CN115967837A

CN115967837A - Method, device, equipment and medium for content interaction based on web course video

Info

Publication number: CN115967837A
Application number: CN202111183309.5A
Authority: CN
Inventors: 王红梅; 潘潇
Original assignee: Guangzhou Shiyuan Electronics Thecnology Co Ltd; Guangzhou Shirui Electronics Co Ltd
Current assignee: Guangzhou Shiyuan Electronics Thecnology Co Ltd; Guangzhou Shirui Electronics Co Ltd
Priority date: 2021-10-11
Filing date: 2021-10-11
Publication date: 2023-04-14

Abstract

The invention relates to a method, a device, equipment and a medium for content interaction based on a web lesson video, wherein the method comprises the following steps: collecting a network course picture when a network course video is played, wherein the network course picture is a static picture displayed at one moment when the network course video is played; determining a picture range to be analyzed on the web lesson picture based on a preset range determination rule; according to the picture range, capturing a partial area of the web lesson picture, and generating a picture from the captured partial area; and determining the content category of the picture according to the content of the picture, determining recommended content according to the content category, and displaying the recommended content when the network course video is played. The invention realizes the independent recommendation of the extended learning content according to the online class picture content, thereby leading students to conveniently and quickly develop the extended learning when watching online classes.

Description

Method, device, equipment and medium for content interaction based on web course video

Technical Field

The invention relates to the technical field of teaching interaction, in particular to a method, a device, equipment and a medium for content interaction based on a web course video.

Background

When a student uses the equipment to watch the online class, the student generally only can watch the content in the online class unilaterally, and the student cannot or is inconvenient to carry out extended learning based on the content in the online class. For example, in a course where students watch a section of course that teaches them to recognize fruits, they can only learn introductions to some fruits in the course, and they cannot continue to learn other knowledge of these fruits. If the extended learning is to be continued, the user can only search in the search engine by himself.

Disclosure of Invention

Embodiments of the present invention are directed to overcoming at least one of the drawbacks (deficiencies) of the prior art, and providing a method, an apparatus, a device, and a medium for performing content interaction based on a web lesson video, so as to solve the problem that learning content cannot be expanded conveniently and quickly when a web lesson video is played.

In a first aspect, an embodiment of the present invention provides a method for performing content interaction based on a web-lesson video, including:

when the online course video is played, acquiring an online course picture, wherein the online course picture is a static picture displayed at one moment when the online course video is played;

determining a picture range to be analyzed on the web lesson picture based on a preset range determination rule;

according to the picture range, capturing a partial area of the web lesson picture, and generating a picture from the captured partial area;

determining the content category of the picture according to the content of the picture;

and determining recommended content according to the content category, and displaying the recommended content when the online lesson video is played.

Optionally, determining, based on a preset range determination rule, a range of pictures to be analyzed in the web lesson picture, including:

and determining a picture range which corresponds to the online lesson picture and needs to be analyzed according to the video characteristics of the online lesson video based on the corresponding relation between the preset video characteristics and the picture range.

Optionally, the video characteristics are a web course type or a web course picture layout;

acquiring an analysis range circled by a user in the web course picture;

and determining a picture range to be analyzed on the web lesson picture according to the analysis range based on a preset range determination rule.

Optionally, determining, based on a preset range determination rule, a picture range to be analyzed in the web lesson picture according to the analysis range, including:

based on a preset shape, taking the minimum picture with the shape as the preset shape and including the analysis range as a picture range needing to be analyzed in the web lesson picture.

Optionally, when the online lesson video is played, acquiring an online lesson picture includes:

when the online class video is played, online class pictures are collected for multiple times at a fixed speed, and one online class picture displayed at the current moment is collected every time.

Optionally, determining the content category of the picture according to the content of the picture includes:

for one online class picture, capturing a partial area of the online class picture corresponding to one picture range to generate a picture;

if the currently generated picture is the first picture generated corresponding to the picture range, determining the content type of the currently generated picture according to the content of the currently generated picture;

if the currently generated picture is not the first picture generated corresponding to the picture range, performing difference detection on the currently generated picture and the picture generated last time and corresponding to the same picture range to obtain a difference degree;

and determining whether the content of the currently generated picture is the content category of the currently generated picture according to the difference degree.

Optionally, determining recommended content according to the content category, and displaying the recommended content when the web lesson video is played, includes:

for the collected web lesson pictures, if the content category of the generated pictures is determined according to the content of the generated pictures, determining recommended content corresponding to the web lesson pictures according to the content category, and displaying the recommended content during the video playing of the web lessons;

if the content type of the generated picture is not determined according to the content of the generated picture, displaying the recommended content which is determined by the corresponding to the web course picture acquired before the current time when the web course video is played.

inputting the pictures into a multi-label classification model, respectively judging the possibility of whether the pictures contain each content type according to the content of the pictures, and outputting the probability of whether the pictures contain each content type; when the probability is larger than a preset probability threshold value, taking the content category corresponding to the probability as the content category of the picture;

or inputting the pictures into a multi-classification model, judging the possibility that the pictures belong to each content category respectively according to the content of the pictures, outputting the probability that the pictures belong to each content category respectively, and taking the content category corresponding to the maximum probability as the content category of the pictures.

Optionally, determining recommended content according to the content category includes:

determining a recommendation category from the content category, the recommendation category comprising one or more elements;

and selecting one or more elements from the recommendation category as recommendation content.

Optionally, selecting one or more elements from the recommendation category as recommendation content includes:

respectively extracting feature information of the element and the picture, wherein the feature information comprises one or more of the following: color, shape;

and selecting one or more elements from the recommendation category as recommended contents according to the matching degree between the elements and the feature information of the pictures.

In a second aspect, an embodiment of the present invention provides an apparatus for content interaction based on a web lesson video, including:

the picture acquisition module is used for acquiring a network course picture when a network course video is played, wherein the network course picture is a static picture displayed at one moment when the network course video is played;

the range determining module is used for determining a picture range to be analyzed in the web lesson picture based on a preset range determining rule;

the picture generating module is used for capturing partial areas of the network course picture according to the picture range and generating pictures from the captured partial areas;

and the recommendation display module is used for determining the content category of the picture according to the content of the picture, determining the recommendation content according to the content category and displaying the recommendation content when the web lesson video is played.

In a third aspect, an embodiment of the present invention provides a computer device, including a processor and a memory, where the memory stores a computer program, and is characterized in that when the processor executes the computer program, the method for performing content interaction based on a web-lesson video is implemented.

In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, where the computer program is executed by a processor to implement the above-mentioned method for performing content interaction based on a web-lesson video.

Compared with the prior art, the embodiment of the invention has the following beneficial effects: the embodiment of the invention collects the online class pictures when the online class video is played, generates the pictures to be analyzed according to the online class pictures and analyzes the contents of the pictures to obtain the recommended contents, and displays the recommended contents when the online class video is played, thereby realizing the purpose of autonomously recommending the extended learning contents according to the online class picture contents, so that students can conveniently and quickly develop the extended learning when watching the online class.

Drawings

Fig. 1 is a flowchart of a method for content interaction based on web lesson video according to embodiment 1 of the present invention.

Fig. 2 is a flowchart of a method for performing content interaction based on a web lesson video according to embodiment 2 of the present invention.

Fig. 3 is a block diagram of an apparatus for content interaction based on a web lesson video according to embodiment 3 of the present invention.

Fig. 4 is a block diagram of another device for content interaction based on web lesson video according to embodiment 3 of the present invention.

Fig. 5 is a diagram of a computer apparatus according to embodiment 3 of the present invention.

Detailed Description

The drawings are only for purposes of illustration and are not to be construed as limiting the invention. For a better understanding of the following embodiments, certain features of the drawings may be omitted, enlarged or reduced, and do not represent the size of an actual product; it will be understood by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.

Example 1

The embodiment provides a method for content interaction based on a web lesson video, which can be applied to equipment playing the web lesson video, or control equipment controlling equipment playing the web lesson video, or can be applied to equipment which does not play the web lesson video or even has no web lesson video playing function and can be matched with the equipment playing the web lesson video. The online lesson video is a video which can be watched by students and used for learning knowledge, and can be recorded in advance or live broadcast in real time. The teaching teachers and some things shown by the teaching teachers, such as courseware showing, teaching aids and the like, can appear in the online lesson videos, reflect the learning contents of students, and can show more learning materials related to the things by identifying the things, so that the students can further expand learning.

As shown in fig. 1, the method for performing content interaction based on a web lesson video according to the embodiment can identify things appearing in the web lesson video, and recommend extensible learning content in time according to the identification, and includes the following steps:

s101, collecting a network course picture when a network course video is played.

The web lesson pictures are static pictures displayed at one moment when the web lesson videos are played. The network course video is dynamically played, and in order to accurately analyze and identify things appearing in the network course video from the dynamic network course video, a static network course picture can be obtained in a network course picture acquisition mode.

Specifically, when the web lesson video is played, the web lesson picture played at the current moment of the web lesson video can be captured as the collected web lesson picture in a screen capture mode; and when the network course video is played, extracting data of a frame of network course video at the current moment from the data stream of the network course video, and processing the data into a frame of network course picture according to the extracted data. The collected web lesson pictures can be stored firstly.

S102, determining a picture range to be analyzed in the network course picture based on a preset range determination rule.

The online lesson pictures are collected from the online lesson videos in playing, and the online lesson videos can generate other contents besides objects capable of reflecting the learning contents of students, such as the faces of lecturers and containers for loading the objects, and the other contents can interfere the analysis accuracy of the objects capable of reflecting the learning contents of the students and even can prevent the objects from being analyzed and identified at all. Therefore, in the web lesson screen, the screen range to be analyzed needs to be determined according to the position where the object reflecting the learning content appears, so as to improve the accuracy of analysis.

The scope determination rule is a preset rule for determining a picture scope to be analyzed in a web lesson picture, and may be manually preset, may be a developer, a student who watches a web lesson video, or may be a parent or a teacher who guides the student to watch the web lesson video.

The screen range is pixel area information or coordinate area information indicating a certain partial area within the screen. For example, for a 400 × 600 pixel class screen, the screen range may be a pixel region having a pixel point (X, Y) in the pixel coordinate system as a midpoint, a length of X pixels, and a width of Y pixels, and the pixel region is a partial region of the class screen. For another example, for a single screen of 9cm × 16cm, the screen range may be a rectangular area formed by 4 points (x 1, y 1), (x 2, y 2), (x 3, y 3), and (x 4, y 4) in the length coordinate system, and the rectangular area may be a partial area of the screen.

Based on the preset range determination rule, the range of the picture to be analyzed determined in one web lesson picture can be one or more, and the ranges of a plurality of pictures determined in the same web lesson picture have no mutual influence and can be overlapped, crossed and the like.

The picture range needing to be analyzed is determined, so that the content to be analyzed can be narrowed from the whole network course picture to the determined picture range, the analysis is more pertinent, the finally made recommendation can more accurately grasp the user requirements, the user experience is improved, the analysis workload can be correspondingly reduced, and the efficiency of analyzing the recommendation is improved.

Specifically, there may be many embodiments how to determine the picture range to be analyzed in the web lesson picture based on the preset range determination rule.

In one embodiment, the preset range determination rule may be a correspondence between a plurality of preset video characteristics and a picture range. Correspondingly, the method for determining the picture range to be analyzed in the web lesson picture based on the preset range determination rule comprises the following steps: and determining a picture range which corresponds to a picture of the online lesson and needs to be analyzed according to the video characteristics of the online lesson video based on the corresponding relation between a plurality of preset video characteristics and the picture range.

The video characteristics are the specific properties of the web lesson video, a plurality of video characteristics can be preset according to the influence on the determined picture range, and the corresponding relation between the plurality of video characteristics and the picture range is preset. When the video characteristics of the web lesson video are determined, the picture range corresponding to the video characteristics can be determined according to the corresponding relation, namely, the picture range needing to be analyzed in the web lesson picture is determined. It will be appreciated that each video characteristic corresponds to at least one picture range.

Through the preset corresponding relation between the video characteristics and the picture range, the picture range to be analyzed of the web lesson picture can be conveniently and rapidly determined according to the video characteristics of different web lesson videos, the special properties of different web lesson videos can be adapted, the picture range can be determined more pertinently, the subsequent web lesson picture partial area intercepted according to the picture range can be used for more accurately positioning the things reflecting the learning content of students, and the recommendation can be more accurate.

The video characteristics can be a class course type or a class picture layout. The class type of the online lesson refers to a class to which the lesson of the online lesson video belongs, and the class can comprise a plurality of lesson subjects or a plurality of lesson subjects, such as languages, english, examination question explanation subjects, review subjects and the like. The course screen layout refers to arrangement planning of each displayed element on the course screen, such as the position and size of each element on the course screen.

The class type and the picture layout of the network classes are the characteristics that the video of the network classes is obvious, and the class type and the picture layout of the network classes can be easier and more accurate when the video characteristics of the network classes are determined. The type of the network course and the layout of the network course pictures have great influence on the network course picture partial area where the learning content of the students appears, and the network course picture partial area is used as a video characteristic to correspond to the range determination rule, so that the accurate determination of the picture range is facilitated.

For the same type of course (like a course subject, or the same course theme, etc.), in the course of playing the web course video, objects needing to be identified and analyzed basically appear in certain specific positions on the web course picture in a fixed size and in a fixed manner. Therefore, when the video characteristics are web lesson types, step S102 may be specifically as follows: presetting a plurality of course types corresponding to the picture range to be analyzed on the online course picture, and determining the picture range to be analyzed on the online course picture according to the course type to which the online course video belongs.

For example, a course with question explanation as a theme, that is, a web course type to which a web course video belongs is question explanation, and in the process of playing the web course video, a web course picture is often displayed in a form that a left half is a question and a right half is a question answer, so that a picture range corresponding to a course to be analyzed of the theme can be preset, and the picture range can be a picture range that the middle upper part of the left half of the web course picture occupies 2/3 of the size of the web course picture of the left half and a picture range that the upper part of the right half occupies 2/3 of the size of the web course picture of the right half.

For the network course picture layout, on one hand, the network course picture layout comprises the whole picture layout of the network course picture, the whole picture layout is formed by each element, for example, a network course video generally shows a course piece by a set position and a set size, and then shows the state expression of a lecturer by another set position and set size, and the course piece, the state expression of the lecturer and the like are the elements forming the whole picture layout. On the other hand, the network course picture layout also includes the internal picture layout of these elements, for example, the internal picture layout of courseware, for the courseware made by the same lecture teacher or the courseware of the same type of course, the picture layout of the courseware display is similar, and objects needing to be identified also can be fixed to some specific positions of the courseware picture with a fixed size. Therefore, when the video characteristic is the web lesson screen layout, step S102 may specifically be: presetting a plurality of network course picture layouts corresponding to picture ranges to be analyzed on network course pictures, and determining the picture ranges to be analyzed on the network course pictures according to the network course picture layouts to which network course videos belong.

For example, the layout of the web lesson picture to which a certain web lesson video belongs is as follows: the whole picture layout of the network course picture is that courseware is displayed on the whole network course picture, the state expression of a teacher is displayed at a corner, and the internal picture layout of the courseware is that objects needing to be learned by students are placed in the middle, so that the picture range corresponding to the picture layout needing to be analyzed can be preset to be the middle part of the network course picture and account for 1/2 of the picture range of the whole network course picture.

It can be understood that the preset several web lesson picture layouts correspond to the picture range to be analyzed in the web lesson picture, may correspond to several whole picture layouts, may correspond to several element internal picture layouts, and may correspond to several whole picture layouts and several element internal picture layouts.

It can also be understood that the same picture range can be uniformly preset for different web lesson types or different web lesson picture layouts.

In another embodiment, the frame range to be analyzed for the web lesson frames may be determined according to the analysis range circled by the user, where the user may be the student who watches the web lesson video, or may be a parent or a teacher who guides the student to watch the web lesson video. Therefore, interactivity between the network course video playing process and the user can be enhanced, and subsequent recommendation can be more suitable for the expanded learning requirements of the user. At this time, the preset range determination rule is specifically a rule for determining a picture range to be analyzed in the web lesson picture according to the analysis range circled by the user, that is, the preset range determination rule is a rule for changing the analysis range into the picture range. Thus, step 102 may be: acquiring an analysis range circled by a user in a web course picture; and determining the picture range to be analyzed on the web course picture according to the analysis range based on a preset range determination rule.

The shapes of the analysis ranges circled by the users are various, based on the hobbies of different users or different preset circled rules, rectangular frames or circular frames can be used for frame selection, irregular tracks of the circled selection can be drawn by hand, the shapes of the picture ranges are not uniform, the shapes of pictures generated by subsequently capturing partial areas of the web lesson pictures are not uniform, and the determination of the content types of the pictures in a uniform method is not facilitated, so that the preset range determination rule can be a preset shape, and the determined picture ranges can be unified into the preset shape no matter what the shapes of the analysis ranges circled by the users are. On the basis of the uniform shape, whether the determined picture range covers the analysis range circled by the user or not is also considered, and the situation that certain content circled by the user cannot be covered in the picture range and is intercepted to generate a picture is prevented, so that the recommendation does not completely meet the requirements of the user.

Accordingly, step S102 may be more specifically: and acquiring an analysis range circled by the user in the web lesson picture, and taking the minimum picture which takes the shape as the preset shape and comprises the analysis range as a picture range needing to be analyzed in the web lesson picture based on the preset shape.

For example, if the preset shape is a rectangle, the analysis range circled by the user on the screen class picture is a circular range, a minimum rectangular picture range including the circular range is selected on the screen class picture, the circular range circled by the user is an inscribed circle of the minimum rectangular picture range, and the minimum rectangular picture range is used as a picture range to be analyzed, so that the analysis range circled by the user can be properly expanded, and the circled content of the user is ensured to be covered in the picture range.

It can be understood that, besides the preset shape and the standard shape of the picture range, the size and the standard size of the picture range can be preset, so that the sizes of the pictures generated by subsequently capturing the web lesson pictures according to the picture range have uniformity.

S103, according to the picture range, capturing partial areas of the web lesson picture, and generating pictures from the captured partial areas.

After the frame range needing to be analyzed is determined by the web lesson frame, a partial area of the web lesson frame in the frame range can be intercepted according to the determined frame range, and a picture can be generated for the extracted partial area. Specifically, after the picture range to be analyzed is determined, the pixel area information or the coordinate area information of a certain partial area in the web lesson picture indicated by the picture range to be analyzed is also determined, the corresponding partial area is cut out from the web lesson picture according to the pixel area information or the coordinate area information, and the cut-out partial area is generated into the picture.

S104, determining the content type of the picture according to the content of the picture, determining recommended content according to the content type, and displaying the recommended content when the network course video is played.

After the pictures are generated by capturing the online lesson pictures, the contents of the pictures can be analyzed to obtain the content category so as to determine the recommended contents which need to be displayed when the online lesson videos are played. The content of a picture refers to various effective information contained in the picture. The recommended content refers to various information recommended to students who are watching the web lesson video or links to the information.

After the picture is generated and before the content category of the picture is determined according to the content of the picture, the picture can be subjected to some pre-processing. The purpose of the picture preprocessing can be to unify the size, brightness, and the like of the picture, so that the feature information of the picture can be uniformly analyzed later.

For the size of the unified picture, taking the picture range as a rectangle as an example, the extracted portion of the screen is rectangular according to the rectangular picture range, and the extracted portion of the rectangle is converted to a specified size M × N, where the specified size may be a specified resolution (in this case, the unit of M and N is a pixel) or a specified physical size (in this case, the unit of M and N is a length unit). The size transformation mode can adopt equal-scale transformation, and the insufficient part after transformation is filled with 0 or 1 pixel. The advantage of using such a size conversion method is that the content of the picture is not deformed, but there are other methods for picture size conversion, such as stretching, scaling, cropping, etc., and the size conversion method of 0 or 1 pixel is just an example, which is not limited in this embodiment.

The content category of the picture is determined according to the content of the picture, and different classification methods can be specifically adopted.

The classification method can input a picture into a multi-label classification model, respectively judge the possibility of whether the picture contains each content type according to the content of the picture, and output the probability of whether the picture contains each content type, when the probability is greater than a preset probability threshold, the picture is judged to contain the content type corresponding to the probability, and the content type corresponding to the probability is used as the content type of the picture.

The multi-label classification model can adopt a conventional network model, such as a typical residual neural network resnet50, resnet101 and the like, and can also adopt a network model formed by combining different deep learning chunks. A loss function of multi-label classification is constructed by adopting the idea of Binary Relevance Bilery Relevance, and a trained multi-label classification model can be obtained by training labeled sample data.

The idea of Binary Relevance Bilery Relevance is to decompose a multi-label learning problem into a plurality of independent two-classification problems, wherein each two-classification problem corresponds to a content category in a category space. For a sample x, binary Relevance Relevance computes each content class y by using respective Binary classifiers _j Class correlation of prediction sample x, class set of prediction sample x:

Y＝{y _j |g _j (x)＞0，1≤j≤q}

specifically, the content category y of the multi-label classification model is preset _j Total q content categories, j =1, 2. Q independent classifiers g are formed in the multi-label classification model _j (x) And training the multi-label classification model through the sample data labeled with the belonged content category. After the pictures are input into the trained multi-label classification model, each secondary classifier g _j (x) Can be used forTo correspondingly predict whether the picture contains the content type y _j Thereby obtaining that the picture belongs to the content category y _j Probability of (p) _j J is more than or equal to 1 and less than or equal to q, and the probability p _j I.e. representing picture containing content category y _j The probability of (c).

The probability threshold p can be preset _{Threshold value} If probability p _j Greater than a predetermined probability threshold p _{Threshold value} Then the inputted picture is considered to contain the content category y _j If there is not a probability p _j Greater than a predetermined probability threshold p _{Threshold value} Then the inputted picture is considered not to contain any preset content category y _j 。

Taking the web lesson video of the cognitive life object course as an example, 3 content categories of the multi-label classification model can be preset: fruits, vehicles, animals; the multi-label classification model has 3 independent classifiers formed: a fruit class II classifier, a vehicle class II classifier and an animal class II classifier.

After the pictures are input into the trained multi-label classification model, the fruit class two classifier can predict whether the pictures contain fruits or not to obtain the probability p of containing the fruits _Fruit Probability of not containing fruit is 1-p _Fruit (ii) a The vehicle class classifier predicts whether the picture contains the vehicle or not to obtain the probability p of containing the vehicle _{Transportation means} Probability 1-p of not containing a vehicle _{Transportation means} (ii) a The animal classifier predicts whether the picture contains animals or not to obtain the probability p of containing the animals _{Animal(s) production} Probability of not containing animal is 1-p _{Animal(s) production} 。

Preset probability threshold p _{Threshold value} =80%. Hypothesis probability p _Fruit =85%, probability p _{Transportation means} =90%, probability p _{Animal(s) production} =13%, then there is p _Fruit ＞p _{Threshold value} 、p _{Transportation means} ＞p _{Threshold value} The inputted picture may be considered to contain fruits and vehicles, and at this time, the content category of the picture may be determined to be fruits and vehicles. Assume again the probability p _Fruit =60%, probability p _{Transportation means} =55% probabilityp _{Animal(s) production} =21%, then p is _Fruit 、p _{Transportation means} 、p _{Animal(s) production} Are not more than p _{Threshold value} The inputted picture may be considered to contain no fruits, vehicles and animals, and the content category of the picture may be determined to be empty at this time.

When the content category of the picture is empty, the recommended content determined according to the content category of the picture can also be empty, and the empty recommended content is displayed or not displayed during the online course video playing.

Because the web lesson picture usually contains one or more content types, compared with other conventional classification models, the multi-label classification model can more completely obtain the content types contained in the pictures and can also deal with the situation that the input pictures do not contain any content types.

Another classification method may be to input the pictures into the multi-classification model, determine the probability that the pictures belong to each content category according to the content of the pictures, output the probability that the pictures belong to each content category, and use the content category corresponding to the maximum probability as the content category of the pictures.

The multi-classification model can adopt a conventional network model, such as a typical residual neural network resnet50, resnet101 and the like, and can also adopt a network model formed by combining different deep learning chunks. Training is carried out through the labeled sample data, and a trained multi-label classification model can be obtained.

Specifically, q-1 content categories y are preset _j J =1, 2., q-1, qth content category

Not belonging to q-1 content categories y _j The multi-label classification model is trained by the sample data labeled with the content category to which the sample data belongs. Inputting the pictures into the trained multi-label classification model to obtain pictures respectively belonging to each content category y _j And &>

Probability of (p) _j J is more than or equal to 1 and less than or equal to q } according to the probability p _j Determine that the picture belongs to q-1 content categories y _j Which one or ones of, or not, belong to the q-1 content categories y _j Any one of (a) and (b). At probability p _q When not the maximum, choose the probability p _j Maximum one or several corresponding content categories y _j As a content category of pictures; at probability p _q At maximum, the content category of the picture is not q-1 content categories y _j Any one of the above.

Similarly, taking the web lesson video of the cognitive living thing course as an example, 4 content categories of the multi-classification model can be preset: fruits, vehicles, animals, and others (not belonging to any of fruits, vehicles, and animals).

After the pictures belong to the trained multi-classification model, the probability p that the pictures respectively belong to fruits, vehicles, animals and other classes can be obtained _Fruit 、p _{Transportation means} 、p _{Animal(s) production} 、p _{Others are} Wherein p is _Fruit +p _{Transportation means} +p _{Animal(s) production} +p _{Others are} =1, then can be according to p _Fruit 、p _{Transportation means} 、p _{Animal(s) production} 、p _Others The relative sizes between these four determine the content category of the picture.

Hypothesis probability p _Fruit ＝56％、p _{Transportation means} ＝34％、p _{Animal(s) production} ＝7％、p _{Others are} =3%, maximum is p _Fruit The next largest is p _{Transportation means} Then it can be determined that the content category of the picture is fruit or fruit and vehicle. Assume again the probability p _Fruit ＝10％、p _{Transportation means} ＝8％、p _{Animal(s) production} ＝22％、p _{Others are} =60%, maximum is p _{Others are} Then the content category of the picture may be determined to be other categories.

Similarly, when the content category of the picture is other categories, the recommended content determined according to the content category of the picture may be empty, and the empty recommended content is displayed or not displayed during the course video playing.

By adopting the multi-classification model, the content classification of the picture can be obtained more accurately, the classification accuracy is higher under the same training data volume, and the situation that the input picture does not belong to any content classification can be solved. And a multi-classification model is adopted, so that the method is suitable for the requirement of recommending only according to one content category in one picture generated from a web lesson picture.

After the content category of the picture is obtained, a two-step recommendation strategy can be adopted: the recommendation category is determined first, and then the recommendation content is determined, so that the recommendation accuracy can be improved.

Determining recommended content according to the content category, specifically comprising: determining a recommendation category according to the content category, the recommendation category comprising one or more elements; one or more elements are selected from the recommendation category as recommendation content.

More specifically, the recommendation category is determined according to the content category, which may be: when the output result of the multi-tag classification model is that the picture contains one or more content categories, it is determined that the recommended category is all or part of the content categories contained in the picture, and when the recommended category is only part of the content categories contained in the picture, the content categories contained in the picture need to be filtered, and the filtering may be performed according to the probability that the picture contains the content categories, or according to other factors, which is not limited in this embodiment. And when the output result of the multi-label classification model is that the picture does not contain any content category, determining that the recommended content is empty and not recommending.

In a specific implementation process, when the multi-label classification model is constructed, the content category may be set according to the recommendation category. In some practical situations, the content category may correspond to the recommendation category one to one, that is, the content category and the recommendation category are the same.

There are various embodiments for selecting one or more elements from the recommendation category as the recommended content.

In one embodiment, one or more elements are randomly selected from the recommendation category as the recommended content. If there are multiple determined recommendation categories, one element may be randomly selected for each recommendation category as the recommendation content.

In another embodiment, feature information of the element and the picture may be extracted separately, and the feature information includes one or more of the following: color, shape; and selecting one or more elements from the recommendation category as recommendation content according to the matching degree of the characteristic information of the elements and the pictures.

It will be appreciated that in such an embodiment, the elements are in the data format of the picture or convertible to the data format of the picture, from which the characteristic information can be extracted and matched to the characteristic information of the picture.

As described above, the feature information of a picture refers to feature information that can represent the content in the picture, and is, for example, a color, a shape, or a combination of both. Correspondingly, the feature information of an element also refers to feature information that can represent the content of the element.

The color may include the type of color contained, the number of types of color, and the like, and the shape may include the type of shape contained, the number of types of shape, and the like.

Extracting feature information of elements or pictures, specifically adopting the following mode: extracting RGB information of each pixel of the element or the picture so as to obtain the color of the element or the picture; the outline of each object in the picture is extracted by utilizing an edge detection algorithm, the edge detection algorithm is based on the principle that discontinuous pixels with large changes are regarded as edges, and OpenCV (open source computer vision and machine learning software library) provides a plurality of picture edge detection functions, such as Laplacian (), sobel (), scharr (), and the like.

The matching degree between the characteristic information of the element and the characteristic information of the picture can be determined by adopting the following modes:

and training a prediction network related to the characteristic information, and inputting the elements and the pictures into the trained prediction network respectively to predict the characteristic information of the elements and the characteristic information of the pictures correspondingly. The characteristic information can be recorded in the form of an information matrix, thereby forming a characteristic information matrix of the elements i of the recommendation category j

Characteristic information matrix v of picture _x Two characteristic information matrices are calculated according to the following formula>

And v _x Similarity of (2):

the above formula, i =1, 2.,. M, m is the total number of elements included in the recommendation category j.

The larger the matching degree between the feature information of the indicating element and the feature information of the picture is, can select>

The largest element or elements serve as recommended content.

The recommended content is determined according to the matching degree of the element and the feature information of the picture, so that the relevance between the recommended content and the picture is higher, the recommendation is more accurate, and the user experience can be improved. For example, if the picture shows an orange, then when the recommended category is a fruit category, then there is a greater preference for choosing the recommended element, than for a banana, for the fruit category, because oranges and oranges have a more similar color and shape than oranges and bananas.

The recommended content is displayed during the online lesson video playing, and specifically, the recommended content can be displayed at a certain fixed position of the online lesson video in a message frame mode according to a random sequence. The determined recommended content corresponding to each recommendation category can be displayed for a fixed time, and different recommendation categories can be displayed in a circulating manner. When a certain recommended content is selected by a user, the recommended content can be aligned to a corresponding content link to display the detailed information of the recommended content; and after returning to continuously playing the online lesson video, continuously displaying the rest other recommended contents which are not selected by the user.

Example 2

Based on the same inventive concept as that of embodiment 1, as shown in fig. 2, this embodiment provides a method for content interaction based on a web lesson video, which is a further optimization of embodiment 1, and specifically includes:

s201, when the network course video is played, network course pictures are collected for many times at a fixed speed, and one network course picture displayed at the current moment is collected every time.

Specifically, the network course pictures are collected for a plurality of times at a fixed rate along the playing time axis of the network course video, and the collection rate of the network course pictures is fixed when one network course video is played. Taking the web lesson video with the time length of 20 minutes as an example, along the playing time axis of the web lesson video, the web lesson pictures can be collected for a plurality of times at a fixed rate of 5 minutes/time, and then the web lesson pictures can be collected for 4 times on the playing time axis of the whole web lesson video.

The fixed rate for collecting the lesson-screening pictures can be a fixed value preset by people, or a fixed value determined according to a rate rule preset by people, and the rate rule can be a rule related to the duration of lesson-screening videos, such as: the fixed rate is in negative correlation with the duration of the network course video, and the value of the fixed rate is increased by 1 minute/time every 5 minutes.

Subsequent steps based on the collected web lesson pictures, requirements on the processing efficiency of the device, and limitations on the hardware performance of the device may be considered when the fixed rate is preset.

The collected multiple web lesson pictures can be stored in the temporary space in sequence and stored in a picture format or a format capable of being converted into pictures, so that the web lesson pictures can be conveniently read from the temporary space in sequence for subsequent picture generation.

The network course pictures are collected at a fixed speed, and once every time a network course picture is collected, one network course picture is subjected to picture generation and content recommendation, so that objects which can be analyzed and identified can be captured from network course videos in real time and continuously, and therefore real-time and continuous content recommendation is carried out.

S202, determining a picture range to be analyzed on a network course picture based on a preset range determination rule.

The picture range needing to be analyzed is determined, the content needing to be analyzed can be reduced from the whole web class picture to the determined picture range, the analysis is more targeted, the finally displayed recommended content can grasp the user requirements more accurately, the calculation amount can be reduced under the condition that the web class pictures are collected for many times, the calculation rate of analysis recommendation is improved, the collection rate of the web class pictures can be increased, and the accuracy of analysis recommendation is improved.

S203, according to the picture range, capturing partial areas of the web course pictures, generating pictures of the captured partial areas, and generating a picture corresponding to the partial area of the web course picture in one picture range for one web course picture.

Step S203 is executed each time the collected web lesson picture is acquired, a partial area of the web lesson picture is captured according to a determined picture range, and the captured partial area is generated into a picture, that is, a web lesson picture and a picture range are processed correspondingly to generate a picture.

And a web lesson picture can determine a plurality of picture ranges to be analyzed, that is, one web lesson picture can be correspondingly processed to generate a plurality of pictures, and different pictures generated by one web lesson picture correspond to different picture ranges.

And S204, judging whether the currently generated picture is the first picture generated corresponding to one picture range, if so, executing the steps S208 to S212, and if not, executing the steps S205 to S206.

Step S204 is executed once for the generated picture corresponding to each frame range of the collected different web lesson pictures, and whether the generated picture is the first picture generated corresponding to the frame range is determined. The first picture generated corresponding to the picture range refers to a picture generated corresponding to the picture range for the first time, that is, a partial area of a first-acquired web class picture is captured according to the picture range, and the first picture is generated according to the captured partial area.

S205, difference detection is carried out on the currently generated picture and the picture which is generated last time and corresponds to the same picture range, and difference degree is obtained.

And when the currently generated picture is judged not to be the first picture generated corresponding to a certain picture range, the currently generated picture is not the picture generated corresponding to the picture range for the first time, the picture generated corresponding to the picture range and generated at the last time exists, and the difference detection is carried out on the currently generated picture and the picture generated at the last time and corresponding to the picture range to obtain the difference degree. Because the currently generated picture is not the first picture generated corresponding to the picture range, and the content analysis and the subsequent determination of recommended content are already performed on the picture corresponding to the picture range before, the difference degree between the previously generated picture and the subsequently generated picture can be detected, when the difference degree is smaller, the content type determination of the currently generated picture is not performed, and when the difference degree is larger, the content type determination of the currently generated picture is performed, so that unnecessary repeated work is avoided, and the calculation resources are not wasted.

In one embodiment, the degree of difference may be calculated as follows: respectively extracting information from the currently generated picture and the last generated picture to form information matrixes Img1 and Img2, and calculating the difference degree epsilon by using the following formula:

ε＝Mean（Img1-Img2）；

in the above formula, mean () is a function of the average value of each element in the matrix.

In another embodiment, the degree of difference may be judged according to the size, position, number, etc. of the areas having difference between the currently generated picture and the last generated picture.

The degree of difference can be evaluated quantitatively or qualitatively.

In a similar manner, the similarity detection may be performed on the currently generated picture and the picture generated last time to obtain the similarity degree, and when the similarity degree is high, the content type of the currently generated picture is not determined, and when the similarity degree is low, the content type of the currently generated picture is determined. The similarity degree and the difference degree can be used for judging the similarity of the generated pictures before and after the two times.

S206, determining whether the content of the currently generated picture is determined according to the content of the currently generated picture according to the difference degree, if so, executing steps S208 to S211, and if not, executing step S207.

A condition on the degree of difference may be set so as to determine whether or not the content category of the currently generated picture is to be determined from the content of the currently generated picture.

When the difference is a quantitative determination, a difference threshold is preset, and it is determined whether the difference is smaller than the preset difference threshold, if so, steps S208 to S212 are executed, otherwise, step S207 is executed.

When the difference is a qualitative judgment, a preferred embodiment is to preset a judgment index, and execute steps S208 to S212 when the judgment index is reached, and execute step S207 when the judgment index is reached.

And S207, displaying the recommended content correspondingly determined by the previously collected web lesson pictures when the web lesson video is played.

Step S207 is executed when the difference between the currently generated picture and the picture corresponding to the same picture range that was generated last time is small, that is, there is a picture generated corresponding to the picture range that was generated last time, which also means that the web lesson picture acquired this time is not acquired for the first time, that is, there is a web lesson picture acquired this time before. When the content category is not determined according to the content of the currently generated picture, the previously acquired web lesson picture can be displayed corresponding to the determined recommended content when the web lesson video is played.

It can be understood that the web lesson pictures collected before this time can be the last time. When the content category is not determined according to the content of the generated picture for the last captured web lesson picture, the web lesson pictures captured before the current time can be captured several times.

The recommended content corresponding to the web lesson pictures acquired before is directly adopted for displaying, so that the recommended content display interruption caused by too small difference degree of pictures generated by the web lesson pictures acquired twice before and after in the web lesson video playing process can be avoided, and the proper recommended content can be continuously provided. In addition, the determined recommended content before multiplexing can reduce the determination time of the recommended content, accelerate the recommendation speed and improve the experience fluency of the recommendation service.

It is understood that the previously determined recommended content may be cached to reuse the previously determined recommended content.

And S208, determining the content type of the currently generated picture according to the content of the currently generated picture.

When the currently generated picture is judged to be the first picture generated corresponding to a certain picture range, step S208 may be executed to determine the content category of the currently generated picture according to the content of the currently generated picture. Because the currently generated picture is the first picture generated corresponding to the picture range, content analysis and subsequent determination of recommended content are not performed on the picture corresponding to the picture range before, the content category of the currently generated picture needs to be determined according to the content of the currently generated picture.

When it is determined that the difference between the pictures corresponding to the same picture range generated this time and the pictures generated last time is large, step S208 may also be executed to determine the content type of the currently generated picture according to the content of the currently generated picture. Since two pictures have a large difference and cannot be multiplexed with previously determined recommended contents, it is necessary to perform content analysis on the currently generated picture and determination of the subsequently recommended contents, and determine the content type of the currently generated picture from the content of the currently generated picture.

Specifically, step S208 may include:

inputting the currently generated picture into a multi-label classification model, respectively judging the possibility of whether the picture contains each content type, outputting the probability of whether the picture contains each content type, and taking the content type corresponding to the probability as the content type of the currently generated picture when the probability is greater than a preset probability threshold; or inputting the currently generated picture into the multi-classification model, judging the possibility that the picture belongs to each content category respectively, outputting the probability that the picture belongs to each content category respectively, and taking the content category corresponding to the maximum probability as the content category of the currently generated picture.

S209, determining a recommendation category according to the content category, wherein the recommendation category comprises one or more elements.

Before the recommended content needing to be displayed is determined, the recommendation category is determined, and the recommendation category is formed by classifying the recommended content, so that the finally determined recommended content can be more accurate.

A recommendation category may have only one element or may have multiple elements.

S210, respectively extracting feature information of the elements and the pictures, wherein the feature information comprises one or more of the following: color, shape.

After the recommendation category is determined, which element in the recommendation category is needed to be determined as the final recommendation content, and the element and the picture can be matched more easily and more accurately through the feature information of the element and the picture which are respectively extracted.

S211, selecting one or more elements from the recommendation category as recommendation contents according to the matching degree of the element and the feature information of the picture.

Specifically, all elements whose matching degrees meet the preset matching conditions may be selected from the recommendation categories as the recommended content, or an element whose matching degree best meets the preset matching conditions may be selected as the recommended content.

The matching degree can be obtained by comprehensively evaluating the matching conditions between colors and the matching conditions between shapes. The preset matching condition may be a preset threshold value of the matching degree, a range of the matching degree, or several logic conditions related to the matching condition of the color and the shape.

S212, the recommended content is displayed when the network course video is played.

The determined recommended content can be displayed while the online lesson video is played, real-time recommendation is achieved, a user can see the recommended content at any time when the online lesson video is played, and even the displayed recommended content can directly enter a link corresponding to the recommended content.

Example 3

Based on the same inventive concept as that in embodiments 1 and 2, as shown in fig. 3, this embodiment provides an apparatus for performing content interaction based on a web lesson video, which can be applied to a device that is playing a web lesson video, or a control device that controls a device that is playing a web lesson video, or a device that is not playing a web lesson video or even has no function of playing a web lesson video but can cooperate with a device that is playing a web lesson video, and the apparatus includes:

the picture acquisition module 301 is configured to acquire a currently played web lesson picture when a web lesson video is played, where the web lesson picture is a static picture displayed at a moment when the web lesson video is played;

a range determining module 302, configured to determine, based on a preset range determining rule, a picture range to be analyzed on a web lesson picture;

the picture generation module 303 is configured to capture a partial area of the web lesson picture according to the picture range, and generate a picture from the captured partial area;

and the recommendation display module 304 is configured to determine the content category of the picture according to the content of the picture, determine the recommendation content according to the content category, and display the recommendation content when the web lesson video is played.

The scope determining module 302 may be implemented in various ways, particularly how to determine the scope of the picture to be analyzed in the web lesson picture based on the preset scope determining rule.

In one embodiment, the range determining module 302 is specifically configured to: and determining a picture range which corresponds to the online lesson picture and needs to be analyzed according to the video characteristics of the online lesson video based on the corresponding relation between the preset video characteristics and the picture range.

The video characteristics can be a class course type or a class picture layout. The class type of the online lesson is a class to which the lesson of the online lesson video belongs, and the class can comprise a plurality of lesson subjects or a plurality of lesson subjects, such as junior middle school Chinese, high school English, examination question explanation subjects, review subjects, and the like. The course screen layout refers to arrangement planning of each displayed element on the course screen, such as the position and size of each element on the course screen.

For the same type of course (like a course subject, or the same course theme, etc.), in the course of playing the web course video, objects needing to be identified and analyzed basically appear in certain specific positions on the web course picture in a fixed size and in a fixed manner. Therefore, when the video characteristics are the class types, the range determining module 302 may be specifically configured to preset a picture range to be analyzed on the class picture corresponding to a plurality of class types, and determine the picture range to be analyzed on the class picture according to the class type to which the class video belongs.

For the web lesson picture layout, the picture layout can comprise the whole picture layout of the web lesson picture and the internal picture layout of each element forming the whole picture layout. Therefore, when the video characteristics are the screen layout of the web lesson, the range determining module 302 may be specifically configured to preset a plurality of screen layouts corresponding to the screen ranges to be analyzed on the screen of the web lesson, and determine the screen ranges to be analyzed on the screen of the web lesson according to the screen layouts to which the video of the web lesson belongs.

In another embodiment, the picture range to be analyzed for the web lesson pictures can be determined according to the analysis range circled by the user, so that the interactivity between the web lesson video playing process and the user can be enhanced, and the subsequent analysis results can better meet the extended learning requirements of the user. At this time, the range determining module 302 may be specifically configured to obtain an analysis range circled by the user on the web lesson picture, and determine a picture range to be analyzed on the web lesson picture according to a preset range determining rule and the analysis range.

More specifically, the range determining module 302 may be specifically configured to obtain an analysis range circled by the user in the web lesson screen, and based on the preset shape, the minimum screen with the shape as the preset shape and including the analysis range in the web lesson screen is used as the screen range to be analyzed. The user can be a student who watches the network course video, and can also be a parent or a teacher and the like which guide the student to watch the network course video.

After the recommendation presentation module 304 obtains the content category of the picture, two recommendation strategies may be adopted: the recommendation category is determined first, and then the recommendation content is determined, so that the recommendation accuracy can be improved. Accordingly, at this time, the recommendation presentation module 304 may specifically include a content category determination unit 341, a recommendation category determination unit 342, a recommendation content determination unit 343, and a recommendation content presentation unit 344.

A content type determining unit 341, configured to determine a content type of the picture according to the content of the picture;

a recommendation category determining unit 342 for determining a recommendation category according to the content category, the recommendation category comprising one or more elements;

a recommended content determining unit 343 configured to select one or more elements from the recommended category as recommended content;

and a recommended content presentation unit 344, configured to present the recommended content during the course video playing.

The content category determining unit 341 may specifically be configured to: the method comprises the steps of inputting a picture into a multi-label classification model, respectively judging whether the picture contains the possibility of each content category, outputting the probability corresponding to each content category, and when the probability is larger than a preset probability threshold value, taking the content category corresponding to the probability as the content category of the picture, or inputting the picture into the multi-label classification model, judging the possibility that the picture belongs to each content category, outputting the probability that the picture belongs to each content category, and taking the content category corresponding to the maximum probability as the content category of the picture.

The recommended content determining unit 343 may be specifically configured to: respectively extracting feature information of the elements and the pictures, wherein the feature information comprises one or more of the following: color, shape; and selecting one or more elements from the recommendation category as recommendation content according to the matching degree of the characteristic information of the elements and the pictures.

In order to capture objects capable of being analyzed and identified from the web lesson video in real time and continuously in the process of playing the web lesson video, so as to perform content recommendation in real time and continuously, the picture acquisition module 301 is specifically configured to acquire a currently played web lesson picture for multiple times at a fixed rate during the playing of the web lesson video, and acquire one web lesson picture displayed at the current moment each time.

The picture generating module 303 captures a partial area of the web lesson picture according to the picture range determined by the range determining module 302 for each captured web lesson picture, and generates a picture from the captured partial area. For a web course picture, a picture range is corresponding to a part of area of the captured web course picture to generate a picture.

If the difference degree of two pictures generated by corresponding to the same picture range of two web lesson pictures acquired twice before and after is very small, repeated analysis is not needed, and the recommended content determined by the analysis of the previous picture is directly multiplexed, so that the analysis time is shortened, and the speed and the experience fluency are accelerated. Preferably, as shown in fig. 4, the apparatus provided in this embodiment may further include a difference detection module 305.

A difference detection module 305, configured to perform difference detection on the currently generated picture and a picture corresponding to the same picture range that is generated last time if the currently generated picture is not the first picture that is generated corresponding to the picture range, so as to obtain a difference degree; determining whether the content type determining unit 41 determines the content type of the currently generated picture according to the content of the currently generated picture according to the difference degree;

the content type determining unit 341 is specifically configured to determine, if the currently generated picture is the first picture generated corresponding to the picture range, or if the difference detecting module 305 determines that the content type of the currently generated picture is determined according to the content of the currently generated picture, the content type of the currently generated picture is determined according to the content of the currently generated picture.

A recommended content display unit 344, configured to, for the currently acquired web lesson picture, determine, according to the content of the generated picture, a recommended content corresponding to the web lesson picture if the content category of the generated picture is determined, and display the recommended content when the web lesson video is played; if the content type of the generated picture is not determined according to the content of the generated picture, displaying the recommended content corresponding to the determined web course picture acquired before the current time when the web course video is played.

The implementation principle and technical effect of the device for performing content interaction based on the web lesson video provided in this embodiment are similar to those of the method embodiment, and are not described herein again.

The present embodiment also provides a computer device, which may be configured as shown in fig. 5, and includes a processor 410 and a memory, for example. The memory includes a nonvolatile storage medium 420, an internal memory 430. The non-volatile storage medium 420 stores an operating system 421 and computer programs 422. The internal memory 430 provides an environment for the operation of an operating system 421 and computer programs 422 in the non-volatile storage medium 420.

The processor 410 is used for providing computing and control capability, and the computer program 422 when executed by the processor 410 implements the method for content interaction based on web lesson videos according to embodiment 1 or embodiment 2. The processor 410 may communicate with external devices via the I/O interface 440.

The external device may include a display 510, and the display 510 may be a liquid crystal display or an electronic ink display. The processor 410 is connected to the display 510 through the I/O interface 440, and when the content interaction method is implemented, the processor 410 may send the determined recommended content to the display 510 through the I/O interface 440 for displaying.

The external device may also include a playback device 520. When the content interaction method is implemented, the playing device 520 is used for playing a web lesson video, and when the web lesson video is played, the collected web lesson pictures are sent to the processor 410 through the I/O interface 440.

The display 510 may be used to display the web lesson video played by the playing device 520 and the recommended content determined when the computer device implements the content interaction method, or may be used to display the recommended content determined when the computer device implements the content interaction method, and the web lesson video is displayed on another display configured by the playing device 520.

The external device may further include an input device 530, where the input device 530 may be a touch layer covered on the display 510, a key, a track ball, or a touch pad disposed on a housing of the computer device, or an external keyboard, a touch pad, or a mouse. When the content interaction method is realized, the input of a user can be received, and the interaction with the user is realized.

The implementation principle and technical effects of the computer device provided in this embodiment are similar to those of the method embodiment, and are not described herein again.

The present embodiment also provides a computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method according to embodiment 1 or embodiment 2.

The computer-readable storage medium provided in this embodiment has similar implementation principles and technical effects to those of the method embodiment, and thus, the implementation principles and the technical effects are not described herein again.

It should be understood that the above-mentioned embodiments of the present invention are only examples for clearly illustrating the technical solutions of the present invention, and are not intended to limit the specific embodiments of the present invention. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention claims should be included in the protection scope of the present invention claims.

Claims

1. A method for content interaction based on web lesson video is characterized by comprising the following steps:

2. The method of claim 1, wherein determining the range of pictures to be analyzed in the course picture based on a preset range determination rule comprises:

3. The method as claimed in claim 2, wherein the video characteristics are a class type or a class picture layout.

4. The method of claim 1, wherein determining the range of pictures to be analyzed in the course picture based on a preset range determination rule comprises:

acquiring an analysis range circled by a user in the network course picture;

5. The method as claimed in claim 4, wherein the determining a frame range to be analyzed in the course frame according to the analysis range based on a preset range determination rule comprises:

6. The method for content interaction based on the web lesson video according to any one of claims 1 to 5, wherein the step of collecting the web lesson pictures during the playing of the web lesson video comprises the following steps:

when the network course video is played, network course pictures are collected for many times at a fixed speed, and one network course picture displayed at the current moment is collected every time.

7. The method of claim 6, wherein determining the content category of the picture according to the content of the picture comprises:

for one network course picture, intercepting a partial area of the network course picture corresponding to one picture range to generate a picture;

and determining whether the content category of the currently generated picture is determined according to the content of the currently generated picture according to the difference degree.

8. The method of claim 7, wherein the determining recommended content according to the content category and displaying the recommended content during the video playback of the web lesson comprises:

for the collected web lesson pictures, if the content category of the generated pictures is determined according to the content of the generated pictures, determining recommended content corresponding to the web lesson pictures according to the content category, and displaying the recommended content when the web lesson video is played;

9. The method for performing content interaction based on web lesson videos as claimed in any one of claims 1 to 5, 7 and 8, wherein the determining the content category of the picture according to the content of the picture comprises:

10. The method for content interaction based on web lesson videos as claimed in any one of claims 1 to 5, 7 and 8, wherein the determining recommended content according to the content category comprises:

11. The method of claim 10, wherein selecting one or more elements from the recommendation category as recommended content comprises:

12. An apparatus for content interaction based on web lesson video, comprising:

the picture acquisition module is used for acquiring a web course picture when a web course video is played, wherein the web course picture is a static picture displayed at one moment when the web course video is played;

the picture generation module is used for intercepting a part of area of the web course picture according to the picture range and generating a picture from the intercepted part of area;

13. A computer device comprising a processor and a memory, the memory storing a computer program, wherein the processor when executing the computer program implements the method for web-based lesson video content interaction according to any one of claims 1 to 11.

14. A computer-readable storage medium, on which a computer program is stored, wherein the computer program, when being executed by a processor, implements the method for content interaction based on web lesson videos of any one of claims 1 to 11.