選択できるのは25トピックまでです。 トピックは、先頭が英数字で、英数字とダッシュ('-')を使用した35文字以内のものにしてください。

image-tracker.ts 12KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403
  1. /*
  2. * encantar.js
  3. * GPU-accelerated Augmented Reality for the web
  4. * Copyright (C) 2022-2025 Alexandre Martins <alemartf(at)gmail.com>
  5. *
  6. * This program is free software: you can redistribute it and/or modify
  7. * it under the terms of the GNU Lesser General Public License as published
  8. * by the Free Software Foundation, either version 3 of the License, or
  9. * (at your option) any later version.
  10. *
  11. * This program is distributed in the hope that it will be useful,
  12. * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  14. * GNU Lesser General Public License for more details.
  15. *
  16. * You should have received a copy of the GNU Lesser General Public License
  17. * along with this program. If not, see <https://www.gnu.org/licenses/>.
  18. *
  19. * image-tracker.ts
  20. * Image Tracker
  21. */
  22. import Speedy from 'speedy-vision';
  23. import { SpeedySize } from 'speedy-vision/types/core/speedy-size';
  24. import { SpeedyMatrix } from 'speedy-vision/types/core/speedy-matrix';
  25. import { SpeedyMedia } from 'speedy-vision/types/core/speedy-media';
  26. import { SpeedyPipeline } from 'speedy-vision/types/core/pipeline/pipeline';
  27. import { SpeedyPromise } from 'speedy-vision/types/core/speedy-promise';
  28. import { SpeedyPipelineNode } from 'speedy-vision/types/core/pipeline/pipeline-node';
  29. import { SpeedyPipelineNodeImageSource } from 'speedy-vision/types/core/pipeline/nodes/images/source';
  30. import { SpeedyPipelineNodeResize } from 'speedy-vision/types/core/pipeline/nodes/transforms/resize';
  31. import { SpeedyPipelineNodeFASTKeypointDetector } from 'speedy-vision/types/core/pipeline/nodes/keypoints/detectors/fast';
  32. import { SpeedyKeypoint } from 'speedy-vision/types/core/speedy-keypoint';
  33. import { VideoSource } from '../../sources/video-source';
  34. import { CanvasSource } from '../../sources/canvas-source';
  35. import { Tracker, TrackerOutput, TrackerResult, Trackable, TrackerType, TrackerResultType } from '../tracker';
  36. import { Session } from '../../core/session';
  37. import { IllegalOperationError, IllegalArgumentError } from '../../utils/errors';
  38. import { Resolution } from '../../utils/resolution';
  39. import { ReferenceImage, ReferenceImageWithMedia } from './reference-image';
  40. import { ReferenceImageDatabase } from './reference-image-database';
  41. import { ImageTrackerState } from './states/state';
  42. import { ImageTrackerInitialState } from './states/initial';
  43. import { ImageTrackerTrainingState } from './states/training';
  44. import { ImageTrackerScanningState } from './states/scanning';
  45. import { ImageTrackerPreTrackingAState } from './states/pre-tracking-a';
  46. import { ImageTrackerPreTrackingBState } from './states/pre-tracking-b';
  47. import { ImageTrackerTrackingState } from './states/tracking';
  48. import { Nullable, Utils } from '../../utils/utils';
  49. import { AREventTarget } from '../../utils/ar-events';
  50. import { ImageTrackerEvent, ImageTrackerEventType } from './image-tracker-event';
  51. import { SpeedyPoint2 } from 'speedy-vision/types/core/speedy-point';
  52. import { Viewer } from '../../geometry/viewer';
  53. import { Pose } from '../../geometry/pose';
  54. import { CameraModel } from '../../geometry/camera-model';
  55. /** A trackable target */
  56. export interface TrackableImage extends Trackable
  57. {
  58. /** the pose of the target */
  59. readonly pose: Pose;
  60. /** the reference image linked to the target */
  61. readonly referenceImage: ReferenceImage;
  62. }
  63. /** Image Tracker result to be consumed by the user */
  64. export class ImageTrackerResult extends TrackerResult
  65. {
  66. /** tracker */
  67. readonly tracker: ImageTracker;
  68. /** trackable targets */
  69. readonly trackables: TrackableImage[];
  70. /** 3D virtual camera */
  71. readonly viewer: Viewer;
  72. /**
  73. * Constructor
  74. * @param tracker
  75. * @param trackables
  76. * @param viewer
  77. */
  78. constructor(tracker: ImageTracker, trackables: TrackableImage[], viewer: Viewer)
  79. {
  80. super();
  81. this.tracker = tracker;
  82. this.trackables = trackables;
  83. this.viewer = viewer;
  84. }
  85. }
  86. /** Image Tracker output */
  87. export interface ImageTrackerOutput extends TrackerOutput
  88. {
  89. /** tracker result to be consumed by the user */
  90. readonly exports?: ImageTrackerResult;
  91. /** keypoints found in this framestep */
  92. readonly keypoints?: SpeedyKeypoint[];
  93. /** optional keypoints for visualizing & testing */
  94. readonly keypointsNIS?: SpeedyKeypoint[];
  95. /** optional polyline for visualizing & testing */
  96. readonly polylineNDC?: SpeedyPoint2[];
  97. /** optional camera model for visualizing & testing */
  98. readonly camera?: CameraModel;
  99. }
  100. /** All possible states of an Image Tracker */
  101. export type ImageTrackerStateName = 'initial' | 'training' | 'scanning' | 'pre-tracking-a' | 'pre-tracking-b' | 'tracking';
  102. /** Options for instantiating an ImageTracker */
  103. export interface ImageTrackerOptions
  104. {
  105. /** resolution of the tracker; it helps define the AR screen space */
  106. resolution?: Resolution;
  107. }
  108. /** Default options for instantiating an ImageTracker */
  109. const DEFAULT_OPTIONS: Readonly<ImageTrackerOptions> = {
  110. resolution: 'sm'
  111. };
  112. /**
  113. * The ImageTracker tracks an image (one at a time)
  114. */
  115. export class ImageTracker extends AREventTarget<ImageTrackerEvent> implements Tracker
  116. {
  117. /** session */
  118. private _session: Nullable<Session>;
  119. /** source of data */
  120. private _source: Nullable<VideoSource | CanvasSource>;
  121. /** all states */
  122. private readonly _state: Record<ImageTrackerStateName, ImageTrackerState>;
  123. /** name of the active state */
  124. private _activeStateName: ImageTrackerStateName;
  125. /** last emitted output of the tracker */
  126. private _lastOutput: ImageTrackerOutput;
  127. /** reference image database */
  128. private readonly _database: ReferenceImageDatabase;
  129. /** the AR resolution size, used in GPU processing, defines the AR screen space */
  130. private _resolution: Resolution;
  131. /**
  132. * Constructor
  133. * @param options
  134. */
  135. constructor(options: ImageTrackerOptions)
  136. {
  137. super();
  138. // the states
  139. this._state = {
  140. 'initial': new ImageTrackerInitialState(this),
  141. 'training': new ImageTrackerTrainingState(this),
  142. 'scanning': new ImageTrackerScanningState(this),
  143. 'pre-tracking-a': new ImageTrackerPreTrackingAState(this),
  144. 'pre-tracking-b': new ImageTrackerPreTrackingBState(this),
  145. 'tracking': new ImageTrackerTrackingState(this),
  146. };
  147. // initial setup
  148. this._session = null;
  149. this._source = null;
  150. this._activeStateName = 'initial';
  151. this._lastOutput = { };
  152. this._database = new ReferenceImageDatabase();
  153. // user settings
  154. options = Object.assign({}, DEFAULT_OPTIONS, options);
  155. this._resolution = options.resolution!;
  156. }
  157. /**
  158. * The type of the tracker
  159. * @deprecated
  160. */
  161. get type(): keyof TrackerType
  162. {
  163. return 'image-tracker';
  164. }
  165. /**
  166. * Check if this tracker is of a certain type
  167. */
  168. is<T extends keyof TrackerType>(type: T): this is TrackerType[T]
  169. {
  170. return type === this.type;
  171. }
  172. /**
  173. * Current state name
  174. */
  175. get state(): ImageTrackerStateName
  176. {
  177. return this._activeStateName;
  178. }
  179. /**
  180. * Reference Image Database
  181. * Must be configured before training the tracker
  182. */
  183. get database(): ReferenceImageDatabase
  184. {
  185. return this._database;
  186. }
  187. /**
  188. * Resolution of the tracker
  189. */
  190. get resolution(): Resolution
  191. {
  192. return this._resolution;
  193. }
  194. /**
  195. * Resolution of the tracker
  196. * @readonly
  197. */
  198. set resolution(resolution: Resolution)
  199. {
  200. // this property is readonly, but this setter has been kept because
  201. // it wasn't readonly in previous versions of the engine. FIXME
  202. }
  203. /**
  204. * Size of the AR screen space, in pixels
  205. * @internal
  206. */
  207. get screenSize(): SpeedySize
  208. {
  209. return this._state[this._activeStateName].screenSize;
  210. }
  211. /**
  212. * Last emitted output
  213. * @internal
  214. */
  215. get _output(): ImageTrackerOutput
  216. {
  217. return this._lastOutput;
  218. }
  219. /**
  220. * Stats related to this tracker
  221. * @internal
  222. */
  223. get _stats(): string
  224. {
  225. const screenSize = this.screenSize;
  226. return `${screenSize.width}x${screenSize.height} ${this.state}`;
  227. }
  228. /**
  229. * Initialize this tracker
  230. * @param session
  231. * @returns promise that resolves after the tracker has been initialized
  232. * @internal
  233. */
  234. _init(session: Session): SpeedyPromise<void>
  235. {
  236. // store the session
  237. this._session = session;
  238. // find a suitable source of data
  239. // XXX also let the user specify a source manually?
  240. for(const source of session.sources) {
  241. // prefer video sources
  242. if(source._is('video')) {
  243. this._source = source;
  244. break;
  245. }
  246. else if(source._is('canvas'))
  247. this._source = source;
  248. }
  249. if(this._source === null)
  250. throw new IllegalOperationError('The image tracker requires a suitable source of data');
  251. // initialize states
  252. for(const state of Object.values(this._state))
  253. state.init();
  254. // done!
  255. return Speedy.Promise.resolve();
  256. }
  257. /**
  258. * Release this tracker
  259. * @returns promise that resolves after the tracker has been released
  260. * @internal
  261. */
  262. _release(): SpeedyPromise<void>
  263. {
  264. // release states
  265. for(const state of Object.values(this._state))
  266. state.release();
  267. // unlink session
  268. this._session = null;
  269. // done!
  270. return Speedy.Promise.resolve();
  271. }
  272. /**
  273. * Update the tracker
  274. * @returns promise
  275. * @internal
  276. */
  277. _update(): SpeedyPromise<void>
  278. {
  279. // validate
  280. if(this._session == null)
  281. return Speedy.Promise.reject(new IllegalOperationError(`Uninitialized tracker`));
  282. // compute the screen size for image processing purposes
  283. // note: this may change over time...!
  284. const media = this._source!._internalMedia;
  285. const screenSize = this._computeScreenSize();
  286. // run the active state
  287. const activeState = this._state[this._activeStateName];
  288. return activeState.update(media, screenSize).then(({ trackerOutput, nextState, nextStateSettings }) => {
  289. // update the output of the tracker
  290. this._lastOutput = trackerOutput;
  291. // need to change the state?
  292. if(this._activeStateName != nextState) {
  293. activeState.onLeaveState();
  294. this._activeStateName = nextState;
  295. this._state[nextState].onEnterState(nextStateSettings || {});
  296. }
  297. });
  298. }
  299. /**
  300. * Get reference image
  301. * @param keypointIndex -1 if not found
  302. * @returns reference image
  303. * @internal
  304. */
  305. _referenceImageOfKeypoint(keypointIndex: number): Nullable<ReferenceImageWithMedia>
  306. {
  307. const training = this._state.training as ImageTrackerTrainingState;
  308. return training.referenceImageOfKeypoint(keypointIndex);
  309. }
  310. /**
  311. * Get reference image index
  312. * @param keypointIndex -1 if not found
  313. * @returns reference image index, or -1 if not found
  314. * @internal
  315. */
  316. _referenceImageIndexOfKeypoint(keypointIndex: number): number
  317. {
  318. const training = this._state.training as ImageTrackerTrainingState;
  319. return training.referenceImageIndexOfKeypoint(keypointIndex);
  320. }
  321. /**
  322. * Get a keypoint of the trained set
  323. * @param keypointIndex
  324. * @returns a keypoint
  325. * @internal
  326. */
  327. _referenceKeypoint(keypointIndex: number): Nullable<SpeedyKeypoint>
  328. {
  329. const training = this._state.training as ImageTrackerTrainingState;
  330. return training.referenceKeypoint(keypointIndex);
  331. }
  332. /**
  333. * Compute the current size of the AR screen space
  334. * Note that this may change over time
  335. * @returns size
  336. */
  337. private _computeScreenSize(): SpeedySize
  338. {
  339. const media = this._source!._internalMedia;
  340. const aspectRatio = media.width / media.height;
  341. const screenSize = Utils.resolution(this._resolution, aspectRatio);
  342. return screenSize;
  343. }
  344. }