ByLinky Johnson 2018-11-15 401
Before shooting, we agreed on the main tone of the film, which was to be reflected in < interchanges>--
1. The beauty of the structure and curve of an interchange
2. Stereoscopic and clearly defined picture impact
3. The contrast between complex interchange and orderly traffic
Instead of mindlessly photographing all the interchange in Shanghai in order to achieve such achievements as "family photo", we first observed the shape of each interchange in Shanghai on the satellite map, and estimated the characteristics of the night picture. We chose only those places that could make us feel "interesting". Among the many photographic concepts, one is the concept of choice, so is aerial photography.
In the recent two months, the weather in Shanghai has not been very awesome. Under the daytime light, the color of overpass is often mainly grey and black, which is not very beautiful. The night scene after lighting is not as good as that after lighting. So we kept all the footage shot at night.
The exact shooting time should also be taken into consideration. The overpass in Shanghai is basically stuck in traffic until 8PM, and some areas are even stuck after 9PM. We want to show a city that is moving, and the shooting of time-lapse material requires high-speed traffic, so there is no need to go too early to get in the traffic jam.
Of course, it's not too late to... It's too late and there's no car... an overpass without a car is dead at night. Therefore, the shooting time is around 8 to 10 PM.
After taking off, the UAV's course depends on the operation of the pilot and flies around the city. To ensure safety, the pilot needs to have a wide observation horizon, is able to see the UAV's position all the time, and ensure that the remote control signal is not blocked by obstacles. Therefore, the choice of take-off point is very important.
"INTERCHANGES" is shot in the vicinity of overpasses. They are large-scale reinforced concrete buildings. Although there is a lot of space at the bottom of the bridge, it is not open enough. So the GPS signals and remote control signals are easily blocked. So we generally choose to take off on the outer surface of the overpass or near the roof.
This work requires UAVs to fly above the traffic flow. We must be responsible for the safety of the drivers below. On the basis of ensuring that the machine is in good condition, we should not fly too close to the bridge deck, otherwise it will affect the normal driving of drivers and cause potential safety hazards. If you need some nearer-looking images, you can achieve similar effects by scaling the images in later editing.
The subjects of the team's shooting are mainly urban. In urban areas with dense population and high buildings, it is necessary to pay special attention to flight safety. Therefore, we have two levels of principles to decide whether to execute a route.
Including the size of the flight space; Whether it will fly to the place where it blocks sight or GPS signal position; Whether there are objects, such as high-voltage lines, signal towers, etc. that may interfere with the UAV; Whether the pilot's flight technology meets the safety requirements of this route. There will be no mistakes such as absenteeism or punching, etc., which will lead to an impact.
We will make an estimate of the picture in our minds and describe the picture taken on the route. If we feel that the picture will not be very good, then we tend to change the route, and there is no point in the unattractive picture. If you need to give the flyer more pressure to take a demanding route for a bad picture, it's even less necessary.
1. Record the return point before taking off;
2. The return altitude must be higher than the maximum nearby building;
3. Keep the UAV out of sight without being blocked by obstacles;
4.The sight of the pilot is mainly to observe the position of the UAV, if the observation position is about to be invisible, try to move the observation position;
Following these route planning principles, zero accident records have been maintained for nearly a year at the time of implementation of the project.
A typical example: We considered flying a lens that let the UAV move forward and rise from the gap between two parallel decks. It should be a picture that rises slowly from the pavement under the bridge to parallel to the traffic flow and eventually becomes a downward angle.
The pilot's skill level is sufficient to execute the route, ensuring that there will be no accidents due to operational errors. However, if you want to see the drone all the time, to ensure that the signal is not blocked, you have to take off from the gap directly under the bridge. The surrounding buildings will block a lot of GPS signals, and you have to fly in attitude mode, unable to record the return point. And it's a curved bridge, and there's still the risk of being blocked from view after the drone flies over it. It violates the safety principle, so it must be abandoned.
In order to ensure that our video can be edited with the highest image quality, we chose to match the X5R head camera with the Inspire 1 RAW to perform the shooting. Compared with the earlier eight-axis s 1000, the more integrated four-axis Inspire 1 can be easily carried by one person and easier to maintain.
The X5R camera brings about the transformation of the whole post-processing process, and puts forward higher requirements for the performance of the equipment. So the team's computer hardware was also upgraded.
Because the X5R footage has an average 1.7Gbps bitstream, the amount of data captured each time will be very large, so we purchased six mechanical hard drives with 2TB capacity, and combined with LSI's 9260 array card to form the RAID 5 disk array. Finally, we got the hard disk space around 10TB.
Compared with other RAID modes, RAID 5 has both access performance and security performance improvements. Compared with the sacrificial ones, RAID 5 is the most suitable solution for our needs, so it is our first choice. However, the boot speed is also slowed down a lot, because every boot has to re-read all the hard disk before entering the system. Then again, show how fast the era has passed, more than a few seconds can not be eaten.
Disk array cards have specialized processors and SSD buffers to cache disk data and transfer data over a wider PCI-E channel. The operation of the disk can be processed directly through the array card, so it does not need to consume CPU and system memory resources, and does not degrade the performance of the disk subsystem as well.
The material has been shot, and the equipment is ready. Then is the video post-processing and editing process.
The X5R camera takes video in raw format, and the raw format means:
1. It is original and uncompressed file. It records the most screen data;
2. More picture data means better picture quality, and more later space.
However, RAW video files are actually stored in raw format photos, each frame of the video is a photo. If this is a one-second, 25 frames per second RAW video, then its file is actually 25 consecutive raw photos, which we call a sequence of images.
The RAW format is a general term, like CR2, NEF, ARW, and so on, all of which is a suffix of RAW file, and there is a generic, RAW file format: DNG file.
The X5r camera stores a sequence of images in DNG format on its solid-state hard disk, but it can also be saved in a TF card, like an X5 camera, after an in-machine compression of mp4 or mov files. We chose to save the material in DNG+mp4 format.
1. Use After Effects (hereinafter referred to as AE) to import RAW image sequence directly, and can directly color the material in the software and other post-processing, render a single material and then use other software to edit. But because AE processes the image sequence, it is done as a picture, which means that it has to apply post-processing to dozens of hundreds of photos at the same time. The performance requirements of hardware devices are high.
After testing, it takes about 20 minutes for a 20-second video sequence to be exported by AE. And each material needs to be processed and rendered separately, which will greatly reduce the work efficiency.
2. Similar to method 1, Lightroom (hereinafter referred to as LR) is used for color mixing and other processing of image sequences. For photographers, LR is more familiar software than AE. Export JPG image sequence after processing. We know that JPG files are much smaller than RAW, and thus can improve some post-processing efficiency, but the picture quality is lost, and the efficiency is not much higher than method 1, which reduces the significance of RAW video.
These two methods, one is the common process of ordinary video processing, the other is the common process of late processing of delayed video. Either way, replacing RAW video becomes an inefficient and clumsy method.
Can't you just do the color matching and editing together?
Good idea, we want to. However, the clip software we use is Final Cut, which cannot be edited directly with DNG files. The software does not support DaVinci Resolve, which handles DNG sequences directly, is a powerful color palette tool with clipping capabilities. But cutting RAW video directly into it, because performance issues can be quite fluid and can greatly reduce efficiency, which is not conducive to fine editing.
The ideal way to use a smooth editing format is to first convert the raw video into a smooth clip format, which takes a lot of time and low efficiency.
The new thing we learned was the concept of "proxy files," a very mature and efficient post-production process. The basic idea is simple: the software doesn't support editing in DNG format, so I'll just use the format it supports.
In theory, we skip the color mixing step and start the editing process by using software like AE or LR or Da Vinci to convert RAW video to supported formats like mov or mp4 before importing the editing software.
So there are two ways.
The first is to import the sequence of pictures into Da Vinci first. Da Vinci will recognize them as video clips, but we don't do color matching operations in Da Vinci at this time. Instead, we use them to render the video material in small formats. As we all know, when the bit rate of video material decreases, the performance requirement is lower and the efficiency is improved. So we take the small document, that is, the proxy document, for editing. Rendering a set of small format material is much faster than rendering RAW material directly. If you are going to use Final Cut clips later, it is recommended to output ProRes-coded proxy files and H.2642-coded proxy files using Premiere.
The second method is faster.
Remember that X5R camera can save RAW videos and compressed MP4 videos at the same time? The MP4 file that is saved at the same time is exactly the same as the RAW file in addition to its poor quality. These MP4 files can facilitate us to observe the shooting effect directly in the early shooting. In the later process, since we edit the whole film first instead of dealing with single material, it is a natural proxy file.
In this way, we can skip the conversion process and just take the proxy files in the mp4 format and cut them, saving a lot of time that the conversion format consumes, which is a lot more efficient.
If you need a fine clip, it's best to render a wave of small format proxy files with a sequence of pictures.
As a result, editing becomes the first step in post-processing, where we select the material, put together the entire film, and add all the special effects except the toning and noise reduction in this step.
The first is the choice of background music. For the emotions we want to express, we tend to choose rhythmic rock music.
The music Glitter And Gold meets our requirements. Lively rhythm and a slightly vicissitude voice all contribute to the video effect.
In fact, "thoughts" is a very subjective thing. How to cut this film? In addition to abiding by some basic laws, it mainly depends on the author's personal preferences and feelings. After a simple selection of about 180 pieces of material, we classified them into categories. Splicing according to the accent rhythm of music. At the climax of music, use a lot of time-lapse material to sublimate emotions. The transition used some conventional similar editing techniques, using the visual inertia of human eyes, so that the picture switching becomes smooth.
INTERCHANGES has a single theme: it's all interchange, and it's all night. To avoid visual fatigue in our viewers, we tried to keep the video length at about 2 minutes. Some simple special effects, such as--
In fact, all of them are the basic special effects of Final Cut Pro. We use moving, zooming, inverting, mirroring and other effects, and also use a few kaleidoscope special effects. It produces the effect similar to the inception, so that the original single viaduct material can gain some sense of form and add some highlights to the picture. However, we are very careful about the use of special effects like kaleidoscope, which is not much in the film. Special effects are a double-edged sword, which can add highlights to the picture and affect our original picture expression.
This is a picture of a compressed, uncolored video, which obviously doesn't render to the final film. The import and export of XML project files is supported in versions after FinalCut Pro X 10.0.1, and Adobe Premiere software now supports the export of XML project files, a very important step in the workflow of proxy files.
Exporting an XML project means that we can move a clip project to other software for further processing, which completely and accurately records the project timeline, screen, and other information. We use DaVinci Resolve 12.5 Studio software (DaVinci) to do color mixing and noise reduction for video.
First, we introduce the image sequence of the raw video into Da Vinci's media pool, and da Vinci automatically treats the image sequence of each material as a video.
Then, we import the XML project to Leonardo Da Vinci, and its source material is the low-quality proxy files, and we modify it and uncheck "automatically import sourceclips into media pool" in the import dialog. Link it to our first imported raw video instead. So, after importing the XML project, Da Vinci is a complete video clip in raw format
However, this late process also requires material management. X5R's raw format file name and mp4 format file name are not one-to-one correspondence. So we need to manually change the filenames to make them consistent, so Da Vinci finds the material that corresponds to each other. After all, we don't want to waste our time looking for things, and good material management applies to all workflows.
Since all the material in this video is a night scene, a balance must be struck between sensitivity and other exposure parameters. To ensure early quality, we use no more than 800 iso, in order to further increase the amount of light to ensure normal exposure, to the maximum aperture, but the final material is often not exposed more than two files. At the same time, according to our experience, x5r photography material has a feature, that is, the night scene in the red and blue saturation, as well as contrast, are relatively high. In the digital age, the significance of the early shooting lies in the recording of information, and the image scene is saved in the form of data, while the meaning of the later stage is the accurate expression of the content. Therefore, in order to express our feelings accurately and achieve the best picture effect, we must adjust the video later.
After importing RAW video into Da Vinci, we can see that the blue part of the image is somewhat strange, which is a state similar to overflow, affecting the image effect.
After research, we found that the default color space is Rec.709. When we changed the color space and Gamma into BMD and put the LUT file prepared for X5R by Dajiang, the problem was solved.
After restoring the highlight parameters and adding some exposure, the picture becomes this effect, and the problem of the blue part is solved.
After solving the problem of blue part, we started to do color mixing. First increase the value of the shadow in Da Vinci to highlight the dark part of the picture, so that it shows more details. The color in the original material is warm, so we adjust the color temperature parameters to normalize the color temperature of the picture. In order to reflect the texture of the bridge, we also improved the details of the middle tone.
The original material has rich color performance, but too many colors will interfere with each other, give people a sense of chaos, affect the aesthetic feeling of the picture. In order to achieve a more uniform color effect, we adopt the method of "first subtract and then add". First of all, reduce the overall saturation of the picture, reduce the impact of unnecessary color. This is a simple and rough adjustment, and the color we need is "mistakenly damaged", so we also need to add some subjective color manually on this basis.
In INTERCHANGES, we also tried to imitate the colors of some film films, adding yellow in highlights and a little cyan in the shadow to achieve better color performance.
After previous toning, the image is very close to what we want, but it also brings more noise, and in order to ensure a good picture quality experience, noise reduction is still done with the powerful Da Vinci software.
There are two versions of the Da Vinci software, which can be downloaded directly from the website. One is the free Lite version, which is already capable of meeting most of the post-video requirements, and the other is the more powerful Studio version, a paid version that uses USB encrypted dogs for encryption.
Most of the functional differences between the two versions are small for the average user, but one of the most significant differences is that the studio version has powerful noise reduction tools, one of Da Vinci's house owners, cool techs. It can be very effective in making the picture purer. For the later quality of the film, we resolutely bought the studio version.
Noise reduction tests the performance of the computer, and processing in advance affects the experience of other post-processing operations, so we put the noise reduction at the last step before the final rendering.
Da Vinci's noise reduction tools, there are two ways: time domain noise reduction and airspace noise reduction.
Noise reduction is a process of achieving a balance between picture noise and picture details. Considering the impact on picture quality, we suggest using time domain noise reduction. However, if the effect of temporal noise reduction is not obvious, the spatial noise reduction can also be used appropriately, but it should be noted that the parameters of spatial noise reduction have a great impact on picture quality.
Since all the materials are shot by the same equipment and similar scenes, we can synchronize the parameters of one piece of material into other materials after processing, and then fine-tune it according to the situation of the scene, so that the color of the whole film tends to be consistent, which effectively improves our work efficiency.
By this point, all the post-processing is done and the final video can be rendered. Since video is primarily distributed across the network, we chose the most popular H.264 code to output the mov file.
Export raw format files and compressed mp4 files-Editing mp4 files with Final Cut Pro-Export XML files to Da Vinci-Replace with RAW files-Toning-Noise reduction-final output.
|You may also want to read:|
|【Tech News】 Xiaomi Mi Max 3 ID photo is exposed, with 5400mAh battery|
|【Tech News】Mobile phone backlight photo camera OPPO R15 hands-on experience|
|OPPO R17 Pro camera performance: ultra clear night shot lives up to expectations|
Aircraft aerial video clip tips