3.1

EL_DM6437 Da vinci video development platform

The DSP demo platform this college used is EL_DM6437 Da Vinci video development board. The EL_DM6437 Da Vinci video development board expands 64Mbytes DDR2SDRAM via a dedicated DDR2 storage controller interface, which brings high throughput, high capacity storage for video applications and the DDR2SDRAM can be extended to the maximum storage 256MByte. 256Mbit NorFlash guarantees a powerful capacity for offline video applications. The board connects with CPU by a video decoder chip. So that CPU can accept external composite video signal input and S signal input. In addition, the signal is automatically detected, which greatly improves video processing flexibility. CPU will also provide video DAC and can change the DDR2 memory data into users’ observable signal through a number of operational amplifiers. It is very convenient. Besides, the board also contains a lot of resources, including audio input and output, 10M/100M adaptive Ethernet interface, USB2.0 interface, real-time clock, 4 bit LED instructions and 4 bit DIP switch input, CAN bus and UART interface a series of functions else like this. The most important thing is that all external memory interface signals are connected to the expansion interface, so that the user can design the external sub-board to expand their own function ^[3-4].

The DSP experiment demo platform image process is as follows. At first, color CCD image collecting output video signal. After passing through the video decoder, we will get digital signal. Then run the algorithm program in the DSP chip to realize real-time image processing. In this paper, we adopt the TMS320DM6437 DSP chip from TI Company. The image is processed and restored into a video signal through a video encoder and then to observe the processing results through video monitor. Through demonstration and application in practice teaching process, students will further understand image processing theory and abstract concept. It will also enhance their learning interests.

3.2

Practice teaching based on EL_DM6437 platform

Take the following image edge detection operator as an example. The traditional image edge detection operators include Roberts, Sobel, Prewitt, Canny and Laplasse etc. When teachers speak the process about each kind of edge operator detecting the image edge. Students will first know the principle and importance of image edge detection. Then they will carry out the image edge detection experiment and analyze the effect on DSP experimental platform. Finally, they need to further study on edge detection theory combined with experiment and expand its application in each field.

(1)Analyze Sobel operator image edge extraction principle combined with experiment Image edge refers to that part where the image intensity changes most obviously. It mainly exists between object and object, object and background, region and region (including different colors). It is also an important basis for image analysis, such as image segmentation, texture feature and shape feature.

The basic idea of edge detection is to use the edge enhancement operator to highlight the local edge of the image, and then define the image edge intensity. And, to extract edge points by setting threshold. Right sobel gray value of each pixel in the image, up, down, left and right neighbors field of weight difference, with the largest neighborhood close. Therefore, the Sobel operator is defined as follows.

Then, using two kernels in Figure 1 as convolutions. One kernel has the biggest influence on usual vertical edge while the other kernel has the biggest influence on horizontal edge. The two convolution maximum is taken as pixel output value and the operation result is an edge image.

Figure 1.

Sobel edge operator

Properly take TH as threshold. If R (i, j) ≥ TH, then it is step edge point. Sobel operator uses pixels up, down, left, right gray weighted algorithm to detect edge according to extremum principle at the edge points.

(2) Sobel operator image extraction based on Da Vinci platform

This part mainly include the following steps. ➀ separately connect computer with emulator, EL_DM6437 with emulator. Using one video cable head to connect P3 or P5 socket on the board. The other end connects with AV input of the display and transfers to the AV mode. Using video cable to connect camera output and P9 socket on the board. Pay attention to set camera output standard to PAL. □ Electrify EL_DM6437 and camera, then turn on CCS and connect. Open the project EdgeDistillUseSobel.pjt. This project feature is, to carry Sobel edge extraction on camera captured images and output to the monitor. Download the program “out” file. The camera is aimed at capturing image site. □ Set the first breakpoint at the 423th line of video_test.c. Click Run, wait till the program stops at the breakpoint. At this time, you can see the image from the monitor as shown in figure 2. Continue to click Run. If succeed, the display will show edge extracted image as shown in Figure 3.

Figure 2.

Original Image

Figure 3.

Sobel operator edge detection results image

(3) Characteristic analysis of Sobel operator image edge detection

Compared with other operators, the Sobel operator has two distinct features. First, due to the introduction of the average factor, it has a smoothing effect on the random noise in the image. Second, since it is the difference between two rows or two columns, the elements on both sides of the edge have been enhanced to make the edges appear thick and bright.

However, the disadvantage is that the edge is coarse, the amount of calculation is increased while the noise resistance is good, and the false edge is detected. So the positioning accuracy is not high. Combined with the above analysis, students are required to be familiar with the DSP system operation process. Take edge detection as an example, students need to know edge detection procedures programming (assembly language or C language), link command file modification, new project creation, project Compile - Link - Load - Run. Finally, students need to analyze the performance of the algorithm by experiment effect. It also requires students to be familiar with other edge detection algorithms, to compare each algorithm by experiments and to form analogical learning habit.