤口面积要求，基于模板匹配的方法定位伤情图片中出现的人脸；
审核所述伤情图片的人脸，包括：判断定位的人脸是否符合人脸要求；如果定位的人脸不符合人脸要求，则发送第三不合格指令给客户端，重新获得拟审核的伤情图片，所述第三不合格指令包括调整拍摄角度或调整拍摄距离；如果定位的人脸符合人脸要求，获得伤者证件，对伤者证件进行人脸检测，获得伤者证件中的人脸；
审核伤情图片中的人脸和伤者证件中的人脸的相似度，包括：将证件中的人脸和伤情图片中的人脸进行相似度对比；判断所述相似度对比的对比结果是否符合相似度要求；如果对比结果不符合相似度要求，发送第四不合格指令给客户端，重新获得拟审核的伤情图片，所述第四不合格指令包括传错单证；如果对比结果符合相似度要求，上传伤情图片。
为了解决上述问题，本申请还提供一种伤情图片自动化审核装置，所述装置包括：
获得部，获得拟审核的伤情图片；
第一审核部，审核所述伤情图片的皮肤区域，包括：皮肤区域检测模块，基于色彩空间阈值分割对伤情图片中皮肤区域进行检测；第一面积判断模块，判断检测到的皮肤区域的面积是否符合面积大小要求，如果检测到的皮肤区域的面积不符合面积大小要求，发送第一不合格指令给客户端，获得部重新获得拟审核的伤情图片，所述第一不合格指令包括光照异常指令；如果检测到的皮肤区域的面积符合面积大小要求，发送信号给伤口区域获得模块；伤口区域获得模块，从检测到的皮肤区域分离出伤口区域并发送给第二审核部；
第二审核部，审核所述伤情图片的伤口区域，包括：伤口区域检测模块，基于色彩空间阈值分割对第一审核部获得的皮肤区域中的伤口区域进行检测；第二面积判断模块，判断分离出的伤口面积是否符合伤口面积要求，如果分离出的伤口面积不符合伤口面积要求，发送第二不合格指令给客户端，获得部重新获得拟审核的伤情图片，所述第二不合格指令包括拍摄距离远；如果分离出的伤口面积符合伤口面积要求，发送信号给第一人脸获得模块；第一人脸获得模块，基于模板匹配的方法定位伤情图片中出现的人脸并发送给第三审核部和相似度获得部；
第三审核部，审核所述伤情图片的人脸，包括：人脸判断模块，判断定位的人脸是否符合人脸要求，如果定位的人脸不符合人脸要求，则发送第三不合格指令给客户端，获得部重新获得拟审核的伤情图片，所述第三不合格指令包括调整拍摄角度或调整拍摄距离；如果定位的人脸符合人脸要求，发送信号给第二人脸获得模块；第二人脸获得模块，获得伤者证件，对伤者证件进行人脸检测，获得伤者证件中的人脸并发送给相似度获得部；
相似度获得部，将证件中的人脸和伤情图片中的人脸进行相似度对比，判断所述相似度对比的对比结果是否符合相似度要求，如果对比结果不符合相似度要求，发送第四不合格指令给客户端，获得部重新获得拟审核的伤情图片，所述第四不合格指令包括传错单证；如果对比结果符合相似度要求，发送信号给上传部；
上传部，上传符合第一审核部、第二审核部、第三审核部和相似度获得部要求的伤情图片。
为了解决上述问题，本申请还提供一种电子设备，所述电子设备包括：
至少一个处理器；以及，
与所述至少一个处理器通信连接的存储器；其中，
所述存储器存储有可被所述至少一个处理器执行的指令，所述指令被所述至少一个处理器执行，以使所述至少一个处理器能够执行如下步骤：
获得拟审核的伤情图片；
审核所述伤情图片的皮肤区域，包括：基于色彩空间阈值分割对伤情图片中皮肤区域进行检测；判断检测到的皮肤区域的面积是否符合面积大小要求；如果检测到的皮肤区域的面积不符合面积大小要求，发送第一不合格指令给客户端，重新获得拟审核的伤情图片，所述第一不合格指令包括光照异常指令；如果检测到的皮肤区域的面积符合面积大小要求，从检测到的皮肤区域分离出伤口区域；
审核所述伤情图片的伤口区域，包括：基于色彩空间阈值分割对伤口区域进行检测；判断分离出的伤口面积是否符合伤口面积要求；如果分离出的伤口面积不符合伤口面积要求，发送第二不合格指令给客户端，重新获得拟审核的伤情图片，所述第二不合格指令包括拍摄距离远；如果分离出的伤口面积符合伤口面积要求，基于模板匹配的方法定位伤情 图片中出现的人脸；
审核所述伤情图片的人脸，包括：判断定位的人脸是否符合人脸要求；如果定位的人脸不符合人脸要求，则发送第三不合格指令给客户端，重新获得拟审核的伤情图片，所述第三不合格指令包括调整拍摄角度或调整拍摄距离；如果定位的人脸符合人脸要求，获得伤者证件，对伤者证件进行人脸检测，获得伤者证件中的人脸；
审核伤情图片中的人脸和伤者证件中的人脸的相似度，包括：将证件中的人脸和伤情图片中的人脸进行相似度对比；判断所述相似度对比的对比结果是否符合相似度要求；如果对比结果不符合相似度要求，发送第四不合格指令给客户端，重新获得拟审核的伤情图片，所述第四不合格指令包括传错单证；如果对比结果符合相似度要求，上传伤情图片。
为了解决上述问题，本申请还提供一种计算机可读存储介质，存储有计算机程序，其中，所述计算机程序被处理器执行时实现如下步骤：
获得拟审核的伤情图片；
审核所述伤情图片的皮肤区域，包括：基于色彩空间阈值分割对伤情图片中皮肤区域进行检测；判断检测到的皮肤区域的面积是否符合面积大小要求；如果检测到的皮肤区域的面积不符合面积大小要求，发送第一不合格指令给客户端，重新获得拟审核的伤情图片，所述第一不合格指令包括光照异常指令；如果检测到的皮肤区域的面积符合面积大小要求，从检测到的皮肤区域分离出伤口区域；
审核所述伤情图片的伤口区域，包括：基于色彩空间阈值分割对伤口区域进行检测；判断分离出的伤口面积是否符合伤口面积要求；如果分离出的伤口面积不符合伤口面积要求，发送第二不合格指令给客户端，重新获得拟审核的伤情图片，所述第二不合格指令包括拍摄距离远；如果分离出的伤口面积符合伤口面积要求，基于模板匹配的方法定位伤情图片中出现的人脸；
审核所述伤情图片的人脸，包括：判断定位的人脸是否符合人脸要求；如果定位的人脸不符合人脸要求，则发送第三不合格指令给客户端，重新获得拟审核的伤情图片，所述第三不合格指令包括调整拍摄角度或调整拍摄距离；如果定位的人脸符合人脸要求，获得伤者证件，对伤者证件进行人脸检测，获得伤者证件中的人脸；
审核伤情图片中的人脸和伤者证件中的人脸的相似度，包括：将证件中的人脸和伤情图片中的人脸进行相似度对比；判断所述相似度对比的对比结果是否符合相似度要求；如果对比结果不符合相似度要求，发送第四不合格指令给客户端，重新获得拟审核的伤情图片，所述第四不合格指令包括传错单证；如果对比结果符合相似度要求，上传伤情图片。
本申请所述伤情图片自动化审核方法、装置、电子设备及计算机可读存储介质实时引导客户在拍摄伤情图片时调整环境光线、距离、拍摄角度，提高一次拍摄符合要求的伤情图片成功概率，从而提升客户体验，采用的是基于色彩空间阈值分割和模板匹配多种轻量级数字图像处理方法，计算量小，可部署于客户手机，减少部署远程服务器的成本，在用户未按下快门之前就实时提醒用户进行相应调整，从而在保证拍摄出有效的伤情图片的同时提升客户的体验，更为重要的是可以离线自动审核。
附图说明
图1是本申请所述伤情图片自动化审核方法的流程图；
图2是本申请一实施例提供的伤情图片自动化审核装置的模块示意图；
图3是本申请一实施例提供的实现伤情图片自动化审核方法的电子设备的内部结构示意图。
本申请目的的实现、功能特点及优点将结合实施例，参照附图做进一步说明。
具体实施方式
应当理解，此处所描述的具体实施例仅仅用以解释本申请，并不用于限定本申请。
图1是本申请所述伤情图片自动化审核方法的流程图，如图1所示，所述伤情图片自 动化审核方法包括：
步骤S1，获得拟审核的伤情图片；
步骤S2，审核所述伤情图片的皮肤区域，包括：基于色彩空间阈值分割对伤情图片中皮肤区域进行检测；判断检测到的皮肤区域的面积是否符合面积大小要求；如果检测到的皮肤区域的面积不符合面积大小要求，发送第一不合格指令给客户端，返回步骤S1，重新获得拟审核的伤情图片，所述第一不合格指令包括光照异常指令；如果检测到的皮肤区域的面积符合面积大小要求，从检测到的皮肤区域分离出伤口区域，执行步骤S3；
步骤S3，审核所述伤情图片的伤口区域，包括：基于色彩空间阈值分割对伤口区域进行检测；判断分离出的伤口面积是否符合伤口面积要求；如果分离出的伤口面积不符合伤口面积要求，发送第二不合格指令给客户端，返回步骤S1，重新获得拟审核的伤情图片，所述第二不合格指令包括拍摄距离远；如果分离出的伤口面积符合伤口面积要求，基于模板匹配的方法定位伤情图片中出现的人脸，执行步骤S4；
步骤S4，审核所述伤情图片的人脸，包括：判断定位的人脸是否符合人脸要求；如果定位的人脸不符合人脸要求，则发送第三不合格指令给客户端，返回步骤S1，重新获得拟审核的伤情图片，所述第三不合格指令包括调整拍摄角度或调整拍摄距离；如果定位的人脸符合人脸要求，获得伤者证件，对伤者证件进行人脸检测，获得伤者证件中的人脸，执行步骤S5；
步骤S5，审核伤情图片中的人脸和伤者证件中的人脸的相似度，包括：将证件中的人脸和伤情图片中的人脸进行相似度对比；判断所述相似度对比的对比结果是否符合相似度要求；如果对比结果不符合相似度要求，发送第四不合格指令给客户端，返回步骤S1，重新获得拟审核的伤情图片，所述第四不合格指令包括传错单证；如果对比结果符合相似度要求，执行步骤S6；
步骤S6，上传符合步骤S2-S5要求的伤情图片。
在一个实施例中，步骤S2中，所述审核所述伤情图片的皮肤区域的步骤包括：
基于色彩空间分析伤情图片的光照；
设定色彩空间中各参数的光照阈值，获得第一判别条件；
判断伤情图片的光照是否满足第一判别条件；
如果伤情图片的光照不满足第一判别条件，发出第一不合格指令，所述第一不合格指令包括光照过暗、光照过曝或不均匀光照；
如果伤情图片的光照满足第一判别条件，基于色彩空间分析伤情图片的光照的均匀性；
设定光照均匀度阈值，获得第二判别条件；
判断伤情图片的光照的均匀性是否满足第二判别条件；
如果伤情图片的光照的均匀性不满足第二判别条件，发出第一不合格指令，所述第一不合格指令包括局部过暗或局部过曝；
如果伤情图片的光照的均匀性满足第二判别条件，基于色彩空间阈值分割伤情图片，获得伤情图片的皮肤区域，获得皮肤区域的面积；
设定皮肤区域面积的阈值，获得第四判别条件；
判断伤情图片的皮肤区域的面积是否满足第四判别条件；
如果伤情图片的皮肤区域的面积不满足第四判别条件，发出第一不合格指令，所述第一不合格指令包括皮肤面积过小和拍摄距离过远；
如果伤情图片的皮肤区域的面积满足第四判别条件，从检测到的皮肤区域分离出伤口区域。
优选地，所述基于色彩空间阈值分割伤情图片，获得伤情图片的皮肤区域的步骤包括：
获得伤情图片的多个色彩空间的数据；
设定皮肤区域对应的多个色彩空间中各参数的阈值，获得第三判别条件；
从伤情图片的多个色彩空间的数据中筛选出满足第三判别条件的多个像素点；
通过形态学闭运算填充所述像素点围成区域的内部空隙，得到一个或多个闭合区域，从而获得皮肤区域。
在一个优选实施例中，步骤S2包括：
步骤S21，将伤情图片从RGB色彩空间转换到HSV色彩空间，获得伤情图片的R通道、G通道、B通道和V通道的均值
Figure PCTCN2020125071-appb-000001
整体环境光照正常的第一判别条件为：
Figure PCTCN2020125071-appb-000002
其中，η R、η G、η B分别为R通道，G通道、B通道的均值上限，如果某通道均值高于上限，说明光照色温存在严重的偏差；η V1、η V2为V通道的均值的下限和上限，低于η V1说明光照过暗，高于η V2说明光照过亮；当满足第一判别条件时，执行步骤S22，当不满足第一判别条件时，根据R通道、G通道、B通道和V通道的均值
Figure PCTCN2020125071-appb-000003
发出不同的第一不合格指令，所述第一不合格指令包括光照过暗、光照过曝或不均匀光照。
步骤S22，环境光照均匀性快速分析，将伤情图片的整图均匀切分为n*n的子块，检查所有子块的V通道均值V ij(i，j＝1，2，3，…，n)是否满足第二判别条件，当满足第二判别条件，执行步骤S23，当不满足第二判别条件，根据所有子块的V通道均值发出不同的第一不合格指令，所述第一不合格指令包括局部过暗或局部过曝，
Figure PCTCN2020125071-appb-000004
其中，η VL是子块与整图相对亮度均值的下限，低于此值说明伤情图片存在局部过暗的问题；η VH是子块与整图相对亮度均值的上限，高于此值说明伤情图片存在局部过曝的问题。
步骤S23，皮肤区域检测，将伤情图片从RGB色彩空间转换至YC rC b、HSV等色彩空间，对于宽高为w*h的伤情图像，将每个像素点用三种色彩空间表示为P ij＝(R，G，B，Y，C r，C b，H，S，V)，筛选出P ij满足第三判别条件的像素点，通过形态学闭运算填充所述像素点围成区域的内部空隙，得到一个或多个闭合区域，从而获得皮肤区域，判断所述皮肤区域的像素点是否符合第四判别条件，如果不满足第四判别条件，根据发出第一不合格指令，所述第一不合格指令包括皮肤面积过小和拍摄距离过远，如果满足第四判别条件，执行步骤S3，其中，第三判别条件为：
Figure PCTCN2020125071-appb-000005
其中，第四判别条件为：
Figure PCTCN2020125071-appb-000006
其中，P ij属于皮肤区域内的一个点，各不等式两边的阈值是针对皮肤设定的，n为满足第三判别条件的像素点的个数，若
Figure PCTCN2020125071-appb-000007
表明皮肤面积过小。
在一个实施例中，步骤S3中，所述审核所述伤情图片的伤口区域的步骤包括：
基于色彩空间阈值分割皮肤区域，获得皮肤区域中的伤口区域，获得伤口区域的面积；
设定伤口区域面积的阈值，获得第六判别条件；
判断伤口区域的面积是否满足第六判别条件；
如果伤口区域的面积不满足第六判别条件，发出第二不合格指令，所述第二不合格指令包括伤口面积过小和拍摄距离过远；
如果伤口区域的面积满足第六判别条件，基于模板匹配的方法定位伤情图片中出现的人脸。
优选地，所述基于色彩空间阈值分割皮肤区域，获得皮肤区域中的伤口区域的步骤包括：
获得皮肤区域的多个色彩空间的数据；
设定伤口区域对应的多个色彩空间中各参数的阈值，获得第五判别条件；
从皮肤区域的多个色彩空间的数据中筛选出满足判别条件第五判别条件的多个像素点；
通过形态学闭运算填充所述像素点围成区域的内部空隙，得到一个或多个闭合区域，从而获得伤口区域。
优选地，所述基于模板匹配的方法定位伤情图片中出现的人脸的步骤包括：
将伤情图片的RGB图像转换为灰度图像；
以目前公开的人脸数据集制作平均脸并作为模板，采用模板匹配算法对所述灰度图像进行扫描检测，检测出人脸。
在一个优选实施例中，步骤S3包括：
步骤S31，将皮肤区域的图像从RGB色彩空间转换至YC rC b、HSV等色彩空间。对于宽高为w*h的伤情图像，将皮肤区域的每个像素点用三种色彩空间表示为p′ ij＝(R′,G′,B′,Y′,C′ r,C′ b,H′,S′,V′)，筛选出p′ ij满足第五判别条件的像素点，通过形态学闭运算填充所述像素点围成区域的内部空隙，得到一个或多个闭合区域，从而获得伤口区域，判断所述伤口区域的像素点是否符合第六判别条件，如果不满足第六判别条件，根据发出第二不合格指令，所述第二不合格指令包括伤口面积过小和拍摄距离过远，如果满足第六判别条件，执行步骤S4，其中，第五判别条件为：
Figure PCTCN2020125071-appb-000008
其中，第六判别条件为：
Figure PCTCN2020125071-appb-000009
其中，p′ ij属于伤口区域内的一个点，各不等式两边的阈值是针对伤口设定的，n′为满足第五判别条件的像素点的个数，若
Figure PCTCN2020125071-appb-000010
表明伤口面积过小。
步骤S32，将伤情图片的RGB图像转换为灰度图像，对于尺寸为w*h的图像，将每一 个RGB像素按下述公式转换为灰度像素(i＝1，2，3…w，j＝1，2，3…，h)：
Gray ij＝R＇ ij*0.299+G＇ ij*0.587+B＇ ij*0.114。
步骤S33，以目前公开的人脸数据集制作“平均脸”并作为模板，采用模板匹配算法对所述灰度图像进行扫描检测，检测出人脸。
在一个实施例中，步骤S4中，所述审核所述伤情图片的人脸的步骤包括：
采用非极大值抑制算法(Non Maximum Suppression)消除同一位置检测重叠的、多余的人脸；
若检测无人脸，发出第三不合格指令，所述第三不合格指令包括伤者调整拍摄距离或调整人脸拍摄角度；
若检测到2张及以上人脸，发出第四不合格指令，所述第四不合格指令包括伤者以外的人员暂离拍摄画面；
若检测到1张人脸，获得伤者证件，对伤者证件进行人脸检测，获得伤者证件中的人脸。
优选地，所述伤者证件为身份证、驾驶证等格式固定(有固定的模板)的证件，进一步，优选地，从身份证截取第二人脸子块，由于身份证版式固定而提取其中头像较为简单，可替换为其他更为简便的方法，例如，因为身份证的头像位置固定，根据相对位置截取头像。
在一个优选实施例中，所述获得伤者证件中的人脸的步骤包括：
将伤者证件的RGB图像转二值化图像；
使用均值滤波器消除部分噪声；
使用图像学的“闭运算”将伤者证件图像的文字和头像填充为方块；
使用伤者证件的模板，以模板匹配算法定位到伤者证件在图像中的位置；
因为头像在伤者证件的相对位置是固定的，根据相对位置截取头像。
在一个实施例中，步骤S5中，所述审核伤情图片中的人脸和伤者证件中的人脸的相似度的步骤包括：
人脸提取，根据检测到的1张人脸相对整图的位置，截取其中的1张人脸作为第一人脸子块；从身份证截取第二人脸子块；
将第一人脸子块和第二人脸子块统一缩放，并分别提取LBP特征(局部二值模式，Local Binary Patterns)；
采用相似度方法根据提取的第一人脸子块和第二人脸子块的LBP特征进行人脸特征匹配获得第一人脸子块和第二人脸子块的相似度；
判断所述相似度是否符合相似度要求；
如果所述相似度不符合相似度要求，发送第四不合格指令给客户端，重新获得拟审核的伤情图片，所述第四不合格指令包括传错单证(例如客户端误上传自己的身份证而不是伤者的身份证，提示客户核查证件是否属于伤者)；
如果所述相似度符合相似度要求，上传伤情图片。
优选地，人脸不同位置的特征区分程度不一样，例如眼部、鼻部位置的特征重要性大于人脸边缘靠近背景位置的特征，根据特征重要性对人脸的不同位置赋予不同的权重，特征重要性越大，权值越大，进一步，优选地，设计固定的权重调节向量W，区分人脸不同位置的权重。
优选地，基于余弦相似度根据提取的第一人脸子块和第二人脸子块的LBP特征进行人脸特征匹配获得第一人脸子块和第二人脸子块的相似度
Figure PCTCN2020125071-appb-000011
其中，F source为从伤情图像中提取的人脸LBP特征向量，F target为从身份证图像中提 取的人脸LBP特征向量，⊙是向量同位置元素相乘操作，‖‖是求向量模长操作，S为第一人脸子块和第二人脸子块的相似度，0≤S≤1，S越大，表明两LBP特征越相近，可用作对人脸相似度的表征。
若S≥S min，即S高于阈值S min，则认为人脸匹配，执行步骤S6；否则发出第四不合格指令。
在本申请所述伤情图片自动化审核方法中，皮肤区域检测(从伤情图片中分离出皮肤区域)。此步骤目的是减少伤口区域检测的误检测概率(伤口只存在于皮肤上)。伤者的皮肤检测不通过是因为光照异常(过暗，过曝，或不均匀光照导致检测出来的皮肤面积过小)，相应地，提示客户调整拍摄环境光线情况；伤口区域检测：从检出的皮肤区域分离出伤口区域，此步骤目的是核实图片中存在受伤的部位，在前述步骤通过的前提下，伤者的皮肤检测不通过是因为镜头与伤者的距离过远，导致检测出来的伤口面积过小，相应地，提示客户调整拍摄距离；人脸检测：定位图像中出现的人脸，在前述步骤通过的前提下，伤者的人脸检测不通过是因为镜头没有拍摄到人脸或人脸未朝向镜头，相应地，提示客户调整拍摄距离或调整拍摄角度，可行的技术方案包括基于模板匹配的人脸检测方法，由于前两个步骤为简化本步骤算法创造了条件，可使用模板匹配算法实现快速人脸检测；人脸相似度比对：将证件(一般为身份证)图片中的人脸和伤情图片中的人脸进行相似度比对，此步骤的目的是核实伤者的身份，抵御骗保等欺诈情况。
本申请所述伤情图片自动化审核方法步骤S2-S5总体计算量较小，所以对设备性能要求不高，移动端CPU可在数秒以内完成处理所需要的计算，满足“离线型”的设计要求,在实现伤情图片全自动化审核的同时，减少高性能服务器的部署成本，进一步达到减少伤情图片审核的根本目的。
另外，对于发布时间较新、性能较强的移动端，本申请所述伤情图片自动化审核方法可实现为在用户未按下快门之前就实时提醒用户进行相应调整，从而在保证拍摄出有效的伤情图片的同时提升客户的体验。
此外，本申请所述伤情图片自动化审核方法可不直接使用伤情图片，例如使用已公开论文推荐的参数完成皮肤检测模型的设计，使用已公开的人脸检测数据集完成人脸检测模型的设计等，从而规避隐私数据的使用问题。
图2是本申请所述伤情图片自动化审核装置装置的构成框图，如图2所示，所述伤情图片自动化审核装置装置100可以安装于电子设备中。根据实现的功能，所述数据稽核装置可以包括获得部110、第一审核部120、第二审核部130、第三审核部140、相似度获得部150和上传部160，第一审核部120包括皮肤区域检测模块121、第一面积判断模块122和伤口区域获得模块123，第二审核部130包括伤口区域检测模块131、第二面积判断模块132和第一人脸获得模块133，第三审核部140包括人脸判断模块141和第二人脸获得模块142。本发所述部/模块是指一种能够被电子设备处理器所执行，并且能够完成固定功能的一系列计算机程序段，其存储在电子设备的存储器中。
在本实施例中，关于各部/模块的功能如下：
获得部110，获得拟审核的伤情图片；
第一审核部120，审核所述伤情图片的皮肤区域，包括：皮肤区域检测模块121，基于色彩空间阈值分割对伤情图片中皮肤区域进行检测；第一面积判断模块122，判断检测到的皮肤区域的面积是否符合面积大小要求，如果检测到的皮肤区域的面积不符合面积大小要求，发送第一不合格指令给客户端，获得部110重新获得拟审核的伤情图片，所述第一不合格指令包括光照异常指令；如果检测到的皮肤区域的面积符合面积大小要求，发送信号给伤口区域获得模块123；伤口区域获得模块123，从检测到的皮肤区域分离出伤口区域并发送给第二审核部130；
第二审核部130，审核所述伤情图片的伤口区域，包括：伤口区域检测模块131，基 于色彩空间阈值分割对第一审核部120获得的皮肤区域中的伤口区域进行检测；第二面积判断模块132，判断分离出的伤口面积是否符合伤口面积要求，如果分离出的伤口面积不符合伤口面积要求，发送第二不合格指令给客户端，获得部110重新获得拟审核的伤情图片，所述第二不合格指令包括拍摄距离远；如果分离出的伤口面积符合伤口面积要求，发送信号给第一人脸获得模块133；第一人脸获得模块133，基于模板匹配的方法定位伤情图片中出现的人脸并发送给第三审核部140和相似度获得部150；
第三审核部140，审核所述伤情图片的人脸，包括：人脸判断模块141，判断定位的人脸是否符合人脸要求，如果定位的人脸不符合人脸要求，则发送第三不合格指令给客户端，获得部110重新获得拟审核的伤情图片，所述第三不合格指令包括调整拍摄角度或调整拍摄距离；如果定位的人脸符合人脸要求，发送信号给第二人脸获得模块142；第二人脸获得模块142，获得伤者证件，对伤者证件进行人脸检测，获得伤者证件中的人脸并发送给相似度获得部150；
相似度获得部150，将证件中的人脸和伤情图片中的人脸进行相似度对比，判断所述相似度对比的对比结果是否符合相似度要求，如果对比结果不符合相似度要求，发送第四不合格指令给客户端，获得部110重新获得拟审核的伤情图片，所述第四不合格指令包括传错单证；如果对比结果符合相似度要求，发送信号给上传部160；
上传部160，上传符合第一审核部120、第二审核部130、第三审核部140和相似度获得部150要求的伤情图片。
在一个实施例中，上述皮肤区域检测模块121包括：
光照获得单元，基于色彩空间分析伤情图片的光照；
第一判别条件设定单元，设定色彩空间中各参数的光照阈值，获得第一判别条件；
光照判别单元，判断伤情图片的光照是否满足第一判别条件；如果伤情图片的光照不满足第一判别条件，发出第一不合格指令，所述第一不合格指令包括光照过暗、光照过曝或不均匀光照；如果伤情图片的光照满足第一判别条件，发送信号给均匀度获得单元；
均匀度获得单元，基于色彩空间分析伤情图片的光照的均匀性；
第二判别条件设定单元，设定光照均匀度阈值，获得第二判别条件；
均匀度判别单元，判断伤情图片的光照的均匀性是否满足第二判别条件；如果伤情图片的光照的均匀性不满足第二判别条件，发出第一不合格指令，所述第一不合格指令包括局部过暗或局部过曝；如果伤情图片的光照的均匀性满足第二判别条件，发送信号给皮肤区域获得单元；
皮肤区域获得单元，基于色彩空间阈值分割伤情图片，获得伤情图片的皮肤区域及皮肤区域的面积。
优选地，皮肤区域获得单元包括：
色彩空间转换子单元，获得伤情图片的多个色彩空间的数据；
第三判别条件获得子单元，设定皮肤区域对应的多个色彩空间中各参数的阈值，获得第三判别条件；
筛选子单元，从伤情图片的多个色彩空间的数据中筛选出满足第三判别条件的多个像素点；
填充单元，通过形态学闭运算填充所述像素点围成区域的内部空隙，得到一个或多个闭合区域，从而获得皮肤区域。
伤口区域检测模块131的实现方式和上述皮肤区域检测模块121的实现方式类似，由于皮肤区域检测模块121已经对伤情图片的光照和均匀度进行了检测，因此伤口区域检测模块131仅仅需要检测伤口区域的面积。
在一个实施例中，上述第一人脸获得模块133包括：
灰度转化单元，将伤情图片的RGB图像转换为灰度图像；
模板匹配单元，以目前公开的人脸数据集制作平均脸并作为模板，采用模板匹配算法对所述灰度图像进行扫描检测，检测出人脸。
在一个实施例中，上述人脸判断模块141包括：
消除单元，采用非极大值抑制算法消除同一位置检测重叠的、多余的人脸；
人脸计数单元，对消除单元处理后的伤情图片中的人脸进行计数，若检测无人脸，发出第三不合格指令，所述第三不合格指令包括伤者调整拍摄距离或调整人脸拍摄角度；若检测到2张及以上人脸，发出第四不合格指令，所述第四不合格指令包括伤者以外的人员暂离拍摄画面；若检测到1张人脸，发送信号给第二人脸获得模块142。
在一个实施例中，相似度获得部150包括：
人脸提取单元，根据检测到的1张人脸相对整图的位置，截取其中的1张人脸作为第一人脸子块；从身份证截取第二人脸子块；
特征提取单元，将第一人脸子块和第二人脸子块统一缩放，并分别提取LBP特征；
相似度匹配单元，采用相似度方法根据提取的第一人脸子块和第二人脸子块的LBP特征进行人脸特征匹配获得第一人脸子块和第二人脸子块的相似度；
相似度判别单元，判断所述相似度是否符合相似度要求；如果所述相似度不符合相似度要求，发送第四不合格指令给客户端，重新获得拟审核的伤情图片，所述第四不合格指令包括传错单证；如果所述相似度符合相似度要求，发送信号给上传部160。
本申请所述伤情图片自动化审核装置能够有效降低伤情图片的审核成本，提高理赔效率，并提升客户体验，基于多种轻量级数字图像处理方法和机器学习方法，可以实现离线型伤情图片审核。
如图3所示，是本申请实现伤情图片自动化审核方法的电子设备的结构示意图。
所述电子设备1可以包括处理器10、存储器11和总线，还可以包括存储在所述存储器11中并可在所述处理器10上运行的计算机程序，如伤情图片自动化审核程序12。
其中，所述存储器11至少包括一种类型的可读存储介质，所述可读存储介质包括闪存、移动硬盘、多媒体卡、卡型存储器(例如：SD或DX存储器等)、磁性存储器、磁盘、光盘等。所述存储器11在一些实施例中可以是电子设备1的内部存储单元，例如该电子设备1的移动硬盘。所述存储器11在另一些实施例中也可以是电子设备1的外部存储设备，例如电子设备1上配备的插接式移动硬盘、智能存储卡(Smart Media Card，SMC)、安全数字(Secure Digital，SD)卡、闪存卡(Flash Card)等。进一步地，所述存储器11还可以既包括电子设备1的内部存储单元也包括外部存储设备。所述存储器11不仅可以用于存储安装于电子设备1的应用软件及各类数据，例如伤情图片自动化审核程序的代码等，还可以用于暂时地存储已经输出或者将要输出的数据。
所述处理器10在一些实施例中可以由集成电路组成，例如可以由单个封装的集成电路所组成，也可以是由多个相同功能或不同功能封装的集成电路所组成，包括一个或者多个中央处理器(Central Processing unit，CPU)、微处理器、数字处理芯片、图形处理器及各种控制芯片的组合等。所述处理器10是所述电子设备的控制核心(Control Unit)，利用各种接口和线路连接整个电子设备的各个部件，通过运行或执行存储在所述存储器11内的程序或者模块(例如伤情图片自动化审核程序等)，以及调用存储在所述存储器11内的数据，以执行电子设备1的各种功能和处理数据。
所述总线可以是外设部件互连标准(peripheral component interconnect，简称PCI)总线或扩展工业标准结构(extended industry standard architecture，简称EISA)总线等。该总线可以分为地址总线、数据总线、控制总线等。所述总线被设置为实现所述存储器11以及至少一个处理器10等之间的连接通信。
图3仅示出了具有部件的电子设备，本领域技术人员可以理解的是，图3示出的结构并不构成对所述电子设备1的限定，可以包括比图示更少或者更多的部件，或者组合某些 部件，或者不同的部件布置。
例如，尽管未示出，所述电子设备1还可以包括给各个部件供电的电源(比如电池)，优选地，电源可以通过电源管理装置与所述至少一个处理器10逻辑相连，从而通过电源管理装置实现充电管理、放电管理、以及功耗管理等功能。电源还可以包括一个或一个以上的直流或交流电源、再充电装置、电源故障检测电路、电源转换器或者逆变器、电源状态指示器等任意组件。所述电子设备1还可以包括多种传感器、蓝牙模块、Wi-Fi模块等，在此不再赘述。
进一步地，所述电子设备1还可以包括网络接口，可选地，所述网络接口可以包括有线接口和/或无线接口(如WI-FI接口、蓝牙接口等)，通常用于在该电子设备1与其他电子设备之间建立通信连接。
可选地，该电子设备1还可以包括用户接口，用户接口可以是显示器(Display)、输入单元(比如键盘(Keyboard))，可选地，用户接口还可以是标准的有线接口、无线接口。可选地，在一些实施例中，显示器可以是LED显示器、液晶显示器、触控式液晶显示器以及OLED(Organic Light-Emitting Diode，有机发光二极管)触摸器等。其中，显示器也可以适当的称为显示屏或显示单元，用于显示在电子设备1中处理的信息以及用于显示可视化的用户界面。
应该了解，所述实施例仅为说明之用，在专利申请范围上并不受此结构的限制。
所述电子设备1中的所述存储器11存储的伤情图片自动化审核问程序12是多个指令的组合，在所述处理器10中运行时，可以实现：
获得拟审核的伤情图片；
审核所述伤情图片的皮肤区域，包括：基于色彩空间阈值分割对伤情图片中皮肤区域进行检测；判断检测到的皮肤区域的面积是否符合面积大小要求；如果检测到的皮肤区域的面积不符合面积大小要求，发送第一不合格指令给客户端，重新获得拟审核的伤情图片，所述第一不合格指令包括光照异常指令；如果检测到的皮肤区域的面积符合面积大小要求，从检测到的皮肤区域分离出伤口区域；
审核所述伤情图片的伤口区域，包括：基于色彩空间阈值分割对伤口区域进行检测；判断分离出的伤口面积是否符合伤口面积要求；如果分离出的伤口面积不符合伤口面积要求，发送第二不合格指令给客户端，重新获得拟审核的伤情图片，所述第二不合格指令包括拍摄距离远；如果分离出的伤口面积符合伤口面积要求，基于模板匹配的方法定位伤情图片中出现的人脸；
审核所述伤情图片的人脸，包括：判断定位的人脸是否符合人脸要求；如果定位的人脸不符合人脸要求，则发送第三不合格指令给客户端，重新获得拟审核的伤情图片，所述第三不合格指令包括调整拍摄角度或调整拍摄距离；如果定位的人脸符合人脸要求，获得伤者证件，对伤者证件进行人脸检测，获得伤者证件中的人脸；
审核伤情图片中的人脸和伤者证件中的人脸的相似度，包括：将证件中的人脸和伤情图片中的人脸进行相似度对比；判断所述相似度对比的对比结果是否符合相似度要求；如果对比结果不符合相似度要求，发送第四不合格指令给客户端，重新获得拟审核的伤情图片，所述第四不合格指令包括传错单证；如果对比结果符合相似度要求，上传伤情图片。
具体地，所述处理器10对上述指令的具体实现方法可参考图1对应实施例中相关步骤的描述，在此不赘述。
进一步地，所述电子设备1集成的模块/单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。所述计算机可读介质可以包括：能够携带所述计算机程序代码的任何实体或装置、记录介质、U盘、移动硬盘、磁碟、光盘、计算机存储器、只读存储器(ROM，Read-Only Memory)。
此外，本申请实施例还提出一种计算机可读存储介质，所述计算机可读存储介质可以 是非易失性，也可以是易失性，计算机可读存储介质中包括计算机程序，该计算机程序被处理器执行时实现如下操作：
获得拟审核的伤情图片；
审核所述伤情图片的皮肤区域，包括：基于色彩空间阈值分割对伤情图片中皮肤区域进行检测；判断检测到的皮肤区域的面积是否符合面积大小要求；如果检测到的皮肤区域的面积不符合面积大小要求，发送第一不合格指令给客户端，重新获得拟审核的伤情图片，所述第一不合格指令包括光照异常指令；如果检测到的皮肤区域的面积符合面积大小要求，从检测到的皮肤区域分离出伤口区域；
审核所述伤情图片的伤口区域，包括：基于色彩空间阈值分割对伤口区域进行检测；判断分离出的伤口面积是否符合伤口面积要求；如果分离出的伤口面积不符合伤口面积要求，发送第二不合格指令给客户端，重新获得拟审核的伤情图片，所述第二不合格指令包括拍摄距离远；如果分离出的伤口面积符合伤口面积要求，基于模板匹配的方法定位伤情图片中出现的人脸；
审核所述伤情图片的人脸，包括：判断定位的人脸是否符合人脸要求；如果定位的人脸不符合人脸要求，则发送第三不合格指令给客户端，重新获得拟审核的伤情图片，所述第三不合格指令包括调整拍摄角度或调整拍摄距离；如果定位的人脸符合人脸要求，获得伤者证件，对伤者证件进行人脸检测，获得伤者证件中的人脸；
审核伤情图片中的人脸和伤者证件中的人脸的相似度，包括：将证件中的人脸和伤情图片中的人脸进行相似度对比；判断所述相似度对比的对比结果是否符合相似度要求；如果对比结果不符合相似度要求，发送第四不合格指令给客户端，重新获得拟审核的伤情图片，所述第四不合格指令包括传错单证；如果对比结果符合相似度要求，上传伤情图片。
本申请之计算机可读存储介质的具体实施方式与上述伤情图片自动化审核方法、装置、电子设备的具体实施方式大致相同，在此不再赘述。
本申请所述面向车险理赔的伤情图片自动化审核、装置、电子设备及计算机可读存储介质基于多种轻量级数字图像处理方法和机器学习方法，可以构建离线型伤情图片审核方案，伤情图片全自动化审核，节省人伤理赔的人力成本和交通成本等；实时引导客户在拍摄伤情图片时调整环境光线、距离、拍摄角度，提高一次拍摄符合要求的伤情图片成功概率，从而提升客户体验；计算量小，可部署于客户手机，减少部署远程服务器的成本；仅在算法研发时使用到极少量客户的伤情图片，尽可能保护用户隐私，满足数据使用的合规要求。
在本申请所提供的几个实施例中，应该理解到，所揭露的设备，装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，所述模块的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式。
所述作为分离部件说明的模块可以是或者也可以不是物理上分开的，作为模块显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。
另外，在本申请各个实施例中的各功能模块可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用硬件加软件功能模块的形式实现。
对于本领域技术人员而言，显然本申请不限于上述示范性实施例的细节，而且在不背离本申请的精神或基本特征的情况下，能够以其他的具体形式实现本申请。
因此，无论从哪一点来看，均应将实施例看作是示范性的，而且是非限制性的，本申请的范围由所附权利要求而不是上述说明限定，因此旨在将落在权利要求的等同要件的含义和范围内的所有变化涵括在本申请内。不应将权利要求中的任何附关联图标记视为限制所涉及的权利要求。
此外，显然“包括”一词不排除其他单元或步骤，单数不排除复数。系统权利要求中陈述的多个单元或装置也可以由一个单元或装置通过软件或者硬件来实现。第二等词语用来表示名称，而并不表示任何特定的顺序。
最后应说明的是，以上实施例仅用以说明本申请的技术方案而非限制，尽管参照较佳实施例对本申请进行了详细说明，本领域的普通技术人员应当理解，可以对本申请的技术方案进行修改或等同替换，而不脱离本申请技术方案的精神和范围。

Claims (20)

  1. 一种伤情图片自动化审核方法，其中，所述方法包括：
    获得拟审核的伤情图片；
    审核所述伤情图片的皮肤区域，包括：基于色彩空间阈值分割对伤情图片中皮肤区域进行检测；判断检测到的皮肤区域的面积是否符合面积大小要求；如果检测到的皮肤区域的面积不符合面积大小要求，发送第一不合格指令给客户端，重新获得拟审核的伤情图片，所述第一不合格指令包括光照异常指令；如果检测到的皮肤区域的面积符合面积大小要求，从检测到的皮肤区域分离出伤口区域；
    审核所述伤情图片的伤口区域，包括：基于色彩空间阈值分割对伤口区域进行检测；判断分离出的伤口面积是否符合伤口面积要求；如果分离出的伤口面积不符合伤口面积要求，发送第二不合格指令给客户端，重新获得拟审核的伤情图片，所述第二不合格指令包括拍摄距离远；如果分离出的伤口面积符合伤口面积要求，基于模板匹配的方法定位伤情图片中出现的人脸；
    审核所述伤情图片的人脸，包括：判断定位的人脸是否符合人脸要求；如果定位的人脸不符合人脸要求，则发送第三不合格指令给客户端，重新获得拟审核的伤情图片，所述第三不合格指令包括调整拍摄角度或调整拍摄距离；如果定位的人脸符合人脸要求，获得伤者证件，对伤者证件进行人脸检测，获得伤者证件中的人脸；
    审核伤情图片中的人脸和伤者证件中的人脸的相似度，包括：将证件中的人脸和伤情图片中的人脸进行相似度对比；判断所述相似度对比的对比结果是否符合相似度要求；如果对比结果不符合相似度要求，发送第四不合格指令给客户端，重新获得拟审核的伤情图片，所述第四不合格指令包括传错单证；如果对比结果符合相似度要求，上传伤情图片。
  2. 根据权利要求1所述的伤情图片自动化审核方法，其中，所述审核所述伤情图片的皮肤区域的步骤包括：
    基于色彩空间分析伤情图片的光照；
    设定色彩空间中各参数的光照阈值，获得第一判别条件；
    判断伤情图片的光照是否满足第一判别条件；
    如果伤情图片的光照不满足第一判别条件，发出第一不合格指令，所述第一不合格指令包括光照过暗、光照过曝或不均匀光照；
    如果伤情图片的光照满足第一判别条件，基于色彩空间分析伤情图片的光照的均匀性；
    设定光照均匀度阈值，获得第二判别条件；
    判断伤情图片的光照的均匀性是否满足第二判别条件；
    如果伤情图片的光照的均匀性不满足第二判别条件，发出第一不合格指令，所述第一不合格指令包括局部过暗或局部过曝；
    如果伤情图片的光照的均匀性满足第二判别条件，基于色彩空间阈值分割伤情图片，获得伤情图片的皮肤区域，获得皮肤区域的面积；
    设定皮肤区域面积的阈值，获得第四判别条件；
    判断伤情图片的皮肤区域的面积是否满足第四判别条件；
    如果伤情图片的皮肤区域的面积不满足第四判别条件，发出第一不合格指令，所述第一不合格指令包括皮肤面积过小和拍摄距离过远；
    如果伤情图片的皮肤区域的面积满足第四判别条件，从检测到的皮肤区域分离出伤口区域。
  3. 根据权利要求2所述的伤情图片自动化审核方法，其中，所述基于色彩空间阈值分割伤情图片，获得伤情图片的皮肤区域的步骤包括：
    获得伤情图片的多个色彩空间的数据；
    设定皮肤区域对应的多个色彩空间中各参数的阈值，获得第三判别条件；
    从伤情图片的多个色彩空间的数据中筛选出满足第三判别条件的多个像素点；
    通过形态学闭运算填充所述像素点围成区域的内部空隙，得到一个或多个闭合区域，从而获得皮肤区域。
  4. 根据权利要求1所述的伤情图片自动化审核方法，其中，所述审核所述伤情图片的伤口区域的步骤包括：
    基于色彩空间阈值分割皮肤区域，获得皮肤区域中的伤口区域，获得伤口区域的面积；
    设定伤口区域面积的阈值，获得第六判别条件；
    判断伤口区域的面积是否满足第六判别条件；
    如果伤口区域的面积不满足第六判别条件，发出第二不合格指令，所述第二不合格指令包括伤口面积过小和拍摄距离过远；
    如果伤口区域的面积满足第六判别条件，基于模板匹配的方法定位伤情图片中出现的人脸。
  5. 根据权利要求4所述的伤情图片自动化审核方法，其中，所述基于色彩空间阈值分割皮肤区域，获得皮肤区域中的伤口区域的步骤包括：
    获得皮肤区域的多个色彩空间的数据；
    设定伤口区域对应的多个色彩空间中各参数的阈值，获得第五判别条件；
    从皮肤区域的多个色彩空间的数据中筛选出满足判别条件第五判别条件的多个像素点；
    通过形态学闭运算填充所述像素点围成区域的内部空隙，得到一个或多个闭合区域，从而获得伤口区域。
  6. 根据权利要求1所述的伤情图片自动化审核方法，其中，所述基于模板匹配的方法定位伤情图片中出现的人脸的步骤包括：
    将伤情图片的RGB图像转换为灰度图像；
    以目前公开的人脸数据集制作平均脸并作为模板，采用模板匹配算法对所述灰度图像进行扫描检测，检测出人脸。
  7. 根据权利要求1所述的伤情图片自动化审核方法，其中，所述审核所述伤情图片的人脸的步骤包括：
    采用非极大值抑制算法消除同一位置检测重叠的、多余的人脸；
    若检测无人脸，发出第三不合格指令，所述第三不合格指令包括伤者调整拍摄距离或调整人脸拍摄角度；
    若检测到2张及以上人脸，发出第四不合格指令，所述第四不合格指令包括伤者以外的人员暂离拍摄画面；
    若检测到1张人脸，获得伤者证件，对伤者证件进行人脸检测，获得伤者证件中的人脸。
  8. 一种伤情图片自动化审核装置，其中，所述装置包括：
    获得部，获得拟审核的伤情图片；
    第一审核部，审核所述伤情图片的皮肤区域，包括：皮肤区域检测模块，基于色彩空间阈值分割对伤情图片中皮肤区域进行检测；第一面积判断模块，判断检测到的皮肤区域的面积是否符合面积大小要求，如果检测到的皮肤区域的面积不符合面积大小要求，发送第一不合格指令给客户端，获得部重新获得拟审核的伤情图片，所述第一不合格指令包括光照异常指令；如果检测到的皮肤区域的面积符合面积大小要求，发送信号给伤口区域获得模块；伤口区域获得模块，从检测到的皮肤区域分离出伤口区域并发送给第二审核部；
    第二审核部，审核所述伤情图片的伤口区域，包括：伤口区域检测模块，基于色彩空间阈值分割对第一审核部获得的皮肤区域中的伤口区域进行检测；第二面积判断模块，判断分离出的伤口面积是否符合伤口面积要求，如果分离出的伤口面积不符合伤口面积要求， 发送第二不合格指令给客户端，获得部重新获得拟审核的伤情图片，所述第二不合格指令包括拍摄距离远；如果分离出的伤口面积符合伤口面积要求，发送信号给第一人脸获得模块；第一人脸获得模块，基于模板匹配的方法定位伤情图片中出现的人脸并发送给第三审核部和相似度获得部；
    第三审核部，审核所述伤情图片的人脸，包括：人脸判断模块，判断定位的人脸是否符合人脸要求，如果定位的人脸不符合人脸要求，则发送第三不合格指令给客户端，获得部重新获得拟审核的伤情图片，所述第三不合格指令包括调整拍摄角度或调整拍摄距离；如果定位的人脸符合人脸要求，发送信号给第二人脸获得模块；第二人脸获得模块，获得伤者证件，对伤者证件进行人脸检测，获得伤者证件中的人脸并发送给相似度获得部；
    相似度获得部，将证件中的人脸和伤情图片中的人脸进行相似度对比，判断所述相似度对比的对比结果是否符合相似度要求，如果对比结果不符合相似度要求，发送第四不合格指令给客户端，获得部重新获得拟审核的伤情图片，所述第四不合格指令包括传错单证；如果对比结果符合相似度要求，发送信号给上传部；
    上传部，上传符合第一审核部、第二审核部、第三审核部和相似度获得部要求的伤情图片。
  9. 一种电子设备，其中，所述电子设备包括：
    至少一个处理器；以及，
    与所述至少一个处理器通信连接的存储器；其中，
    所述存储器存储有可被所述至少一个处理器执行的指令，所述指令被所述至少一个处理器执行，以使所述至少一个处理器能够执行如下步骤：
    获得拟审核的伤情图片；
    审核所述伤情图片的皮肤区域，包括：基于色彩空间阈值分割对伤情图片中皮肤区域进行检测；判断检测到的皮肤区域的面积是否符合面积大小要求；如果检测到的皮肤区域的面积不符合面积大小要求，发送第一不合格指令给客户端，重新获得拟审核的伤情图片，所述第一不合格指令包括光照异常指令；如果检测到的皮肤区域的面积符合面积大小要求，从检测到的皮肤区域分离出伤口区域；
    审核所述伤情图片的伤口区域，包括：基于色彩空间阈值分割对伤口区域进行检测；判断分离出的伤口面积是否符合伤口面积要求；如果分离出的伤口面积不符合伤口面积要求，发送第二不合格指令给客户端，重新获得拟审核的伤情图片，所述第二不合格指令包括拍摄距离远；如果分离出的伤口面积符合伤口面积要求，基于模板匹配的方法定位伤情图片中出现的人脸；
    审核所述伤情图片的人脸，包括：判断定位的人脸是否符合人脸要求；如果定位的人脸不符合人脸要求，则发送第三不合格指令给客户端，重新获得拟审核的伤情图片，所述第三不合格指令包括调整拍摄角度或调整拍摄距离；如果定位的人脸符合人脸要求，获得伤者证件，对伤者证件进行人脸检测，获得伤者证件中的人脸；
    审核伤情图片中的人脸和伤者证件中的人脸的相似度，包括：将证件中的人脸和伤情图片中的人脸进行相似度对比；判断所述相似度对比的对比结果是否符合相似度要求；如果对比结果不符合相似度要求，发送第四不合格指令给客户端，重新获得拟审核的伤情图片，所述第四不合格指令包括传错单证；如果对比结果符合相似度要求，上传伤情图片。
  10. 根据权利要求9所述的电子设备，其中，所述至少一个处理器执行的所述审核所述伤情图片的皮肤区域的步骤包括：
    基于色彩空间分析伤情图片的光照；
    设定色彩空间中各参数的光照阈值，获得第一判别条件；
    判断伤情图片的光照是否满足第一判别条件；
    如果伤情图片的光照不满足第一判别条件，发出第一不合格指令，所述第一不合格指 令包括光照过暗、光照过曝或不均匀光照；
    如果伤情图片的光照满足第一判别条件，基于色彩空间分析伤情图片的光照的均匀性；
    设定光照均匀度阈值，获得第二判别条件；
    判断伤情图片的光照的均匀性是否满足第二判别条件；
    如果伤情图片的光照的均匀性不满足第二判别条件，发出第一不合格指令，所述第一不合格指令包括局部过暗或局部过曝；
    如果伤情图片的光照的均匀性满足第二判别条件，基于色彩空间阈值分割伤情图片，获得伤情图片的皮肤区域，获得皮肤区域的面积；
    设定皮肤区域面积的阈值，获得第四判别条件；
    判断伤情图片的皮肤区域的面积是否满足第四判别条件；
    如果伤情图片的皮肤区域的面积不满足第四判别条件，发出第一不合格指令，所述第一不合格指令包括皮肤面积过小和拍摄距离过远；
    如果伤情图片的皮肤区域的面积满足第四判别条件，从检测到的皮肤区域分离出伤口区域。
  11. 根据权利要求10所述的电子设备，其中，所述至少一个处理器执行的所述基于色彩空间阈值分割伤情图片，获得伤情图片的皮肤区域的步骤包括：
    获得伤情图片的多个色彩空间的数据；
    设定皮肤区域对应的多个色彩空间中各参数的阈值，获得第三判别条件；
    从伤情图片的多个色彩空间的数据中筛选出满足第三判别条件的多个像素点；
    通过形态学闭运算填充所述像素点围成区域的内部空隙，得到一个或多个闭合区域，从而获得皮肤区域。
  12. 根据权利要求9所述的电子设备，其中，所述至少一个处理器执行的所述审核所述伤情图片的伤口区域的步骤包括：
    基于色彩空间阈值分割皮肤区域，获得皮肤区域中的伤口区域，获得伤口区域的面积；
    设定伤口区域面积的阈值，获得第六判别条件；
    判断伤口区域的面积是否满足第六判别条件；
    如果伤口区域的面积不满足第六判别条件，发出第二不合格指令，所述第二不合格指令包括伤口面积过小和拍摄距离过远；
    如果伤口区域的面积满足第六判别条件，基于模板匹配的方法定位伤情图片中出现的人脸。
  13. 根据权利要求12所述的电子设备，其中，所述至少一个处理器执行的所述基于色彩空间阈值分割皮肤区域，获得皮肤区域中的伤口区域的步骤包括：
    获得皮肤区域的多个色彩空间的数据；
    设定伤口区域对应的多个色彩空间中各参数的阈值，获得第五判别条件；
    从皮肤区域的多个色彩空间的数据中筛选出满足判别条件第五判别条件的多个像素点；
    通过形态学闭运算填充所述像素点围成区域的内部空隙，得到一个或多个闭合区域，从而获得伤口区域。
  14. 根据权利要求9所述的电子设备，其中，所述至少一个处理器执行的所述基于模板匹配的方法定位伤情图片中出现的人脸的步骤包括：
    将伤情图片的RGB图像转换为灰度图像；
    以目前公开的人脸数据集制作平均脸并作为模板，采用模板匹配算法对所述灰度图像进行扫描检测，检测出人脸。
  15. 根据权利要求9所述的电子设备，其中，所述至少一个处理器执行的所述审核所述伤情图片的人脸的步骤包括：
    采用非极大值抑制算法消除同一位置检测重叠的、多余的人脸；
    若检测无人脸，发出第三不合格指令，所述第三不合格指令包括伤者调整拍摄距离或调整人脸拍摄角度；
    若检测到2张及以上人脸，发出第四不合格指令，所述第四不合格指令包括伤者以外的人员暂离拍摄画面；
    若检测到1张人脸，获得伤者证件，对伤者证件进行人脸检测，获得伤者证件中的人脸。
  16. 一种计算机可读存储介质，存储有计算机程序，其中，所述计算机程序被处理器执行时实现如下步骤：
    获得拟审核的伤情图片；
    审核所述伤情图片的皮肤区域，包括：基于色彩空间阈值分割对伤情图片中皮肤区域进行检测；判断检测到的皮肤区域的面积是否符合面积大小要求；如果检测到的皮肤区域的面积不符合面积大小要求，发送第一不合格指令给客户端，重新获得拟审核的伤情图片，所述第一不合格指令包括光照异常指令；如果检测到的皮肤区域的面积符合面积大小要求，从检测到的皮肤区域分离出伤口区域；
    审核所述伤情图片的伤口区域，包括：基于色彩空间阈值分割对伤口区域进行检测；判断分离出的伤口面积是否符合伤口面积要求；如果分离出的伤口面积不符合伤口面积要求，发送第二不合格指令给客户端，重新获得拟审核的伤情图片，所述第二不合格指令包括拍摄距离远；如果分离出的伤口面积符合伤口面积要求，基于模板匹配的方法定位伤情图片中出现的人脸；
    审核所述伤情图片的人脸，包括：判断定位的人脸是否符合人脸要求；如果定位的人脸不符合人脸要求，则发送第三不合格指令给客户端，重新获得拟审核的伤情图片，所述第三不合格指令包括调整拍摄角度或调整拍摄距离；如果定位的人脸符合人脸要求，获得伤者证件，对伤者证件进行人脸检测，获得伤者证件中的人脸；
    审核伤情图片中的人脸和伤者证件中的人脸的相似度，包括：将证件中的人脸和伤情图片中的人脸进行相似度对比；判断所述相似度对比的对比结果是否符合相似度要求；如果对比结果不符合相似度要求，发送第四不合格指令给客户端，重新获得拟审核的伤情图片，所述第四不合格指令包括传错单证；如果对比结果符合相似度要求，上传伤情图片。
  17. 根据权利要求16所述的计算机可读存储介质，其中，所述计算机程序被处理器执行时实现的所述审核所述伤情图片的皮肤区域的步骤包括：
    基于色彩空间分析伤情图片的光照；
    设定色彩空间中各参数的光照阈值，获得第一判别条件；
    判断伤情图片的光照是否满足第一判别条件；
    如果伤情图片的光照不满足第一判别条件，发出第一不合格指令，所述第一不合格指令包括光照过暗、光照过曝或不均匀光照；
    如果伤情图片的光照满足第一判别条件，基于色彩空间分析伤情图片的光照的均匀性；
    设定光照均匀度阈值，获得第二判别条件；
    判断伤情图片的光照的均匀性是否满足第二判别条件；
    如果伤情图片的光照的均匀性不满足第二判别条件，发出第一不合格指令，所述第一不合格指令包括局部过暗或局部过曝；
    如果伤情图片的光照的均匀性满足第二判别条件，基于色彩空间阈值分割伤情图片，获得伤情图片的皮肤区域，获得皮肤区域的面积；
    设定皮肤区域面积的阈值，获得第四判别条件；
    判断伤情图片的皮肤区域的面积是否满足第四判别条件；
    如果伤情图片的皮肤区域的面积不满足第四判别条件，发出第一不合格指令，所述第 一不合格指令包括皮肤面积过小和拍摄距离过远；
    如果伤情图片的皮肤区域的面积满足第四判别条件，从检测到的皮肤区域分离出伤口区域。
  18. 根据权利要求16所述的计算机可读存储介质，其中，所述计算机程序被处理器执行时实现的所述审核所述伤情图片的伤口区域的步骤包括：
    基于色彩空间阈值分割皮肤区域，获得皮肤区域中的伤口区域，获得伤口区域的面积；
    设定伤口区域面积的阈值，获得第六判别条件；
    判断伤口区域的面积是否满足第六判别条件；
    如果伤口区域的面积不满足第六判别条件，发出第二不合格指令，所述第二不合格指令包括伤口面积过小和拍摄距离过远；
    如果伤口区域的面积满足第六判别条件，基于模板匹配的方法定位伤情图片中出现的人脸。
  19. 根据权利要求16所述的计算机可读存储介质，其中，所述计算机程序被处理器执行时实现的所述基于模板匹配的方法定位伤情图片中出现的人脸的步骤包括：
    将伤情图片的RGB图像转换为灰度图像；
    以目前公开的人脸数据集制作平均脸并作为模板，采用模板匹配算法对所述灰度图像进行扫描检测，检测出人脸。
  20. 根据权利要求16所述的计算机可读存储介质，其中，所述计算机程序被处理器执行时实现的所述审核所述伤情图片的人脸的步骤包括：
    采用非极大值抑制算法消除同一位置检测重叠的、多余的人脸；
    若检测无人脸，发出第三不合格指令，所述第三不合格指令包括伤者调整拍摄距离或调整人脸拍摄角度；
    若检测到2张及以上人脸，发出第四不合格指令，所述第四不合格指令包括伤者以外的人员暂离拍摄画面；
    若检测到1张人脸，获得伤者证件，对伤者证件进行人脸检测，获得伤者证件中的人脸。
PCT/CN2020/125071 2020-08-27 2020-10-30 伤情图片自动化审核方法、装置、电子设备及存储介质 WO2021147435A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010879301.1A CN112037235B (zh) 2020-08-27 2020-08-27 伤情图片自动化审核方法、装置、电子设备及存储介质
CN202010879301.1 2020-08-27

Publications (1)

Publication Number Publication Date
WO2021147435A1 true WO2021147435A1 (zh) 2021-07-29

Family

ID=73585822

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/125071 WO2021147435A1 (zh) 2020-08-27 2020-10-30 伤情图片自动化审核方法、装置、电子设备及存储介质

Country Status (2)

Country Link
CN (1) CN112037235B (zh)
WO (1) WO2021147435A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115252124A (zh) * 2022-09-27 2022-11-01 山东博达医疗用品股份有限公司 一种基于伤情图片数据分析的缝线用量预估方法及系统

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114240364A (zh) * 2021-12-13 2022-03-25 深圳壹账通智能科技有限公司 一种工伤自动审核的方法、装置、计算机设备及存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130245400A1 (en) * 2010-04-27 2013-09-19 Solar System Beauty Corporation Abnormal skin area calculating system and calculating method thereof
CN109345396A (zh) * 2018-09-13 2019-02-15 医倍思特（北京）医疗信息技术有限公司 一种智能人伤理赔管理系统
CN110009509A (zh) * 2019-01-02 2019-07-12 阿里巴巴集团控股有限公司 评估车损识别模型的方法及装置
CN110009508A (zh) * 2018-12-25 2019-07-12 阿里巴巴集团控股有限公司 一种车险自动赔付方法和系统
CN110503403A (zh) * 2019-08-27 2019-11-26 陕西蓝图司法鉴定中心 针对损伤程度的分析及鉴定智能自动化系统及方法

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105761260B (zh) * 2016-02-15 2018-09-28 天津大学 一种皮肤图像患处分割方法
CN109509102B (zh) * 2018-08-14 2023-04-11 深圳平安医疗健康科技服务有限公司 理赔决策方法、装置、计算机设备和存储介质
CN109544103A (zh) * 2018-10-30 2019-03-29 平安医疗健康管理股份有限公司 一种理赔模型的构建方法、装置、服务器及存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130245400A1 (en) * 2010-04-27 2013-09-19 Solar System Beauty Corporation Abnormal skin area calculating system and calculating method thereof
CN109345396A (zh) * 2018-09-13 2019-02-15 医倍思特（北京）医疗信息技术有限公司 一种智能人伤理赔管理系统
CN110009508A (zh) * 2018-12-25 2019-07-12 阿里巴巴集团控股有限公司 一种车险自动赔付方法和系统
CN110009509A (zh) * 2019-01-02 2019-07-12 阿里巴巴集团控股有限公司 评估车损识别模型的方法及装置
CN110503403A (zh) * 2019-08-27 2019-11-26 陕西蓝图司法鉴定中心 针对损伤程度的分析及鉴定智能自动化系统及方法

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115252124A (zh) * 2022-09-27 2022-11-01 山东博达医疗用品股份有限公司 一种基于伤情图片数据分析的缝线用量预估方法及系统
CN115252124B (zh) * 2022-09-27 2022-12-20 山东博达医疗用品股份有限公司 一种基于伤情图片数据分析的缝线用量预估方法及系统

Also Published As

Publication number Publication date
CN112037235A (zh) 2020-12-04
CN112037235B (zh) 2023-01-10

Similar Documents

Publication Publication Date Title
US20220012468A1 (en) Method, system and computer program for comparing images
WO2019104930A1 (zh) 一种身份鉴定方法、电子装置及计算机可读存储介质
CN101609500B (zh) 出入境数字人像相片质量评估方法
WO2021147435A1 (zh) 伤情图片自动化审核方法、装置、电子设备及存储介质
US11281892B2 (en) Technologies for efficient identity recognition based on skin features
KR102469720B1 (ko) 전자 장치 및 이를 이용한 눈의 충혈도 판단 방법
CN105120167A (zh) 一种证照相机及证照拍摄方法
US20200104567A1 (en) Obstruction detection during facial recognition processes
US20130329970A1 (en) Image authentication apparatus, image processing system, control program for image authentication apparatus, computer-readable recording medium, and image authentication method
CN106982426A (zh) 一种远程实现旧卡实名制的方法与系统
WO2021151277A1 (zh) 目标物损伤程度判定方法、装置、电子设备及存储介质
KR101301821B1 (ko) 안색 정보 생성 장치 및 그 방법, 안색 정보를 이용한 건강 상태 판단 장치 및 그 방법, 건강 분류 함수 생성 장치 및 그 방법
US11678180B2 (en) Iris recognition workflow
CN104143086A (zh) 人像比对在移动终端操作系统上的应用技术
CN112801062B (zh) 直播视频识别方法、装置、设备及介质
US9349071B2 (en) Device for detecting pupil taking account of illuminance and method thereof
CN106557729A (zh) 用于处理人脸图像的设备和方法
CN108875477A (zh) 曝光控制方法、装置和系统及存储介质
WO2021217853A1 (zh) 损伤图像智能定损方法、装置、电子设备及存储介质
US20230252119A1 (en) Authentication device, authentication method, and recording medium
KR20080047190A (ko) 자동 역광 보정 기능을 구비한 이미지 촬영기기 및 그 방법
CN112541899B (zh) 证件的残缺检测方法、装置、电子设备及计算机存储介质
JP2019211921A (ja) 物体認識システムおよび物体認識方法
JP2019113959A (ja) 鼻紋照合システム、鼻紋照合方法及び鼻紋照合プログラム
JP2016110467A (ja) 照合システム、照合方法、対面型照合システムおよびウォークスルー型照合システム

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20915802

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20915802

Country of ref document: EP

Kind code of ref document: A1