Search

CN-121999246-A - Intelligent image processing system based on multi-technology fusion

CN121999246ACN 121999246 ACN121999246 ACN 121999246ACN-121999246-A

Abstract

The invention discloses an intelligent image processing system based on multi-technology fusion, which relates to the technical field of artificial intelligence and image processing intersection, and discloses an intelligent image processing system based on multi-technology fusion and an application method thereof. The image acquisition unit adopts a multi-type sensor to adapt to different scenes, the intelligent preprocessing unit combines filtering, correction and color optimization algorithm to eliminate image interference, and the characteristic extraction and segmentation unit realizes accurate extraction of a target area based on deep learning and a traditional algorithm fusion technology. The system can be widely applied to the fields of industrial quality inspection, intelligent transportation, medical diagnosis, security monitoring and the like, greatly improves the image processing efficiency and accuracy, and reduces the manual dependence.

Inventors

  • GE DINGJIA

Assignees

  • 上海精鲲计算机科技有限公司

Dates

Publication Date
20260508
Application Date
20260130

Claims (10)

  1. 1. The intelligent image processing system based on multi-technology fusion is characterized by comprising an image acquisition unit, an intelligent preprocessing unit, a feature extraction and segmentation unit, a multi-model identification unit, a dynamic tracking and analysis unit and a result output and control unit which are respectively connected with the multi-model identification unit and the dynamic tracking and analysis unit in a data way, wherein the image acquisition unit is used for acquiring image data under different scenes and transmitting the image data to the intelligent preprocessing unit; the intelligent preprocessing unit performs noise suppression, geometric correction and color optimization processing on the image data and then sends the image data to the feature extraction and segmentation unit; The characteristic extraction and segmentation unit extracts target characteristics in the image, segments a target area and transmits the target characteristics to the multi-model identification unit; the multi-model identification unit carries out category judgment on the target area and synchronizes the identification result to the dynamic tracking and analysis unit and the result output and control unit; The dynamic tracking and analyzing unit captures and predicts the track of the moving target in real time; The result output and control unit stores and displays the identification result and the track data and outputs a control signal to external equipment.
  2. 2. The intelligent image processing system based on multi-technology fusion according to claim 1, wherein the image acquisition unit comprises at least one image acquisition sensor, and the image acquisition unit further comprises an interface adaptation module, wherein the interface adaptation module can switch a single-frame image acquisition mode or a continuous video acquisition mode according to scene requirements.
  3. 3. The intelligent image processing system based on multi-technology fusion according to claim 1, wherein the intelligent preprocessing unit comprises a noise suppression subunit, a geometric correction subunit and a color optimization subunit, the noise suppression subunit adopts a time domain filtering algorithm or a frequency domain filtering algorithm, the time domain filtering algorithm adopts mean filtering, median filtering or Gaussian filtering, and the frequency domain filtering algorithm is selected from low-pass filtering or band-pass filtering.
  4. 4. The intelligent image processing system based on multi-technology fusion according to claim 1, wherein the feature extraction and segmentation unit comprises a feature extraction subunit and a region segmentation subunit, and the region segmentation subunit adopts a mode of fusing a deep learning algorithm with a traditional segmentation algorithm.
  5. 5. The intelligent image processing system based on multi-technology fusion according to claim 1, wherein the multi-model recognition unit is internally provided with a model recognition model library, the model recognition model library comprises a statistical model recognition model, a syntactic model recognition model and a fuzzy model recognition model, and the statistical model recognition model is based on a probability statistical theory and realizes recognition by calculating probability distribution of target features.
  6. 6. The intelligent image processing system based on multi-technology fusion of claim 1, wherein the dynamic tracking and analysis unit comprises a visual switching subunit, a track capturing subunit, and a motion prediction subunit.
  7. 7. The intelligent image processing system based on multi-technology fusion according to claim 1, wherein the result output and control unit comprises a data storage subunit, a visual presentation subunit and a device control subunit.
  8. 8. The intelligent image processing system based on multi-technology fusion according to claim 1, further comprising a model training and optimizing unit, wherein the model training and optimizing unit is respectively in data connection with the feature extraction and segmentation unit and the multi-model recognition unit.
  9. 9. The intelligent image processing system based on multi-technology fusion according to claim 1, further comprising a remote monitoring and interaction unit in data connection with the result output and control unit.
  10. 10. The intelligent image processing system based on multi-technique fusion according to claim 1, wherein the feature extraction and segmentation unit extracts the impurity region in the cloth image by a Blob analysis algorithm.

Description

Intelligent image processing system based on multi-technology fusion Technical Field The invention relates to the technical field of artificial intelligence and image processing, in particular to an intelligent image processing system based on multi-technology fusion, which is suitable for various scenes needing high-precision and high-efficiency image detection and identification, such as industrial production, traffic management, medical diagnosis, safety protection and the like. Background With the popularization of the digitizing technology, the application of the image processing technology in various industries is becoming wider. The traditional image processing system mainly adopts a single algorithm or model, and has the following limitations that firstly, the image acquisition equipment has poor interface compatibility, is difficult to adapt to image acquisition requirements under different scenes, so that the system universality is low, secondly, the image preprocessing means is single, interference such as noise, geometric deformation, color deviation and the like cannot be effectively eliminated, the subsequent processing precision is influenced, thirdly, the target segmentation depends on artificial design features or the traditional algorithm, the segmentation effect on complex background and irregular targets is poor, fourthly, the target recognition model is fixed, targets with various forms and fuzzy features are difficult to deal with, the recognition robustness is insufficient, fifthly, the dynamic target tracking lacks stereoscopic vision support, the track capturing precision is low under the conditions of target shielding and rapid motion, and thirdly, the system functions are relatively independent, the linkage control and remote operation capability with external equipment are lacked, and the automation and intelligent degree are to be improved. In order to solve the problems, the invention provides an intelligent image processing system based on multi-technology fusion, which realizes dual promotion of image processing precision and efficiency and meets the application requirements of multiple scenes by integrating multi-type acquisition equipment, multi-algorithm preprocessing, a segmentation method of deep learning and the fusion of the traditional technology, a multi-model recognition mechanism, single-binocular switching tracking and full-flow data interaction and control. Disclosure of Invention The invention provides an intelligent image processing system based on multi-technology fusion, which comprises an image acquisition unit, an intelligent preprocessing unit, a feature extraction and segmentation unit, a multi-model identification unit, a dynamic tracking and analysis unit and a result output and control unit which are respectively connected with the multi-model identification unit and the dynamic tracking and analysis unit in a data mode, wherein the image acquisition unit is used for acquiring image data in different scenes and transmitting the image data to the intelligent preprocessing unit, the intelligent preprocessing unit performs noise suppression, geometric correction and color optimization processing on the image data and then transmits the image data to the feature extraction and segmentation unit, the feature extraction and segmentation unit extracts target features in an image and segments target regions and transmits the target features to the multi-model identification unit, the multi-model identification unit performs category judgment on the target regions and synchronizes identification results to the dynamic tracking and analysis unit and the result output and control unit, the dynamic tracking and analysis unit performs real-time capturing and prediction on tracks of moving targets, and the result output and control unit stores and displays identification results and track data and outputs control signals to external equipment. The image acquisition unit comprises at least one image acquisition sensor, and an interface adaptation module, wherein the interface adaptation module can switch a single-frame image acquisition mode or a continuous video acquisition mode according to scene requirements. The intelligent preprocessing unit comprises a noise suppression subunit, a geometric correction subunit and a color optimization subunit, wherein the noise suppression subunit adopts a time domain filtering algorithm or a frequency domain filtering algorithm, the time domain filtering algorithm is selected from mean filtering, median filtering or Gaussian filtering, and the frequency domain filtering algorithm is selected from low-pass filtering or band-pass filtering. Preferably, the feature extraction and segmentation unit comprises a feature extraction subunit and a region segmentation subunit, wherein the region segmentation subunit adopts a fusion mode of a deep learning algorithm and a traditional segmentation algorithm Preferably, the multi-model reco