An artificial intelligence-based pre-control method and system for range hoods

Abstract

The present invention relates to the field of artificial intelligence, and discloses an artificial-intelligence-based pre-control method and system for a range hood. A cookware recognition model based on artificial intelligence and used for recognizing the type of cookware, a cooking action detection model used for detecting cooking actions, and a fume detection model used for detecting fumes are constructed. When the range hood is started, a cookware image is collected, the type of the cookware is recognized through the cookware recognition model, and the height of the air suction opening of the range hood is adjusted according to the different types of cookware. According to the amount of fumes that can be generated by different cooking actions, a corresponding air volume increment Voldelta is set for each cooking action. The cooking actions are periodically detected through the cooking action detection model. When a cooking action is detected, the air volume of the range hood is set. If no cooking action is detected after a time T from the start of detecting the cooking action, the fumes are detected through the fume detection model, the fumes are quantified, and the air volume of the range hood is controlled according to the amount of fumes.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and particularly to a method and system for pre-controlling an oil fume extractor based on artificial intelligence. Background Art

[0002] When removing oil fume during traditional cooking, it is necessary for a person to manually adjust the air volume of the oil fume extractor according to the size of the oil fume. Since the air volume cannot be adjusted in time, it will cause a person to inhale excessive oil fume, which will seriously affect health in the long term.

[0003] Currently, there is a technical solution that calculates cooking behavior based on the amplitude change of the cooking temperature curve, captures the moment of sudden change in the temperature of the stove area, realizes oil fume prediction, and thus adjusts the suction of the oil fume extractor to an appropriate gear in advance to suck in the oil fume in advance.

[0004] Some other oil fume extractors will adopt the method of oil fume recognition, collect and detect the oil fume image in real time, judge the size of the oil fume, and control the air volume of the oil fume extractor according to the size of the oil fume.

[0005] For the air volume control based on oil fume recognition, since it can only be detected after the oil fume rises, there is a certain delay between the rise of the oil fume and the adjustment of the air volume to an appropriate size; it is impossible to adjust the air volume of the oil fume extractor in advance, resulting in the lag of oil fume elimination. Summary of the Invention

[0006] In order to solve the above technical problems, the present invention provides a method and system for pre-controlling an oil fume extractor based on artificial intelligence. <X

[0007] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0008] An oil fume extractor pre-control method based on artificial intelligence, wherein the oil fume extractor has a suction port capable of lifting, includes the following steps:

[0009] Step 1: Construct a pan recognition model based on artificial intelligence for identifying the type of pan, a cooking action detection model for detecting cooking actions, and an oil fume detection model for detecting oil fume;

[0010] Step 2: When the oil fume extractor is started, collect a pan image, identify the type of the pan through the pan recognition model, and adjust the height of the suction port of the oil fume extractor according to the type of the pan: the greater the amount of oil fume generated by the pan, the lower the height of the suction port;

[0011] Step 3: Set the air volume increase value Vol corresponding to each cooking action according to the size of the oil fume that can be generated by different cooking actions; delta ; Periodically detect the cooking action through the cooking action detection model. When a cooking action is detected, set the air volume Vol of the oil fume extractor change= max((Vol curr + Vol delta ), Vol max ); Vol curr is the current air volume of the range hood, Vol delta is the air volume increase value, and Vol max is the maximum air volume of the range hood; where the cooking action refers to an action that will affect the amount of oil fume generated during cooking after the action is performed during the cooking process;

[0012] Step Four: If no cooking action is detected after the time T when starting to detect the cooking action, detect the oil fume through the oil fume detection model, quantify the oil fume, and control the air volume of the range hood according to the amount of oil fume.

[0013] Specifically, the cooking actions include opening the pot lid, and / or adding ingredients into the pot, and / or stirring the ingredients in the pot, and / or adding other raw materials except ingredients.

[0014] Specifically, when adjusting the height of the air suction opening of the range hood according to different types of pots, the pots include frying pans, frying pans, flat pans, and steamers, which are divided into the highest gear, medium gear, and lowest gear in ascending order of the height of the air suction opening of the range hood. When the pot is a frying pan, the height of the air suction opening is the highest gear; when the pot is a frying pan or a flat pan, the height of the air suction opening is the medium gear; when the pot is a steamer, the height of the air suction opening is the lowest gear.

[0015] Specifically, when setting the air volume increase value Vol delta corresponding to each cooking action according to the amount of oil fume generated by different cooking actions, the cooking actions include opening the pot lid, and / or adding ingredients into the pot, and / or stirring the ingredients in the pot, and / or adding water into the pot, and / or adding oil into the pot. The air volume increase values are in descending order of five gears, four gears, three gears, two gears, and one gear, and each gear corresponds to a specific set value; when the cooking action is opening the pot lid, the air volume increase value takes the value corresponding to the fifth gear; when the cooking action is adding ingredients into the pot, the air volume increase value takes the value corresponding to the fourth gear; when the cooking action is stirring the ingredients in the pot, the air volume increase value takes the value corresponding to the third gear; when the cooking action is adding water into the pot, the air volume increase value takes the value corresponding to the second gear; when the cooking action is adding oil into the pot, the air volume increase value takes the value corresponding to the first gear.

[0016] Specifically, in Step One, the steps of constructing the oil fume detection model are as follows:

[0017] Collect oil fume images generated during the cooking process;

[0018] Label the image oil fume images, divide the oil fume images into oil fume parts and smoke-free parts, and make an oil fume detection data set;

[0019] Input the oil fume detection data set into the artificial intelligence model for model training;

[0020] In step four, when quantifying the oil fume and controlling the air volume of the range hood according to the amount of oil fume, if there is an oil fume part in the real-time collected oil fume image, extract the dark channel image of the oil fume part, filter the pixel values pixel of the dark channel image, and the pixel values that meet the following conditions are identified as oil fume particles: pixel ∈ [50, 220]; then control the air volume change of the range hood to Vol change1 ;

[0021] Vol change1 = max(Vol, Vol curr );

[0022] where m is the number of oil fume particles.

[0023] A pre-control system for a range hood based on artificial intelligence, the range hood has an air suction port that can be lifted, including:

[0024] A model training module, which constructs a cookware recognition model based on artificial intelligence and used to identify the type of cookware, a cooking action detection model used to detect cooking actions, and an oil fume detection model used to detect oil fume;

[0025] A cookware recognition control module, which collects cookware images when the range hood is started, identifies the type of cookware through the cookware recognition model, and adjusts the height of the air suction port of the range hood according to the different types of cookware: the greater the amount of oil fume generated by the cookware, the lower the height of the air suction port;

[0026] A cooking action detection control module, which sets the air volume increase value Vol delta ; according to the size of the oil fume that can be generated by different cooking actions, and periodically detects the cooking actions through the cooking action detection model. When a cooking action is detected, set the air volume of the range hood Vol change = max((Vol curr + Vol delta ), Vol max ); Vol curr is the current air volume of the range hood, Vol delta is the air volume increase value, Vol max is the maximum air volume of the range hood; where the cooking action refers to an action that will affect the amount of oil fume generated during cooking when the action is performed during the cooking process;

[0027] The oil fume quantification control module, if no cooking action is detected after the time T when starting to detect the cooking action, detects the oil fume through the oil fume detection model, quantifies the oil fume, and controls the air volume of the range hood according to the amount of oil fume.

[0028] The pre-control system for the air volume of the range hood and the pre-control method for the air volume of the range hood in the present invention adopt corresponding technical solutions, and the preferred solutions proposed for the method are also fully applicable to the system.

[0029] Compared with the prior art, the beneficial technical effects of the present invention are:

[0030] The present invention combines pot recognition, cooking action detection, and oil fume detection to formulate a complete air volume regulation process for the range hood; the air suction position of the range hood in the present invention can be lifted and lowered, and the initial height is set through pot recognition, reasonably formulating the initial height of the air suction position of the range hood. The cooking action detection and oil fume detection are carried out alternately and complement each other, which can not only predict the change of oil fume and eliminate the time delay between the increase of oil fume and the increase of air volume, but also realize the timely adjustment of oil fume through the detection of oil fume when the cooking action is missed. Description of the Drawings

[0031] Figure 1 It is a flow chart of the working principle of the present invention. Detailed Embodiments

[0032] The following is a detailed description of a preferred embodiment of the present invention with reference to the drawings.

[0033] An artificial intelligence-based pre-control method for a range hood includes the following steps:

[0034] Step 1: Construct a pot recognition model based on artificial intelligence and used to identify the type of pot, a cooking action detection model used to detect cooking actions, and an oil fume detection model used to detect oil fume.

[0035] The pot recognition model, the cooking action detection model, and the oil fume detection model are all based on artificial intelligence models and are obtained through image collection, manual annotation, and training.

[0036] The training process of the pot recognition model includes: 1. Collect data: Collect common household pots on the market, divide them into different categories such as steamers, frying pans, and woks, and use common annotation software such as labelImg to annotate the data set; 2. Select a target detection model (such as faster-rcnn, yolov3, ssd, etc.) and train the model; 3. Adjust different parameters to obtain different models, and select a model with better performance from these models for deployment.

[0037] The training process of the cooking action detection model includes: 1. Collect cooking data: Collect data during the home cooking process, classify it into different cooking types such as adding ingredients and stir-frying, and use common annotation software such as labelImg to annotate the dataset; 2. Select a target detection model (such as faster-rcnn, yolov3, ssd, etc.) and train the model; 3. Adjust different parameters to obtain different models, and select a model with better performance from these models for deployment.

[0038] The training process of the oil fume detection model includes: 1. Collect oil fume images generated during cooking; 2. Use common annotation software such as label Img to annotate the oil fume images, divide the oil fume images into the oil fume part and the smoke-free part, and make an oil fume detection dataset; 3. Select a target detection model (such as faster-rcnn, yolov3, ssd, etc.), adjust different parameters to obtain different models, and select a model with better performance from these models for deployment.

[0039] Step 2: Collect images of the cookware when the range hood is started, identify the type of cookware through the cookware recognition model, and adjust the height of the air suction port of the range hood according to the different types of cookware.

[0040] The range hood has an air suction port that can be lifted. In this embodiment, the air suction port can be driven to lift by a servo drive mechanism. In other embodiments of the present invention, it is also possible to adopt a method of setting different air suction ports at different heights.

[0041] In this embodiment, the cookware includes a wok, a frying pan, a flat pan, and a steamer. The height of the air suction port is divided into the highest gear, the middle gear, and the lowest gear in ascending order according to the height of the air suction port of the range hood, that is, the air suction port height corresponding to the highest gear is the lowest, and the air suction port height corresponding to the lowest gear is the highest; because the oil fume generated during stir-frying in a wok is the largest, the oil fume generated by a steamer is the smallest, and the frying pan and the flat pan are between the two, so when the cookware is a wok, the height of the air suction port is the highest gear; when the cookware is a frying pan or a flat pan, the height of the air suction port is the middle gear; when the cookware is a steamer, the height of the air suction port is the lowest gear. In other embodiments of the present invention, other corresponding modes of cookware-air suction port positions can also be adopted, and are not limited to the above corresponding modes.

[0042] Step 3: Set the air volume increase value Vol corresponding to each cooking action according to the amount of oil fume that can be generated by different cooking actions delta ; Through the cooking action detection model, periodically detect the cooking action. When a cooking action is detected, set the air volume of the range hood Vol change = max((Vol curr + Vol delta ), Vol max ); Volcurr is the current air volume of the range hood, Vol delta is the increased value of the air volume, Vol max is the maximum air volume of the range hood; where the cooking action refers to an action that will affect the amount of cooking fumes generated after the action is performed during the cooking process.

[0043] In this embodiment, the cooking actions include opening the pot lid, adding ingredients into the cooking utensil, turning the ingredients in the cooking utensil, adding water into the cooking utensil, adding oil (or soy sauce) into the cooking utensil. The increased values of the air volume are in five levels, four levels, three levels, two levels, and one level in descending order of numerical values, and each level corresponds to a specific set value; when the cooking action is opening the pot lid, the increased value of the air volume takes the value corresponding to the fifth level; when the cooking action is adding ingredients into the cooking utensil, the increased value of the air volume takes the value corresponding to the fourth level; when the cooking action is turning the ingredients in the cooking utensil, the increased value of the air volume takes the value corresponding to the third level; when the cooking action is adding water into the cooking utensil, the increased value of the air volume takes the value corresponding to the second level; when the cooking action is adding oil into the cooking utensil, the increased value of the air volume takes the value corresponding to the first level.

[0044] The following table lists the levels of the increased values of the air volume corresponding to different cooking actions. The larger the level value, the larger the increased value of the air volume, and the increased value of the air volume can be set or calibrated as needed. In other embodiments of the present invention, different cooking actions, the corresponding relationships between different cooking actions and the increased values of the air volume, and the specific increased values of the air volume corresponding to different levels can also be set.

[0045] Cooking action Gear corresponding to the increased air volume value Lift the pot lid 5 Add ingredients 4 Stir-fry 3 Add water 2 Add oil, soy sauce, etc. 1

[0046] Step 4: If no cooking action is detected after the time T when starting to detect the cooking action, the range hood is used to detect the cooking fumes, and the cooking fumes are quantified, and the air volume of the range hood is controlled according to the amount of the cooking fumes. The detection of the cooking action and the detection of the cooking fumes are carried out alternately. The cooking action is detected every period of time (T1). If no cooking action is detected after detecting the cooking action for a period of time (T), the cooking fumes are detected; the next detection of the cooking action can be carried out after or simultaneously with the detection of the cooking fumes

[0047] The cooking fume image is collected in real time. If there is a cooking fume part in the collected cooking fume image in real time, the dark channel image of the cooking fume part is extracted, and the pixel values pixel of the dark channel image are filtered. The pixel values that meet the following conditions are recognized as cooking fume particles: pixel ∈ [50, 220]; then the air volume of the range hood is controlled to change to Vol change1 ;

[0048] Vol change1 = max(Vol, Vol curr )

[0049] Among them m is the number of oil fume particles.

[0050] An intelligent pre-control system for range hoods, comprising:

[0051] A model training module, which constructs a cookware recognition model based on artificial intelligence and used for recognizing cookware types, a cooking action detection model for detecting cooking actions, and an oil fume detection model for detecting oil fumes;

[0052] A cookware recognition control module, which collects cookware images when the range hood is started, recognizes the type of cookware through the cookware recognition model, and adjusts the height of the air suction port of the range hood according to different cookware types;

[0053] A cooking action detection control module, which sets the air volume increase value Vol corresponding to each cooking action according to the amount of oil fume generated by different cooking actions delta ; Through the cooking action detection model, the cooking action is periodically detected. When a cooking action is detected, the air volume of the range hood Vol is set change = max((Vol curr + Vol delta ), Vol max ); Vol curr is the current air volume of the range hood, Vol delta is the air volume increase value, and Vol max is the maximum air volume of the range hood; where the cooking action refers to an action that will affect the amount of oil fume generated during cooking after the action is performed during cooking;

[0054] An oil fume quantification control module, if no cooking action is detected after the time T when starting to detect the cooking action, the oil fume is detected through the oil fume detection model, quantified, and the air volume of the range hood is controlled according to the amount of oil fume.

[0055] The technical solutions adopted by the air volume pre-control system for range hoods and the air volume pre-control method in the present invention correspond to each other, and the preferred solutions proposed for the method are also fully applicable to the system.

[0056] As Figure 1 shown, the control method in the present invention runs on an AI development board.

[0057] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above-described exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention. Therefore, in any aspect, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be embraced within the present invention, and any reference signs in the claims should not be regarded as limiting the claims involved.

[0058] In addition, it should be understood that although this specification is described according to embodiments, not every embodiment only contains an independent technical solution. This narrative way of the specification is only for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. An artificial intelligence-based pre-control method for a range hood. The range hood has an air suction opening that can be lifted, and the method includes the following steps: Step 1: Construct a cookware recognition model based on artificial intelligence for identifying the type of cookware, a cooking action detection model for detecting cooking actions, and a fume detection model for detecting fumes; Step 2: When the range hood is started, collect an image of the cookware, identify the type of cookware through the cookware recognition model, and adjust the height of the air suction opening of the range hood according to the type of cookware: the greater the amount of fumes generated by the cookware, the lower the height of the air suction opening; among them, the cookware includes a wok, a frying pan, a flat pan, and a steamer, which are divided into the highest gear, the middle gear, and the lowest gear in ascending order of the height of the air suction opening of the range hood. When the cookware is a wok, the height of the air suction opening is the highest gear; when the cookware is a frying pan or a flat pan, the height of the air suction opening is the middle gear; when the cookware is a steamer, the height of the air suction opening is the lowest gear; Step 3: Set the increased air volume value corresponding to each cooking action according to the amount of oil fume generated by different cooking actions ; Periodically detect the cooking actions through the cooking action detection model. When a cooking action is detected, set the air volume of the range hood ; is the current air volume of the range hood, is the increased air volume value, is the maximum air volume of the range hood; where the cooking action refers to an action that will affect the amount of oil fume generated during cooking after the action is implemented during the cooking process; Step 4: If no cooking action is detected after the time T when starting to detect cooking actions, detect fumes through the fume detection model, quantify the fumes, and control the air volume of the range hood according to the amount of fumes.

2. The pre-control method of the range hood based on artificial intelligence according to claim 1, wherein: The cooking actions include opening the pot lid, and / or adding ingredients into the cookware, and / or stirring the ingredients in the cookware, and / or adding other raw materials except ingredients.

3. The pre-control method of the range hood based on artificial intelligence according to claim 1, wherein: Set the air volume increment value corresponding to each cooking action according to the amount of fume generated by different cooking actions When, the cooking actions include opening the pot lid, and / or adding ingredients into the cooking pot, and / or stirring the ingredients in the cooking pot, and / or adding water into the cooking pot, and / or adding oil into the cooking pot. The air volume increment values are, in descending order of magnitude, five gears, four gears, three gears, two gears, and one gear, and each gear corresponds to a specific set value; when the cooking action is opening the pot lid, the air volume increment value takes the value corresponding to the fifth gear; when the cooking action is adding ingredients into the cooking pot, the air volume increment value takes the value corresponding to the fourth gear; when the cooking action is stirring the ingredients in the cooking pot, the air volume increment value takes the value corresponding to the third gear; when the cooking action is adding water into the cooking pot, the air volume increment value takes the value corresponding to the second gear; when the cooking action is adding oil into the cooking pot, the air volume increment value takes the value corresponding to the first gear.

4. The pre-control method of the range hood based on artificial intelligence according to claim 1, characterized in that, In Step 1, the steps for constructing the fume detection model are as follows: Collect fume images generated during the cooking process; Annotate the fume images, divide the fume images into a fume part and a fume-free part, and make a fume detection data set; Input the fume detection data set into an artificial intelligence model for model training; In Step 4, when quantifying the fume and controlling the fume extractor's air volume according to the fume volume, if there is a fume part in the real-time collected fume image, the dark channel image of the fume part is extracted, and the pixel values of the dark channel image are filtered. Pixel values meeting the following conditions are identified as fume particles: ; then control the change of the fume extractor's air volume to ; ； Among them , is the number of oil fume particles.

5. An artificial intelligence-based pre-control system for a range hood, the range hood having an air suction opening capable of lifting, characterized in that, Including: A model training module that constructs a cookware recognition model based on artificial intelligence for identifying the type of cookware, a cooking action detection model for detecting cooking actions, and a fume detection model for detecting fumes; A cookware recognition control module that collects an image of the cookware when the range hood is started, identifies the type of cookware through the cookware recognition model, and adjusts the height of the air suction opening of the range hood according to the type of cookware: the greater the amount of fumes generated by the cookware, the lower the height of the air suction opening; among them, the cookware includes a wok, a frying pan, a flat pan, and a steamer, which are divided into the highest gear, the middle gear, and the lowest gear in ascending order of the height of the air suction opening of the range hood. When the cookware is a wok, the height of the air suction opening is the highest gear; when the cookware is a frying pan or a flat pan, the height of the air suction opening is the middle gear; when the cookware is a steamer, the height of the air suction opening is the lowest gear; The cooking action detection and control module sets the increased air volume value corresponding to each cooking action according to the amount of oil fume generated by different cooking actions. ; Periodically detect the cooking action through the cooking action detection model. When a cooking action is detected, set the air volume of the range hood. ; is the current air volume of the range hood, is the increased air volume value, is the maximum air volume of the range hood; where the cooking action refers to an action that will affect the amount of oil fume generated during cooking after the action is implemented during the cooking process; A fume quantification control module that, if no cooking action is detected after the time T when starting to detect cooking actions, detects fumes through the fume detection model, quantifies the fumes, and controls the air volume of the range hood according to the amount of fumes.

Images