Image processing method and system based on three-view-field cloud precipitation particle imaging

Abstract

The invention relates to the technical field of image processing, in particular to an image processing method and system based on three-view-field cloud precipitation particle imaging, and aims to realize high-precision three-dimensional space reconstruction of precipitation particle imaging images by comprehensively considering weather environment state information and combining an advanced image processing technology. By means of the method, the defects of a traditional processing method in the aspect of three-dimensional reconstruction are overcome, more comprehensive and deep data support is provided for meteorological research and forecast, and therefore the technical progress in the field of meteorological research is promoted.

Claims

1. An image processing method based on three-field cloud precipitation particle imaging, characterized in that: Applied to an image processing system, the method comprises: Determining weather environment status information of a target meteorological monitoring area, and determining a three-dimensional particle space reconstruction instruction set by the weather environment status information; Acquire a regional particle imaging set, wherein the regional particle imaging set includes a plurality of precipitation particle imaging images, different precipitation particle imaging images have corresponding particle classification labels, and the precipitation particle imaging images correspond to at least one weather environment state in the weather environment state information; Performing environmental state grouping according to the weather environment state pointed to by each precipitation particle imaging image, determining the environmental state set corresponding to each weather environment state, and all weather environment states in the same environmental state set are bound to the same precipitation particle imaging label, and the precipitation particle imaging label is used to distinguish the precipitation particle imaging image pointed to by the environmental state set; If there is a target environment state set whose pointed precipitation particle imaging image meets the three-dimensional model processing requirements, the three-dimensional particle space reconstruction instruction is used to perform three-dimensional particle space reconstruction on the target precipitation particle imaging image pointed to by the target environment state set.

2. The method according to claim 1, characterized in that The particle classification label includes at least: a particle attribute sub-label and a motion trajectory sub-label; the acquisition of the regional particle imaging set includes: acquiring the attribute mapping imaging image of each precipitation particle matching the particle attribute sub-label, and the motion trajectory imaging image of each precipitation particle matching the motion trajectory sub-label.

3. The method according to claim 2, characterized in that The step of grouping the environmental states according to the weather environmental states pointed to by the precipitation particle imaging images to determine the environmental state set corresponding to each weather environmental state includes: Environmental state clustering is performed based on the weather environment state pointed to by each precipitation particle attribute mapping imaging image and the weather environment state pointed to by each precipitation particle motion trajectory imaging image, and the environmental state set corresponding to each weather environment state is determined; wherein the weather environment states in the same environmental state set are all associated with the same precipitation particle attribute mapping imaging image and the same precipitation particle motion trajectory imaging image.

4. The method according to claim 3, characterized in that The target meteorological monitoring area has corresponding target climate change keywords; the environmental state grouping is performed according to the weather environment state pointed to by each precipitation particle attribute mapping imaging image and the weather environment state pointed to by each precipitation particle motion trajectory imaging image, and the environmental state set corresponding to each weather environment state is determined, including: Determine, based on the correlation vectors between each particle group category feature and each climate change keyword in each precipitation particle attribute mapping imaging image, a plurality of target particle group category features matching the target climate change keyword, and a target precipitation particle attribute mapping imaging image to which each target particle group category feature belongs; According to the motion trajectory features pointed to by each of the particle group category features in each of the precipitation particle motion trajectory imaging images, determine the target motion trajectory features pointed to by each of the target particle group category features, and the target precipitation particle motion trajectory imaging image to which the target motion trajectory features pointed to by each of the target particle group category features belong; According to the weather environment state pointed to by each target precipitation particle attribute mapping imaging image and the weather environment state pointed to by each target precipitation particle motion trajectory imaging image, the environment state grouping is performed to determine the environment state set corresponding to each weather environment state; Among them, the weather environment states in the same environment state set are all associated with the same target precipitation particle attribute mapping imaging image, and are associated with the same target precipitation particle motion trajectory imaging image.

5. The method according to any one of claims 2 to 4, characterized in that: The particle classification label also includes: a time series description sub-label; The acquiring of the regional particle imaging set further includes: acquiring a time series mapping image of each precipitation particle matching the time series description sub-tag; The method of clustering the environmental states according to the weather environmental states pointed to by each precipitation particle imaging image to determine the environmental state set corresponding to each weather environmental state includes: clustering the environmental states according to the weather environmental states pointed to by each precipitation particle attribute mapping imaging image, the weather environmental states pointed to by each precipitation particle motion trajectory imaging image, and the weather environmental states pointed to by each precipitation particle time series mapping image to determine the environmental state set corresponding to each weather environmental state; wherein the weather environmental states in the same environmental state set are all associated with the same precipitation particle attribute mapping imaging image, the same precipitation particle motion trajectory imaging image, and the same precipitation particle time series mapping image.

6. The method according to claim 1, characterized in that The three-dimensional particle space reconstruction indication includes a multi-view geometric three-dimensional reconstruction model, and the multi-view geometric three-dimensional reconstruction model is used to integrate the multi-view states of the weather environment state; The three-dimensional particle space reconstruction is performed on the target precipitation particle imaging image pointed to by the target environment state set through the three-dimensional particle space reconstruction indication, including: Performing multi-view state integration on the target weather environment state in the target environment state set by using the multi-view geometric three-dimensional reconstruction model; Based on the target weather environment state after the multi-view state integration, a three-dimensional particle space reconstruction is performed on the target precipitation particle imaging image associated with the target environment state set; Among them, the multi-view state integration of the target weather environment state in the target environment state set is performed through the multi-view geometric three-dimensional reconstruction model, including: determining the state description keywords of the target weather environment state, and determining the target multi-view geometric three-dimensional reconstruction model that matches the state description keywords; performing multi-view state integration of the target weather environment state in the target environment state set through the target multi-view geometric three-dimensional reconstruction model; the state description keywords at least include: the type of weather environment state and the number of timing description sub-tags; the number of timing description sub-tags is used to describe: the number of label categories of timing description sub-tags; the types of weather environment state include at least: rainfall state, rain and snow state and snow crystal state.

7. The method according to claim 6, characterized in that The step of performing multi-view state integration on the target weather environment state in the target environment state set by using the target multi-view geometric 3D reconstruction model includes: Grouping the target precipitation particle imaging images pointed to by the target environmental state set into a first precipitation particle imaging image and a second precipitation particle imaging image; Determine, according to the first precipitation particle imaging image, a target multi-mode cross-state label and a target temporal description sub-label corresponding to the target weather environment state; The target weather environment state in the target environment state set is subjected to multi-view state synthesis through the target multi-view geometric three-dimensional reconstruction model, the target multi-mode cross-state label and the target temporal description sub-label.

8. The method according to claim 7, characterized in that The state description keywords include: the rainfall state; the target weather environment state in the target environment state set is subjected to multi-view state synthesis through the target multi-view geometric three-dimensional reconstruction model, the target multi-mode cross-state label and the target time series description sub-label, including: generating a state update heat map of the target weather environment state in combination with the target multi-mode cross-state label and the target time series description sub-label; determining the trend jump characteristics of the state update heat map of the target weather environment state, and correcting the trend jump characteristics of the state update heat map of the target weather environment state to obtain the target weather environment state after the multi-view state synthesis; Wherein, the state description keywords include: the rain and snow state; the target weather environment state in the target environment state set is integrated with multiple views through the target multi-view geometric three-dimensional reconstruction model, the target multi-mode cross state label and the target time series description sub-label, including: based on the feature transformation rule corresponding to the state mapping model, the target multi-mode cross state label and the target time series description sub-label are transformed to obtain the target multi-mode cross state label after feature transformation and the target time series description sub-label after feature transformation, the target multi-mode cross state label after feature transformation and the target time series description sub-label after feature transformation point to the type of weather environment state which is the rainfall state; combining the target multi-mode cross state label after feature transformation and the target time series description sub-label after feature transformation, generating a state update heat map of the target weather environment state; determining the trend jump characteristics of the state update heat map of the target weather environment state, and correcting the trend jump characteristics of the state update heat map of the target weather environment state to obtain the target weather environment state after the multi-view state integration; Among them, the state description keywords include: the snow crystal state; the multi-view state synthesis of the target weather environment state in the target environment state set through the target multi-view geometric 3D reconstruction model, the target multi-mode cross-state label and the target timing description sub-label, including: based on different attention indicators, feature transformation of the target multi-mode cross-state label and the target timing description sub-label to obtain the target multi-mode cross-state label after feature transformation and the target timing description sub-label after feature transformation; the target multi-mode cross-state label after feature transformation and the target timing description sub-label after feature transformation The types of weather environment states pointed to by the sub-tags are all related to the weather environment states of rainfall states of different attention indicators; the target multi-mode cross-state label after the feature transformation and the target timing description sub-tag after the feature transformation are feature quantized to obtain the target multi-mode cross-state label of the quantized vector state and the target timing description sub-tag of the quantized vector state; the trend jump characteristics of the target weather environment state are corrected through the target multi-mode cross-state label of the quantized vector state and the target timing description sub-tag of the quantized vector state to obtain the target weather environment state after the multi-view state is integrated.

9. The method according to claim 7, characterized in that: The state description keyword includes a number of the temporal description sub-tags; the multi-view state synthesis of the target weather environment state in the target environment state set is performed through the target multi-view geometric three-dimensional reconstruction model, the target multi-mode cross state tag and the target temporal description sub-tag, including: generating error evaluation data through the first precipitation particle imaging image and the second precipitation particle imaging image; Determining the error weight of the error assessment data by an imaging offset discrimination rule; The target weather environment state in the first precipitation particle imaging image is updated using the error weight.

10. An image processing system, characterized in that: The method comprises at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the method according to any one of claims 1 to 9.

Images