CN-122027886-A - Shooting gesture recommending method and device and electronic equipment

CN122027886ACN 122027886 ACN122027886 ACN 122027886ACN-122027886-A

Abstract

The invention discloses a shooting gesture recommending method, a shooting gesture recommending device and electronic equipment, wherein the method comprises the steps of acquiring a real-time shooting picture, and respectively carrying out scene semantic recognition and target object feature extraction on the real-time shooting picture to obtain a target scene identifier and target object parameters; selecting at least two candidate gesture templates from a preset gesture template library based on a target scene identifier, matching target object parameters with preset matching parameters of each candidate gesture template to determine a recommended gesture template, scaling and adjusting characteristic points of the recommended gesture template based on the proportion of a real-time shooting picture, overlapping and displaying the adjusted characteristic points on a preview interface in a preset visual mode, detecting the similarity between the current gesture of the target object and the recommended gesture template in real time, and triggering a shooting prompt under the condition that the similarity meets preset conditions. The gesture recommendation method and device achieve gesture recommendation scene and personalized adaptation, improve user gesture matching efficiency and photographing experience, and ensure accurate triggering of photographing time.

Inventors

TANG JIALE
YE JILONG
LI JUNTAO

Assignees

浙江商识信息技术有限公司

Dates

Publication Date: 20260512
Application Date: 20260204

Claims (10)

1. A shooting pose recommendation method, comprising: acquiring a real-time shooting picture, and respectively carrying out scene semantic recognition and target object feature extraction on the real-time shooting picture to obtain a target scene identifier and a target object parameter; Selecting at least two candidate gesture templates from a preset gesture template library based on the target scene identification; matching the target object parameters with preset matching parameters of each candidate gesture template to determine a recommended gesture template; Scaling and adjusting the characteristic points of the recommended gesture template based on the proportion of the real-time shooting picture, and superposing and displaying the adjusted characteristic points on a preview interface in a preset visual mode; And detecting the similarity between the current gesture of the target object and the recommended gesture template in real time, and triggering a photographing prompt under the condition that the similarity meets a preset condition.
2. The method according to claim 1, wherein the performing scene semantic recognition and target object feature extraction on the real-time shot image to obtain a target scene identifier and a target object parameter respectively includes: performing scene semantic recognition on the real-time shooting picture through a lightweight neural network model, outputting probability distribution results of multiple types of scenes, and determining the target scene identification according to the category with the highest confidence in the probability distribution results; And extracting standard joint point coordinates of the target object from the real-time shooting picture through a gesture estimation algorithm, and calculating to obtain the target object parameters based on the standard joint point coordinates.
3. The method of claim 2, wherein the target object parameters include a target object height ratio and a body type category, and the calculating the target object parameters based on the standard joint point coordinates includes: calculating to obtain the height ratio of the target object through the ratio of the vertical coordinate difference value of the upper and lower end joint points of the target object to the height of the real-time shooting picture; calculating a shoulder width pixel value and a hip width pixel value of the target object based on the standard articulation point coordinates; And mapping the ratio of the shoulder width pixel value to the hip width pixel value to a preset interval, and dividing the body type category.
4. A method according to claim 3, wherein the preset gesture template library comprises at least two gesture templates to be selected and corresponding scene identifiers to be selected, and the selecting at least two candidate gesture templates from the preset gesture template library based on the target scene identifiers comprises: Selecting at least two gesture templates to be selected corresponding to the scene identification to be selected matched with the target scene identification from the preset gesture template library; And determining at least two gesture templates to be selected as the candidate gesture templates.
5. The method of claim 4, wherein the predetermined matching parameters include a list of fit types and ideal height duty cycles, and wherein said matching the target object parameters to the predetermined matching parameters of each of the candidate gesture templates to determine a recommended gesture template comprises: determining a body type matching result based on the body type category and the aptamer list, and determining the duty ratio matching result based on the target object height duty ratio and the ideal height duty ratio; determining the comprehensive score of each candidate gesture template based on the body type matching result, the duty ratio matching result and a preset weight coefficient; and determining the recommended gesture template from the candidate gesture templates according to the comprehensive scores.
6. The method of claim 5, wherein scaling the feature points of the recommended gesture template based on the scale of the real-time captured image comprises: Mapping the normalized feature point coordinates of the recommended gesture template from a normalized coordinate system to a coordinate system of the real-time shooting picture through coordinate transformation; and taking the ratio of the height ratio of the target object to the ideal height ratio of the recommended gesture template as a scaling factor to scale and adjust the characteristic points.
7. The method of claim 6, wherein displaying the adjusted feature point in a preview interface in a preset visual form in a superimposed manner, comprises: constructing a gesture framework of the target object based on the adjusted characteristic points; And superposing and displaying the gesture framework on the real-time shooting picture of the preview interface in a semitransparent line mode.
8. The method according to claim 1, wherein detecting, in real time, a similarity between a current gesture of the target object and the recommended gesture template, and triggering a photographing prompt if the similarity satisfies a preset condition, includes: determining N joint point coordinates of the current gesture of the target object and N joint point coordinates corresponding to the recommended gesture template; Calculating a normalized average joint point error between the joint point coordinates of the current gesture of the target object and the joint point coordinates of the recommended gesture template, and taking the inversely proportional numerical value of the normalized average joint point error as the similarity; And calculating the normalized average joint point error by taking the shoulder-hip distance as a normalized reference, wherein the similarity meeting the preset condition comprises that the similarity is continuously greater than or equal to a preset threshold and reaches a set duration.
9. A photographing posture recommending apparatus, characterized by comprising: the image and feature acquisition module is used for acquiring a real-time shooting image, and respectively carrying out scene semantic recognition and target object feature extraction on the real-time shooting image to obtain a target scene identifier and target object parameters; The candidate template screening module is used for selecting at least two candidate gesture templates from a preset gesture template library based on the target scene identification; The recommended template matching module is used for matching the target object parameters with preset matching parameters of the candidate gesture templates so as to determine recommended gesture templates; the characteristic point adjustment and display module is used for performing scaling adjustment on the characteristic points of the recommended gesture template based on the proportion of the real-time shooting picture, and overlapping and displaying the adjusted characteristic points on a preview interface in a preset visual mode; The similarity detection and prompt module is used for detecting the similarity between the current gesture of the target object and the recommended gesture template in real time, and triggering a photographing prompt under the condition that the similarity meets the preset condition.
10. An electronic device, the electronic device comprising: At least one processor; and a memory communicatively coupled to the at least one processor; Wherein the memory stores a computer program executable by the at least one processor to enable the at least one processor to perform the shooting pose recommendation method of any of claims 1-8.

Description

Shooting gesture recommending method and device and electronic equipment Technical Field The invention relates to the technical field of computer vision and man-machine interaction, in particular to a shooting gesture recommending method and device and electronic equipment. Background With the popularization of the photographing function of the smart phone, the demand of users for high-quality self-photographing is increasing, however, the common users often lack the expertise of composition and gesture design, so that the photographing effect is difficult to reach expectations. Most of existing photographing APP only provides static filters or fixed gesture schematic diagrams, so that gesture guidance requirements of users cannot be fundamentally met, meanwhile, a mainstream image comparison method (such as image difference, structural similarity index SSIM and the like) is extremely sensitive to factors such as illumination change, photographing angles and shielding objects, false alarm or missing report is easy to generate, and the traditional point-to-point characteristic comparison is extremely sensitive to small shake of a camera and slight deviation of the gesture of a target object, and small pixel dislocation can cause large-area false change alarm, so that the reliability of photographing guidance is further affected. The defects of the existing photographing auxiliary tool are characterized in that firstly, scene adaptability is poor, a generalized gesture template cannot be adapted to composition requirements of different scenes such as beach, credentials, parties and the like, secondly, personalized design is lacking, individual differences such as heights and body types of users are not considered, recommended gestures are difficult to imitate by users, thirdly, interaction feedback is lacking, gestures are displayed only through static pictures, real-time gesture guidance is not available, users cannot judge whether the gestures of the users are in place, fourthly, false triggering rate is high, part of APP only relies on simple face detection to prompt photographing, coordination of the overall gestures is ignored, and thirdly, data utilization is insufficient, high-quality cases successfully photographed by the users are not effectively precipitated and utilized and cannot be used for optimizing subsequent gesture recommendation strategies. Therefore, the current technical field needs an intelligent photographing guiding method which can integrate scene understanding, human body parameter measurement and real-time gesture comparison, realizes the 'what you see is what you get' immersive guiding experience through multi-dimensional technology cooperation, and really solves a plurality of pain points of the existing tool. Disclosure of Invention The invention provides a shooting gesture recommending method, a shooting gesture recommending device and electronic equipment, which are used for realizing scene and personalized adaptation of gesture recommendation, improving the gesture matching efficiency and shooting experience of a user and ensuring accurate triggering of shooting time. According to an aspect of the present invention, there is provided a photographing posture recommending method including: Acquiring a real-time shooting picture, and respectively carrying out scene semantic recognition and target object feature extraction on the real-time shooting picture to obtain a target scene identifier and target object parameters; selecting at least two candidate gesture templates from a preset gesture template library based on the target scene identification; Matching the target object parameters with preset matching parameters of each candidate gesture template to determine a recommended gesture template; scaling and adjusting the characteristic points of the recommended gesture template based on the proportion of the real-time shooting picture, and superposing and displaying the adjusted characteristic points on a preview interface in a preset visual mode; And detecting the similarity between the current gesture of the target object and the recommended gesture template in real time, and triggering a photographing prompt under the condition that the similarity meets the preset condition. According to another aspect of the present invention, there is provided a photographing posture recommending apparatus including: the picture and feature acquisition module is used for acquiring a real-time shooting picture, and respectively carrying out scene semantic recognition and target object feature extraction on the real-time shooting picture to obtain a target scene identifier and target object parameters; The candidate template screening module is used for selecting at least two candidate gesture templates from a preset gesture template library based on the target scene identification; the recommended template matching module is used for matching the target object parameters with preset matching