Method
The method used for this experiment is outlined in the following steps:
1. Calibration of the experimental setup. Calibration is necessary to ensure that the images obtained are focussed and the scale factor necessary for processing of the images is obtained. Calibration can be done in a number of ways. The two ways considered during this experiment are: use of a calibration target and real-time analysis of the correlation vectors in the flow field analysed by the computer.
The first way makes use of a calibration target, given in Figure 2.
Figure 2 – Captured image of the calibration target
This calibration target is placed parallel to the light sheet and approximately in the middle. The calibration target consists or a grid of dots with a bigger central dot surrounded by four smaller dots. The distance between two dots on this calibration target is given as 5 mm. Images are captured of the calibration target when placed in the light sheet. These images are analysis by the computer and the scale factor is deduced by comparing the apparent distance between the dots provided by the perspective of the CCD camera and the actual distance of 5 mm. Before the images are captured, the camera is focussed such that all the dots appear sharp. This is confirmed by analysing the histogram of the greyscale image and ensuring that the signal peaks present are only black and white.
The second calibration method that can be employed is real-time analysis of the correlation vectors in the direction flow field processed by the computer. Typically noise signals are represented by chaotic and large vectors in this direction field. By focussing the camera, the volume of these vectors in the direction field can be minimised. When perfectly focussed theoretically there should not be any of these noise correlation vectors present. This however in practice is difficult to ensure when working under tight time constraints so a level of tolerance is adopted.
2. Measurement of the flow field. Once calibration of the setup is achieved, measurement can commence. The laser is switched on the time delay between pulses is set. The fog generator is switched on and the flow field expelled from the box is captured by the CCD camera. The settings used by the Dantec software are provided in the results. The computer ensures that the laser pulse and the camera capture are synchronised. The delay between successive captures is automatically controlled by the software. The images obtained are recorded on the computer system, ready for post-processing.
3. The final step necessary is post-processing. During this phase, the images captured during the measurement phase are analysed. Each pair of images taken are processed using cross-correlation. Using cross-correlation the images are broken down into smaller and smaller interrogation areas until the specified size is achieved (for example 32x32 px). Each interrogation area is processed using statistical correlation techniques whereby the displacement of the seeding particles is analysed and compared. This allows for the visualisation of the flow field. The resulting direction vector field is provided in the results.
Once the cross-correlation vectors are obtained, the next step necessary is the use of various filters to eliminate the “bad” vectors. These vectors are typically anomalies in the flow field resulting from low seeding particle concentration and general signal noise. The filters used are: peak filter, range filter, and moving average filter. The peak filter eliminates vectors by considering the ratio between the signal peak and the noise peaks. If this ratio drops below a certain value, then the vector is eliminated. The ratio used is given in the results. The range filter ensures that the magnitude of the vectors remains within a set range determined by qualitative analysis of the “good” vectors. The moving average filter filters out “bad” vectors by taking a sample set of vectors around one vector and calculating the mean. If the discrepancy between the actual vector and the average calculated by considering the surrounding vectors is greater than a given ratio, the central vector is either replaced by the mean or completely eliminated. Use of these three filters allows for a great deal of noise to be eliminated from the flow field, however in practice this is never perfect and a certain number of “bad” vectors persist. If necessary, these vectors can be removed manually, however this is a laborious task.
4. The above three phases are carried out for two different velocities of the fog generated.
MethodThe method used for this experiment is outlined in the following steps:1. Calibration of the experimental setup. Calibration is necessary to ensure that the images obtained are focussed and the scale factor necessary for processing of the images is obtained. Calibration can be done in a number of ways. The two ways considered during this experiment are: use of a calibration target and real-time analysis of the correlation vectors in the flow field analysed by the computer. The first way makes use of a calibration target, given in Figure 2. Figure 2 – Captured image of the calibration targetThis calibration target is placed parallel to the light sheet and approximately in the middle. The calibration target consists or a grid of dots with a bigger central dot surrounded by four smaller dots. The distance between two dots on this calibration target is given as 5 mm. Images are captured of the calibration target when placed in the light sheet. These images are analysis by the computer and the scale factor is deduced by comparing the apparent distance between the dots provided by the perspective of the CCD camera and the actual distance of 5 mm. Before the images are captured, the camera is focussed such that all the dots appear sharp. This is confirmed by analysing the histogram of the greyscale image and ensuring that the signal peaks present are only black and white. วิธีการปรับเทียบสองที่สามารถทำงานได้วิเคราะห์แบบเรียลไทม์ของเวกเตอร์ความสัมพันธ์ในฟิลด์ขั้นตอนทิศทางการประมวลผล โดยคอมพิวเตอร์ โดยปกติเสียงสัญญาณจะแทน ด้วยเวกเตอร์ขนาดใหญ่ และวุ่นวายในฟิลด์ทิศทางนี้ โดย focussing กล้อง ปริมาณเวกเตอร์เหล่านี้ในฟิลด์ทิศทางสามารถเป็นกระบวน เมื่อสมบูรณ์ focussed ครั้งแรกราคา ไม่ควรมีของเวกเตอร์ความสัมพันธ์ของเสียงเหล่านี้มีอยู่ อย่างไรก็ตามในทางปฏิบัติอยู่เมื่อทำงานภายใต้ข้อจำกัดเวลาแน่นเพื่อให้ถึงระดับยอมรับยาก2. วัดฟิลด์ขั้นตอน เมื่อทำการปรับเทียบการตั้งค่า การประเมินสามารถเริ่ม สลับเลเซอร์ในเวลา ตั้งหน่วงเวลากะพริบ สร้างหมอกเป็นสลับกับ และฟิลด์ไหลออกมาจากกล่องจับภาพจากกล้อง CCD มีการตั้งค่าซอฟต์แวร์ของ Dantec ในผลลัพธ์ คอมพิวเตอร์ช่วยให้มั่นใจว่า พัลส์เลเซอร์และกล้องจับอยู่เพื่อ ซอฟต์แวร์โดยอัตโนมัติควบคุมการหน่วงเวลาระหว่างจับต่อเนื่อง ภาพที่ได้ถูกบันทึกไว้ในระบบคอมพิวเตอร์ พร้อมสำหรับการประมวลผลหลัง 3.ขั้นตอนสุดท้ายที่จำเป็นถูกประมวลผล ระหว่างขั้นตอนนี้ มี analysed ภาพจับระหว่างขั้นตอนการประเมิน แต่ละคู่ของภาพถ่ายจะดำเนินการโดยใช้ความสัมพันธ์ระหว่าง ใช้ภาพจะแบ่งออกเป็นพื้นที่ขนาดเล็ก และขนาดเล็กการสอบปากคำจนได้ขนาดที่ระบุความสัมพันธ์ระหว่าง (เช่น 32 x 32 px) แต่ละพื้นที่มืดจะถูกประมวลผลโดยใช้เทคนิคทางสถิติสหสัมพันธ์โดยใช้ analysed และเปรียบเทียบปริมาณกระบอกสูบของอนุภาค seeding นี้ช่วยให้การสร้างมโนภาพของฟิลด์ลำดับ ฟิลด์เวกเตอร์ทิศทางผลลัพธ์ไว้ในผลลัพธ์ Once the cross-correlation vectors are obtained, the next step necessary is the use of various filters to eliminate the “bad” vectors. These vectors are typically anomalies in the flow field resulting from low seeding particle concentration and general signal noise. The filters used are: peak filter, range filter, and moving average filter. The peak filter eliminates vectors by considering the ratio between the signal peak and the noise peaks. If this ratio drops below a certain value, then the vector is eliminated. The ratio used is given in the results. The range filter ensures that the magnitude of the vectors remains within a set range determined by qualitative analysis of the “good” vectors. The moving average filter filters out “bad” vectors by taking a sample set of vectors around one vector and calculating the mean. If the discrepancy between the actual vector and the average calculated by considering the surrounding vectors is greater than a given ratio, the central vector is either replaced by the mean or completely eliminated. Use of these three filters allows for a great deal of noise to be eliminated from the flow field, however in practice this is never perfect and a certain number of “bad” vectors persist. If necessary, these vectors can be removed manually, however this is a laborious task.4. The above three phases are carried out for two different velocities of the fog generated.
การแปล กรุณารอสักครู่..
